Google Part I: Origins of Search. How the Best Business in Human History Happened (Audio)

All right, David. Last episode we’re doing in our studios before Radio City.

Oh, that’s right. How do you feel?

We’re about to go from the stage of one, very, very small audience of one, to the very, very big stage where, if we make a mistake, no one will notice.

Yeah, and we just re-record it, and it’s like it never happened. We should try that at Radio City. Just be like, “Ah, strike that. All right, let’s take that again.”

Yeah. Hey, this is authentically Acquired, you guys. This is how we do it. You’re getting a look at the inside.

Probably not, though. All right, let’s do it.

Artificial intelligence is the story of our time. It is definitively the next trillion-dollar technology wave after PCs, the internet, and mobile. And to understand AI, you have to understand the company most responsible for its technical foundation and the wave that came before it: Google.

This episode begins our multipart Google saga. Finally, as I’m sure many of you out there are saying right now, Google has been the front door to the entire internet for 25 years now—a quarter century. But it wasn’t always this way.

No, it was not.

Back in 1998, when Google was founded, there were a dozen other search engines that already existed. And there were a variety of different business models, most of which were not very interesting.

Yeah. None of which were very interesting.

So today we will try to answer the question: Why did Google work? And once it did, how did it go from clever technology and a nice product to the single greatest business of all time?

I’m not being facetious, listeners. Google—and I should say Alphabet today—generates more net income, or profit, than any other U.S. company: more than Apple, Microsoft, ExxonMobil, JPMorgan Chase, and Berkshire Hathaway. This is a cash gusher. It is A, super-high gross margin; B, in a giant market; and C, according to the U.S. government as of today, a monopoly in that market, with 90% market share. Those are 3 enormous numbers multiplied together to create that “most profitable company in the U.S.” stat that I threw out earlier.

Well, I’m glad we don’t need to get to the government and all that until much, much later in our series. But, yeah, this is the creation of the most beautiful business of all time.

And Google’s market position has been seemingly unassailable, at least until really this year, until the AI wars really heated up. So, of course, it was that clean user experience that everyone talks about, with just a search box on the homepage, and the focus on the users and the high-quality, fast search that spread virally through word of mouth.

But a good product is far from the only reason that Google became dominant. So today we’ll tell the story of why, while Google was nowhere near the first search engine, it was the last.

Well, Microsoft may take a little issue with that with Bing, but only a little issue.

A few percentage points of issue.

A few percentage points of issue. Yeah.

David Rosenthal, when is the first time any human being searched for anything ever?

I have no idea. That’s a great question. Not that far back. Where do we start?

1. Google Founders Take Shape

Yeah. Well, you know what? We should ask Google. But to do that, we need to tell the story of Google. And that story starts in March of 1973 in Lansing, Michigan, where Larry Page is born as the second child and second son to Carl and Gloria Page.

Larry, of course, grows up in Lansing because his dad, Carl Page Sr., is a professor of computer science in nearby East Lansing at Michigan State University.

Now, before my dad, who went to MSU and is an MSU alum, gets too excited here, I regret to inform him, my dad, and you, Ben, that Carl got his PhD from Michigan.

I’m sorry about that.

And he would send his son there as well.

Yeah. Both of his sons.

Unfortunately, Larry’s mom also went to Michigan, also got a CS degree there, and also teaches programming as a programming instructor at MSU.

Pretty Michigan-heavy.

Pretty Michigan-heavy, but a pretty amazing childhood in the early to mid-’70s here for Larry and his older brother. It may not have been unique—I’m sure there were a few other households in America, in the world, that grew up with both of their parents steeped in computers as computer science professors—but it was really pretty unique.

Incredibly unique. Are you kidding me? Larry Page grew up with 2 computer science academics as parents in the ’70s, which would have meant that his parents would have needed to start in the ’50s. Very, very few households were like that, right at the same time as the PC era was coming online and Microsoft was emerging. To just have that be your daily existence growing up as a kid—how incredible is that?

Amazing. So, even more so, for the 1979–1980 academic year, when Larry is 6 and 7 years old, his dad does a sabbatical year at Stanford. So the whole family goes out and lives in Palo Alto and Stanford, in early Silicon Valley, and that makes a big impression on young Larry.

Then he continues to be influenced by this because I mentioned his older brother, Carl Jr., who’s 9 years older than him. Carl goes to Michigan, majors in CS just like Larry would, and then, when he graduates, he goes out to the West Coast—actually to the Pacific Northwest—and he works fairly early at Microsoft and then Mentor Graphics down in Oregon. Carl Jr. would ultimately also come down to Silicon Valley and play a little role in this story, as we will see in a little bit.

But back to Larry here. Part of the reason I say all this—and I want to include Sergey, too, in what I’m about to say, even though we haven’t introduced him yet in the story—is that I think there’s this perception today that Larry and Sergey were these bumbling academic guys who weren’t really business-minded, and Google was a research project, and this all sort of happened by accident: that they built the best business of all time.

Absolutely freaking not. We want to dispel that notion right now. You know, one of the things we heard over and over again talking to people in research is how hugely ambitious the 2 of them were, not just for the products they were building, but for Google, for the business.

Larry’s a different generation and a very different personality than Mark Zuckerberg, but you should think about his ambition and his desire to build a huge, world-changing company at the same level as him.

Yes. To your point, Google did not happen by accident. Or, another reasonable comparison: the generation before, Bill Gates. I think publicly, Larry and Sergey don’t get the same sort of ethos that those 2 get, but the same fire was there.

Larry would say later—this is a quote from him, probably when he was 12—“I knew I was going to start a company eventually. I wanted to make the world better, and in order to do that, you need to do more than just invent things.” And he would say another time later, “You need to use business and entrepreneurship to make these things real. It’s not enough just to invent them.”

And he’s alluding to 2 things there. One is, companies are the vehicles by which you bring ideas to the masses. And 2 is, in a capitalist society, a company is the vehicle that can accumulate profits, which then you can reinvest to build something of great scale and ambition.

Yes. And Larry got this all along, and Sergey did too.

Okay, so as we all know, Larry goes to Michigan for undergrad. He graduates in 1995 and then goes off to Stanford to get his PhD in computer science, where he has a fateful meeting with his business partner, friend, soulmate, Sergey Brin.

And isn’t the way this all went down that Larry was visiting and Sergey was already in the program?

I think Sergey was leading some kind of tour to try to sell Larry on joining the program. And Larry, even though he’s the new guy there, is challenging Sergey at every little corner. He’s bringing up, “Wouldn’t a better system be this?” And they’re talking about cities and transportation and civic design.

And they're almost bickering back and forth in this verbal sparring over who's smarter, even though they just met. That's sort of the story that I read, and I don't know—what did you and I read, probably 6 or 7 books between us, on the history of Google?

Yeah. Many, many versions of this story are out there.

As I was doing the research, I talked to one of our good friends, Anna Patterson, who's been in the Google orbit for a very long time. She was an early employee, left, started a company, Google reacquired it, and she was a VP of engineering there for a long time. She told me another little bit of this story that has, I think, never been told publicly before.

Who?

Well, it turns out Anna was a postdoc at Stanford at this time, and she's the one who organized this new students' weekend. She organized Larry and Sergey meeting, and she told me that the first night of the weekend—the first event, the first time that they actually met—was at drinks at the British Bankers Club in Menlo Park. This sort of magical friendship, bickering back and forth but a real partnership, was started there, and it was already going that night.

Larry and Sergey ended up shutting the bar down that night, and another famous local Stanford alum picked up the tab for the group. Can you guess who that person was?

I don't know. Lay it on me.

Charles Schwab.

No way.

Yep. He apparently lived locally in the area, went to the British Bankers Club all the time, and he'd do this. He'd just see Stanford students there and be like, "All right, I got you guys."

That's awesome.

So Charles Schwab funded the first date of Larry and Sergey.

Yeah. Amazing.

And Charles Schwab accounts would go on to hold billions and billions and billions of dollars' worth of Google stock because of that. Amazing. But yes, I think the spirit of all these stories is true. It was an instant, electric friendship and partnership between the 2 of them, and that would carry on forever.

I mean, they shared an office at Google. I don't think we've really covered founders before who were true partners in the way that Larry and Sergey are. It reminds me a little bit of you and me, on a very different scale.

Yes, on a very different scale. But that's interesting: equal co-founders. I'm racking my brain. Maybe Bill Gates and Paul Allen, but even then, it became very clear very quickly that Bill Gates was the guy.

Yeah, that was reflected in their equity ownership.

Exactly. I'm sure we've covered other companies where there was equal equity ownership among founders, but where it was a real, true partnership. One plus one equaled like 100. It's fascinating.

Yeah, maybe Jensen, Curtis, and Chris at NVIDIA, but over time, it kind of became Jensen.

Yeah, you're right. This is sort of unique for us to be covering true founder partners who made it decades together. I mean, even Warren and Charlie—Charlie was never full-time at Berkshire Hathaway, right? And certainly owned way less.

The closest I can think of is maybe Capital Cities with Tom Murphy and Dan Burke.

Yeah, it's really interesting.

Okay, so Sergey—what's his story?

Sergey was also born in 1973, a few months later in August, in some place even colder than Michigan: Moscow, which then, of course, was part of the Soviet Union. Sergey's family was Jewish. The Soviet Union was not exactly a great time and place to be Jewish, or probably to really be anything there.

His family lived in a 3-room apartment in Moscow, in what I assume was state-allocated housing. They shared it with his paternal grandmother, but his dad was an extremely talented mathematician. When Sergey was 4 years old, his dad attended an international mathematics conference. This was in 1977 or 1978, and he realized, "Oh, I've got to get my family to the West. We've got to get out of here."

It took him 2 years to be able to immigrate out of the Soviet Union, but they eventually came to the United States. His dad became a math professor at the University of Maryland. His mom became a researcher at NASA's Goddard Space Flight Center.

Okay. So Larry's parents are both computer science professors. Sergey's dad is a math professor, and his mom works for NASA. The pool is small of people with backgrounds like that in the '80s.

Yes. As you would imagine, Sergey is very precocious. He graduates high school at age 16, goes to the University of Maryland, gets his undergraduate degree in both math and computer science in 3 years, graduates at age 19, and then, of course, gets into Stanford for his PhD.

Do you know what he did in the summer before coming to Stanford?

Ooh, I have no idea.

He interned at Wolfram Research.

No way. Really?

Stephen Wolfram.

Oh, shout-out to friend of the show, Stephen.

I know, I know. The developers of Mathematica and Wolfram Alpha, I guess, eventually another search engine.

So yes, Sergey was every bit Larry's intellectual equal, every bit his sparring partner, and maybe a little more zany, too. More of a love for rollerblading than Larry, let's put it that way.

Yes. And frankly, it kind of makes sense now that you're starting to get the picture of these 2 over time. In the far future, closer to today, Sergey is the one doing stuff like Google Glass and skydiving videos. The zany parts of Google are sort of Sergey's DNA, and the really honed products—the products that make the business work—are a little bit more Larry's DNA. But to be honest, there's incredible overlap everywhere between the 2 of them.

2. PageRank Changes Search

Yeah. Okay. So, fall of 1995: Larry arrives at Stanford. Sergey's already there. Terry Winograd is his PhD adviser, and Larry and Sergey are already building this great friendship. That academic year, Larry presents a dissertation topic to Terry, his adviser, in collaboration with Sergey.

The idea they have is that this World Wide Web thing seems to be becoming a thing. Here we're in 1995. We're, what, a year after Netscape was started, a couple of years after Mosaic. I have a stat for you on this era of the internet.

Oh, yeah. Lay it on me.

This is from John Battelle's book The Search, which is excellent. From 1993 to 1996, the web grew from 130 sites to more than 600,000. If you compute that rate of growth over that 4-year period, it is 723% growth year over year for 4 years. That is exponential.

Yeah, this is the version of the Jeff Bezos realization where he's like, "I've got to leave D. E. Shaw and I've got to build Amazon." Nothing like this has ever happened before, right? The internet is a phenomenon like no other. If it keeps going, I mean, it's amazing for something to grow 700% year over year at all, but it happens. For that to keep happening year over year over year for half a decade, that is worth quitting your job, dropping everything, changing your whole life.

Yep. And just a couple of years before, 2 other PhD students at Stanford had started this thing called Yahoo. So I think perhaps inspired by what Yahoo was doing, Larry proposes that they're going to work on this idea of a system that will allow people to make annotations and notes directly on websites instead of in a centralized directory like Yahoo, because Yahoo was human-hand-curated commentary—a directory of websites.

The idea is, "Oh, this could be a decentralized annotation system where anybody can say what's interesting about a website."

Oh, that's funny. I thought I knew the whole history of Google. I somehow missed this.

So Terry, Larry's adviser, is like, "Okay, I like the problem space, shall we say, of the web for your dissertation here, Larry, but why don't you go refine this idea a little more and come back to me?"

Larry goes off, and, of course, he's collaborating with Sergey on this. As they think about it, they realize that there's sort of a fundamental flaw in what they were planning for this annotation system: for a big site like, say, The New York Times, it's just going to get overrun with hundreds, thousands, or millions of users commenting. You need a way to separate the wheat from the chaff, so to speak, of the comments. You need to be able to have the good ones rise to the top. You need a way to rank them, you might say.

And so Larry has a quote here:

"It wasn't that we intended to build a search engine. We built a ranking system to deal with annotations. We wanted to annotate the web. Build a system so that after you'd viewed a page, you could click and see what smart comments other people had about it. But how do you decide who gets to annotate a big site like Yahoo? We needed to figure out how to choose which annotations people should look at, which meant we needed to figure out which other sites contained comments that we should classify as authoritative. Hence, PageRank."

Ah, yes. We should say PageRank. It is ironic that the things that were being ranked were web pages, because the actual "Page" in PageRank is named for Larry Page, not for the pages they would rank.

Right, exactly. So Larry goes back to Terry and he's like, "Okay, this ranking idea seems like a really interesting computer science problem. The annotation thing seems messy. Why don't you just focus on rankings?"

Larry goes back and ultimately has the breakthrough leap: "Oh, we should apply rankings to web pages themselves." Larry says, "Wow, the big problem here is not annotation. We should use it not for ranking annotations but for ranking searches." Ding, ding, ding, ding, ding. And thus, at least the germ of the idea for PageRank as we all know it today is born.

So essentially, the mechanics of what the idea is are to try to rank websites based on how authoritative they are, based on how credible they are. And this is something that had been done somewhere before the web, very close to home for all these Stanford folks: academia.

Academia, of course.

Yes. How important is a research paper? Well, that depends on how many other people cited the research paper. And in particular, not just how many raw numbers of other papers cite a research paper—how many important papers cite your research paper? If you're in an important journal, what do those papers cite? And this becomes the inspiration for how they're going to do the ranking of web pages.

This was actually a research field before the web: the study of academic citations. There was already a body of work around how to do this.

Oh, interesting. I didn't know that. It makes sense. It's like how Hollywood loves making movies about Hollywood; academia loves doing papers about papers.

Yeah. So there are some examples to look at of how one might use references or citations to weight importance. We're almost all the way there to the huge leap that would become PageRank, then BackRub, and ultimately Google. But there's still one missing piece: they've got the theory of how to do this, but what's a citation on the web?

Well, they realize it's a link. A hyperlink is the exact same model as an academic citation. And not only is it the same as a citation, it's even better because there's metadata embedded within the link, which is the anchor text. Anytime you click a link, anybody who's creating a link can make anchor text for it. If you're writing, you can write whatever you want, and then just hit Command-K on your keyboard and make that text into a link. That's pretty easy to identify as metadata on an HTML page. So you get not only a citation of the link, but a few words of what the author of that link thought about it.

Yes. When someone is linking to you, they often do a better job of describing your website than you do on the page itself. Anybody who's just looking at your website to try to figure out what it's about, the actual words on your website tend not to do as good a job as everyone who links to you in aggregate. What words did they use to describe your website?

This whole thing is a genius idea, and we're going to talk about all the work they had to do to implement it. It basically works right away. The notion is, "Hey, what is the output if we try to create a system that ranks all websites for authoritativeness based on how many other reputable websites are linking to it?" Then later on, they can use the anchor text. But right now, just this ranking system—it spits out a list that's sorted exactly as you would hope: the most authoritative websites first, and all the crap all the way at the bottom.

Terry's like, "Yeah, do this for your dissertation." "Great. Let's do this project." Reputation on the web—that's going to be valuable.

So there's one more thing to making this brilliant PageRank idea work, which is a little problem: the way the web is architected. Any given web page only shows outgoing links. There's no way to query a page and say, "Who links to me?" You can only query a page and say, "Who do you link to?"

Right. It's like, it's easy for me to answer the question, "Who's in my phone book in my phone today?" It's hard for me to answer the question, "Whose phone books am I in?"

Exactly. And so the only way you could figure that out is if you somehow went and got a copy of everybody's phone book, right? Then you could backtrace all the links. Well, that's what they do. Google could not have been built at any other time in history.

Yes, because the web was actually small enough that you could go suck it all up then.

Exactly. It was small enough that, as a research project, it was not totally insane.

Just a little bit insane.

It's still pretty insane to say, "Oh, I'm going to go crawl the entire internet, make a copy of every web page out there, store it in something—which we'll get to—and then trace back all the links, and then reverse-compute all the links to answer that one seemingly simple question of what web pages link to me?"

If you had tried to undertake this as a brand-new project even just a year or a couple of years later, it would have been impossible. The web would have already gotten so big that to just start de novo and create an indexed copy like this—a full copy, even one year later—would have been tens of millions of dollars, and within a couple of years would have been hundreds of millions of dollars.

Yeah. Prohibitively expensive. So we're in what, 1996, 1997 here?

Yep. We're in 1996, kind of the back half of that first academic year of Larry's at Stanford.

3. BackRub Becomes Google

Great. So with Terry encouraging Larry and Sergey to go forth and undertake this ambitious project, they set up a page on the Stanford internet. They decide they're going to call the project BackRub, since it uses backlinks for ranking web pages. So they spin up backrub.stanford.edu, and Larry writes the first implementation of PageRank and the crawler to go do this. He writes it in Java, and it's super buggy and basically doesn't work. So they ask one of their friends there at Stanford, a guy named Scott Hassan, to help them. Scott's a better coder than Larry and Sergey are, so he codes it up in Python, and it actually sort of works.

Now, Scott is not an employee of Google and never would become an employee of Google, because it's not a company yet. It's a research project, right? They hadn't even come up with the name Google.

I mean, nothing about this is a search engine.

Exactly. So if you go to that project homepage, you can find cached versions of this on the internet, or images. I think there's even a recreation of it out there.

Yes, with the really weird black-and-white picture of a back rub with the red text over it.

I think that picture actually was on backrub.stanford.edu.

It looks like somebody's back. They certainly didn't search Google Images for it.

Exactly. The text on the page says, "BackRub is a 'web crawler' which is designed to traverse the web. Currently, we are developing techniques to improve web search engines."

So, yeah, Ben, they're not thinking of BackRub as a search engine yet. They're thinking of BackRub as just an implementation of a crawler and the PageRank algorithm.

Oh, that's interesting.

But they have the insight that this method of ranking, if it turns out to be better, could contribute to better search engines.

Yep.

Because again, even though Larry really wants to start a company, he's thinking he's going to get his PhD and then go forth here. Kind of like Mark Zuckerberg got years into Facebook before realizing, "Oh, this is my company." So Larry and Sergey are building BackRub here.

I mentioned Scott Hassan, their friend who helps code it up in Python, and he doesn't end up joining Google.

What happened to Scott?

Scott leaves while all this is going on and starts a company, because it's 1996, 1997. You're here in Silicon Valley, right? The bubble is inflating. What do you do? You go start a company. It's your obligation to go smash this piñata.

So Scott leaves and starts a company called eGroups. That company a couple of years later ends up getting acquired by Yahoo and becomes Yahoo Groups for about $400 million.

Do you know who co-founded eGroups with Scott?

Man, you're stumping me today. No.

Larry's older brother, Carl Page.

Oh, that's this company. Wow.

Yes. So here we are: we've got Larry and Sergey doing this research project to improve search engines. Meanwhile, their buddy who helped code that and Larry's older brother just went and started a company, and then they raised money. Sequoia and Mike Moritz would end up funding eGroups, and then they sold it for $400 million to the leading web company at the time.

That seems like a good idea.

Yeah. Larry and Sergey start thinking, "Ooh, wait, maybe we should do something commercial around this. This seems like it would have value."

So that leads them, in the spring of 1997 before school is out, to start shopping this BackRub technology around to the other existing search engines at the time. They're not yet thinking that this could be a company. They're thinking, "We're going to sell this technology to another search engine. They're going to pay us a lot of money for it. We'll go help implement it, and then we'll go back and finish our PhDs."

Because they effectively have it working at this point, even though it's not hardened. They have a crawler that has run on a small number of websites. They have a very modest index that has been created. It's not all efficient and everything, but they sort of have the proof of, "Hey, look, this ranking is actually a good ranking of how authoritative these websites are."

But other than some demo proof-of-concepts, they haven't yet built the consumer version of, "Hey, you can go query this thing and anybody can use it." So they're shopping it around that spring, that summer. They get a bunch of meetings. They meet with Infoseek, Lycos, and all the existing search engines.

And we should say there was a large list of other search engines, portals, internet properties, and search-engine-like things that existed and had traffic.

Yes. Archie, Gopher, AltaVista, HotBot, Inktomi, Lycos, Yahoo, Excite, Infoseek—the list goes on and on and on.

So the closest they get that spring and summer is with Excite. The story is amazing. Supposedly—this is according to In the Plex, Steven Levy's book, a great book that we used as a source for the episode—they end up getting a meeting with Vinod Khosla, a legendary founder of Sun Microsystems. By this point in time, he's one of the top VCs in the Valley. He's at Kleiner Perkins alongside John Doerr. The 2 of them are running the firm, and Vinod is on the board of Excite.

And so somehow Larry and Sergey—and I think they bring Scott along—get a meeting with Vinod, and they hammer out a deal that Excite is going to license this BackRub search technology from the 2 of them for about $1 million. Part of that's in cash, and part of that's in Excite stock. Larry and Sergey are going to come work at Excite that summer, implement BackRub for their search—basically make Excite into Google—and then they're going to leave and go back to Stanford in the fall.

They get so far that they run a side-by-side test of Excite's search results, the original algorithm, and then the BackRub algorithm. The legend goes that they're demoing this test to Excite's CEO as a final step to finalizing this deal.

And the results are so relevant with BackRub. You get exactly what you search for, exactly what you want. It’s right there: you click, and you go to it. The usual Excite search is bad. You have to click around; you go forward, you come back. You spend a lot of time on the site.

And the CEO is like, “Why on earth would we move to your algorithm? I want people to stay on my site. I make money when people stay on my site. I don’t want them to leave my site. You guys are crazy. Get out of here. I’m killing the whole deal.”

It is amazing that as early as 1997, this very important conflict of interests is teased out. This is basically why Google beat Yahoo. I mean, there’s a lot more to it, but the portals all had this mentality of, “We want to build more and more and keep people on our site, in our ecosystem, continuing to look at our banner ads.” Google, from this point basically forever, was about how quickly they could deliver someone something relevant so that person could leave Google and have a good experience finding what they really wanted.

Now, all this is laughable in hindsight but made total sense at the time, because what was the business paradigm for all these sites? What was the model? It was banner ads. It was CPM, cost per thousand views. You wanted page views and impressions on your site. What BackRub was doing was kneecapping your page views, right? They were dramatically reducing the number of page views and allocating them to other properties on the web so people could leave your property.

This deal was never going to happen with Excite or anybody else because it broke the business model, right? It was not strategic for them. It was a conflict of interest to implement this type of better ranking, better search, faster technology.

So, back to the lab, right? No deal.

Yeah, exactly. That fall, 1997, Larry and Sergey go back to school. All these get-rich-quick deals have fallen apart. Nobody wants BackRub. Larry’s just like, “Well, fuck it. All right. This is how I believe search should work. We’re going to build this thing ourselves here at Stanford.”

Quote from him: “We couldn’t get anyone interested in buying BackRub. We did get offers, but they weren’t for much money. So we said, whatever. We went back to Stanford to work on it some more. These companies weren’t going to focus on search. They were becoming portals, i.e., they wanted eyeballs and page views. They didn’t understand search, and they weren’t technology people.”

Yeah.

And when he says they talked to others, they tried to sell PageRank to Yahoo for $1 million and were rejected. There are going to be chapters of this story that you can mark by the different times that Yahoo discussed buying Google and didn’t. But this is the very first.

They showed the technology to Infoseek. That also didn’t happen. Infoseek was bought by Disney and later shut down. They showed the technology to Lycos. They were shopping it all around town.

So they get back to campus. They’re like, “All right, we’re going to build a real search engine ourselves. First order of business: the name BackRub. You’re probably not going to fly.”

Yeah. BackRub describes the technical underpinnings, but it doesn’t really describe the searching.

Yeah. So they’re casting about, trying to find the right name to encapsulate what they’re doing—this sort of new, good form of search. The story is that Larry’s dorm mate suggests the term Google.

Oh, wait. Do you know the name before this?

Oh, no. Do I know something you don’t?

Oh, this is great. Go for it.

The name is Whatbox.

Whatbox? Whatbox? It rolls right off the tongue. You know, it’s a box that you type stuff into. Sort of a question.

I got it. But they decided that it sounded too close to a porn site, so they decided not to go with it.

Well, hey, I guess if Facebook can be a product name and a company—

Yeah, maybe we could all be Whatboxing things. I mean, we are all WhatsApping, so to be fair.

Yeah, it wasn’t that crazy. It wasn’t that crazy. But yes, Whatbox out. Next name: how did Google come about?

Larry’s dorm mate suggests that they might want to use the term googol, G-O-O-G-O-L, which is the mathematical term for 1 followed by 100 zeros, 10 to the 100th power.

Yes.

The legend is that Larry loves the name. Sergey likes it. They go to register the domain name, and Larry misspells it, thinking that Google is spelled G-O-O-G-L-E.

Really? I thought Go.com was taken.

Oh, maybe that’s it. Maybe that’s it. Like anything here, there are a lot of legends floating around.

But the misspelling is actually great, because you kind of should spell it the way that other people are most likely to spell it, right?

Right, exactly. So this is before you’ve got Google’s “Did you mean?” in the search box.

Yes. Spelling was important.

So Sergey designs the homepage and makes the first logo using GIMP, the open-source drawing program.

Oh, and you can tell it was drawn using GIMP.

A lot of people can probably think of the earliest Google logo they’ve ever seen. Even the real nerds out there are like, “Oh, yeah, I know about that one. That was really colorful, before the drop-shadow thing.” There’s even one before that that’s completely illegible, and this is the one that we’re talking about.

But it was rainbow-colored.

Yes, it was. We’ll link to it in the show notes and on social media. It’s fun to look at.

Yep.

But basically, that homepage design—a colorful logo and a search box—that was it then, and that’s it today, from 1997 onward.

That 1997–98 academic year is when they’re building BackRub into Google. By spring quarter of that year, Google.com is doing 10,000 queries a day. It starts to spread virally, first on the Stanford campus and then to other academic universities and communities. People get wind of what they’re doing, and then it starts spreading into Silicon Valley. It’s bringing the Stanford network to its knees with all the traffic happening on Google.com out of Stanford.

At one point, they actually bring down the Stanford network. This is how fast it all happened. It’s all during this academic calendar, where they’re taking BackRub and working it into a search box. There’s now a keyword that they’re ranking things for, not just arbitrarily ranking them. That keyword relies heavily on the anchor-text description.

There are these 2 early key innovations: weighting results based on backlinks, and using descriptions from anchor text. It really did just work. The technical underpinnings are extremely difficult. They’re having to do things like steal computers from other research projects. Do you know about the loading-dock stuff, David?

Oh, yeah. Yeah, yeah. There are these famous stories of other researchers who had ordered computers but weren’t going to start their projects for a few months. So Larry and Sergey would grab them off the loading dock, spin them up, and use them for Google for a few months until they needed to hand them over to the other research projects.

David, to your point, they bring down the Stanford network because there’s so much traffic. There’s so much demand for something that is just clearly a better way of ranking websites than what everybody else was doing. Everyone else was basically just using keywords on pages and saying, “Well, what pages exist out there with the word ‘dog’?” If they had a whole bunch of instances of “dog,” that was going to be the top of your results for “dog,” no matter how authoritative they were. Obviously, that’s a problem.

This was just a better way to do search, and they were really starting to sop up Stanford’s network bandwidth. At one point, they were using about half the bandwidth of the entire university to serve Google.com pages. And David, to your point, it wasn’t a heavy website. It was a white page with an image and a search box, and when you went to the results page, there were no images. If they were consuming an incredible amount of bandwidth for something that was so asset-light, people were using the crap out of this thing.

Yep.

So here we are at the end of the 1998 academic year. It’s clear this is going to be a company. This has to be a company. Stanford is about to tip over if it stays a project any longer. Stanford has been very kind to say, “We’re going to keep housing all the infrastructure for this thing,” but at some point this needs to be a company so that they can get it off Stanford’s network and fund it on their own.

And nobody should shed a tear for Stanford here, because as part of the tech transfer to spin it out of the university, they end up getting about 1% of the company or something like that. Stanford did very, very well for their largesse here.

Yes. So Larry and Sergey go to a professor in the computer science department at Stanford named Dave Cheriton. Dave had started an Ethernet company called Granite Systems with Andy Bechtolsheim from Sun while also staying on as a professor at Stanford. He was a founder of Granite Systems and had stayed as a professor. Cisco had just acquired Granite for $220 million.

Larry and Sergey are like, “Oh, okay. Dave is one of our professors. He knows how to do this stuff.” And Dave is like, “Well, why don’t you talk to Andy about how we could spin this out and make it into a company?”

So Dave emails Andy that evening. Andy replies right away. He’s like, “Sure, I’m kind of busy tomorrow, but how about we meet at your house at 8 a.m.? I’ll come by on my way to the office.”

4. The Google Seed Round

And thus begins the story of Google’s legendary seed financing round and the crazy cast of characters involved in it.

All right, David, the Google seed round.

Yes. Here we go. So, at 8:00 a.m. the next morning, Larry and Sergey rouse themselves out of bed over on the Stanford campus, head over to downtown Palo Alto to Dave Cheriton’s house, and Andy Bechtolsheim drives up. He’s like, “All right, I’m in a hurry. Show me what you’ve got.” They demo Google for him. Andy loves it. He’s like, “Great, I’m in. $100,000.”

Larry and Sergey are like, “But we weren’t talking about raising money. We just wanted some advice to start a company.” Andy’s like, “Great. I’ll go get the check from my car.” He writes a check to Larry and Sergey, made out to Google Inc., for $100,000. He basically just throws it at them, hops in his car, and takes off. Google Inc. does not exist yet. This is actually true. This actually happened.

Andy’s like, “You guys figure this out. That’s your problem, not mine. I’m good for the money.” There are no investment documents and no valuation. It’s just, “Here’s $100,000.”

I assume I will get something for my investment.

Yep, exactly. And this was the forcing function for Google Inc. to get founded. Larry and Sergey need to be able to spin up an entity, have that entity own the intellectual property from Stanford, and set up a bank account such that they can deposit this check before it expires.

It takes a couple of months to get all this done, which, depending on who you ask, is either very good or very bad. In the intervening months, Dave himself decides to throw in another $100,000 to the funding. Larry and Sergey meet a former Netscape guy named Rahm Shriram, who starts advising them on starting this company and spinning things out. Longtime Acquired listeners will recognize this name from our Amazon episode.

Oh yes. Rahm had left Netscape and joined a startup called Junglee that Amazon.com then acquired. Rahm had a little bit of liquidity, so he throws in $250,000 into the round. And he’s like, “Hey, do you guys want to meet Jeff?”

Jeff Bezos, who at this point is kind of interesting. You think about Amazon and Google as equally old companies. Jeff is sort of the elder statesman of the internet. His company was started in ’94. This is ’98. Amazon just went public the year before.

It’s kind of crazy that at this point in time, it was little old Larry and Sergey, grad students, meeting the public-company CEO Jeff Bezos.

Yeah, yeah. It’s almost like when your 6-month-old baby is hanging out with a 14-month-old. You’re like, “Oh my God, they’re so different,” but they’re going to be classmates in a couple of years. They’re basically the same age.

Exactly. So Rahm arranges a meeting, and the next time Jeff is in Silicon Valley, they all meet at Rahm’s house. Similar to Andy Bechtolsheim, Jeff’s like, “Great, Rahm, what are you in for? $250,000? I’m in for $250,000, too.”

All in, they end up raising $1 million at a $10 million post-money valuation. And yes, Jeff Bezos does a quarter of Google’s seed round.

It’s so crazy. We knew about this in the past because we talked about it on the Amazon episode, but whenever I heard someone reference Jeff Bezos as an angel investor in Google, I always thought, “Well, yeah, but you look at any of these startup cap tables and there are 50 founder friends in addition to the main VC. That’s not surprising at all.” But Jeff is a quarter of the money in the seed round.

Yeah, one of 4 investors.

Is that right?

That’s right. Yep. It’s Andy, Dave, Rahm, and Jeff. Jeff has never said whether he sold any of his Google shares along the way, but by my math, if he didn’t, that stake is worth about $20 billion today. And even if he did sell at the IPO, he turned that $250,000 into something like $200 million at the IPO.

Right, right, right. Nice returns. And Amazon stock was probably in the dumpster when Google went public, so he could have used the money.

Anyway, Google is now an official company. They’ve got $1 million in cash from their crazy seed round. It seems like they might burn through that pretty fast because of their business model of trying to store the entire internet on their servers. But you’ve got $1 million. You’re no longer a Stanford project. You’re spun out. You’ve got investors. What’s next?

Well, first, office space. Famously, they go find space in a Menlo Park garage, a house owned by Susan Wojcicki, who was a manager at Intel and soon would herself become an early Google employee and eventually CEO of YouTube.

Yeah, exactly. But besides office space, it’s time to put your product out into the world. I think it’s worth drilling in here: What was the state of search in 1998, when Google becomes a company?

5. Search Before Google

Yep. I think it’s worth saying a little bit more about 2 other players to set the context. The first is AltaVista. Folks who are old enough might remember. AltaVista, pre-Google, was pretty good. AltaVista has a fascinating history. This is wild. I didn’t know this until doing research for this episode. Do you know where AltaVista came from?

It was pretty good.

I do, but most people don’t. DEC—Digital Equipment Corporation. DEC’s Western Research Laboratory, which was its Palo Alto Research Lab, its equivalent of Bell Labs. But DEC—I mean, the company we talked all about in our Microsoft series, where Dave Cutler came from, who wrote Windows NT—you know, legendary, legendary company, right? Hardcore enterprise, big-hardware computing company. The minicomputer company Judy Faulkner wrote Epic on a DEC minicomputer, right?

Yep. This is where AltaVista came from. And the big insight that they had was that you can crawl the web to build the index in parallel.

Before AltaVista, all the other web crawlers out there that were building search engines were just single-threaded processes. You’d crawl 1 page, and then you’d crawl another page, and then you’d crawl another page. The internet was small enough back then that people didn’t really think to do it any other way, because remember, it was all growing so fast.

It’s a super-parallelizable process. Why not?

And this makes sense because DEC hardware would be pretty well suited for this. They’ve got a big enterprise appliance that they need applications for.

Exactly. This is why it was a research project at DEC. It was a way to show off the power of their latest enterprise-class servers that they were hawking.

So, if you think about the competitive vectors of search—what makes one search engine better than others—it’s not just the ranking. Today, people think about PageRank and Google, and the innovation in the ranking and the relevancy. That was the most important thing. There are actually 2 other vectors that are critical.

One is speed. How fast are you going to return the results? We take that for granted today, but not too long before Google’s founding, search was a thing where you’d kick off a query, go do something else, and wait for it to come back.

Right. It was like AI today. You’re doing deep research, and you’re just like, “Okay, great. Send a query. Go get some lunch. Come back.”

Exactly. We’ll get more into speed in a minute. But the other attribute that’s super important is the index. How big is the index of pages that you’re searching across?

Before AltaVista and the parallelization of crawling, all the indexes of all the other search engines were super small. Maybe 1 million pages was the biggest. You could have the best search engine in the world, but if you’re only getting a small percentage of the actual sites out there, it’s not going to be that useful.

And the funny thing about this period of time, too, is that very rarely were people actually updating their index. So, when something would say, “1 million pages crawled,” that was cumulative. They just kept adding new sites to it and weren’t doing many updates.

Right, right, right. Oh man, how times have changed. So, when AltaVista spins out of DEC and launches as a commercial company, an entity in and of itself, its big claim to fame is its index. It has 16 million pages in its index versus its competitors.

AltaVista still kind of sucked at relevancy and speed. The information-retrieval algorithm that AltaVista and others were using was highly based on how many times a given query word appeared on the page.

So if you wanted to rank highly for dog food, you just spam “dog food” in invisible text all over your page, or even not invisible—you just want to make the dog-foodiest dog food page on the web.

Exactly. And the other thing that, I guess, AltaVista was probably fine at but not good was speed. And that was a particular problem. I mean, AltaVista had great hardware from DEC, and really expensive hardware, too.

That’s a main thing to underscore here: yeah, they’re doing this cool parallelization thing, which leads to a bigger index, but if they had to be their own company, it is an extremely expensive company to run to have all that DEC hardware for search, which we should say doesn’t have a great business model yet.

The business model is just banner ads: low price, not very targeted. That would all come later. And so the whole search market—why would anyone take it seriously? Because to date, it doesn’t feel like there’s a good business there, and to do a really good job at it, it would be very expensive to run.

Yes, it is funny. As you can imagine, the powers that be at DEC are pushing really hard on, “Can we sell more of these boxes?” Do they really want to be the ones developing the best search engine, or really what they want is for other people to be developing stuff like this, see it as a proof of concept, and then start their own companies to buy more and more DEC hardware?

Yes, exactly. Ben, search as an industry was not interesting. One, because the economic upside was capped, and two, also, people loved directories and portals and Yahoo. Yahoo was the big player, not AltaVista or Excite or Lycos or Infoseek or any of these others.

So Yahoo was the site that was taking off like wildfire. They’d gone public in 1996 at a $1 billion market cap. By 1998, when Google.com was launching as a company, Yahoo was a $20 billion public company. This was the juggernaut.

And what was so great about Yahoo? Yes, it was started by Jerry Yang and David Filo, 2 other Stanford PhDs, but it wasn’t technology-driven. It started as Jerry and David’s Guide to the World Wide Web. It didn’t start as their academic research.

What Yahoo was was exactly that. It was a hand-curated guide, a directory to the internet. It was kind of like the Yellow Pages, with even better annotations explaining why you would want to look at a particular site. And that’s what people thought: technology search engines would never be able to replace human curation and human thought about what the most interesting sites on the web were.

Well, this is why Larry’s original idea was this annotation idea. It was humans who were going to rank things. And for the size that the web was at the time, Yahoo was correct.

Yes, when you have a small number of total websites, curating them is interesting. But when you have 10,000 times more websites, with more niches that people are interested in, a directory is not going to be an efficient way to surface what people are looking for. If you believed that the internet was going to get as big as it did, search became a more interesting front door.

But for this period of time, a directory was an amazing front door to the internet.

Yep. Now here we are in 1998. The internet is already big enough that, yes, it’s clear there are a lot more interesting web pages out there. So Yahoo has search to address that.

Exactly. So the model was hybrid. All these portals, and Yahoo included, went hybrid: when you searched on Yahoo, the results you got at the top of the page were their hand-curated, directory-driven results, and then they backfilled with a search engine. And so they would partner with these search engines to provide backfill results. People thought that this was the ideal solution.

So interesting. It is the epitome of just good-enough technology. It was so different from Google. Google wants to be the very best technology solution for a problem, the most elegant. And I think the Yahoo solution was very much, “Search just has to be good enough.” The curation is sort of the thing that matters.

We’re a media company, with enough human editors to cover all the big categories, but the business is showing banner ads. They may or may not be relevant to whatever page you happen to be looking at right now, and we’re effectively a media company that has search just in case.

Yes. Okay. So that’s what’s going on at Yahoo and all the portals. The opinion of Larry and Sergey is: we don’t want to do a homepage like that. We don’t want to clutter it up. Our whole point is to help people find what they want.

Which, of course, raises the question: what’s the business then? Because if you’re not keeping people on-site to see your banner ads, the only moment that you really have is on the search-box page and on the search-results page.

And they were extremely against—well, really, ads generally. They didn’t think it was good for users, but they were especially against banner ads.

And there’s this sort of scary thing, which doesn’t seem scary now because we know how it played out, but just imagine trying to evaluate this company. It’s growing like wildfire. Everyone’s using it. There’s 1 known business model for this entire sector. It’s not a great one, but it is known. And these guys are dead set against using it.

But on the other hand, let me pitch it to you a different way. These guys are building the front door to the internet, which was just growing 700% year over year. So isn’t that going to be really valuable?

Yes, but we don’t know how yet. But the problem is actually even more dire than what you’re saying. As usage is growing, they need more infrastructure, but they’re not making any money, right?

And for each piece of this, you need more infrastructure. You need the crawler to go and crawl the whole web and store—not entire web pages, but little pieces of web pages that you can reference from your index. You need the index itself. You need to serve the web pages up for when people are doing the searches.

There are a bunch of components of this infrastructure that all need to scale, and they all need to scale differently.

Yep. Which brings us to really what is the second big reason why Google worked so well and became the Google we all know today. One is accurate, relevant, fast search results and PageRank and everything we’ve been covering. Two, though, is the infrastructure to actually make this whole thing work and scale efficiently.

6. Google Builds Its Infrastructure

Yes. So right after they raised the angel round, Larry and Sergey went out and recruited just unbelievable, top-tier engineers and computer scientists to come rewrite the code and work on this infrastructure problem. So pretty quickly they got Urs Hölzle and then Jeff Dean, who are just these absolute legends. They are both still at Google today.

Urs is now a Fellow, but he ran all of Google’s infrastructure from 1999 until 2023. Before joining, he’d done his PhD at Stanford, and he was a professor at UCSB. He’d also written the primary Java virtual machine that Sun used as the official Java virtual machine.

Oh wow.

And Larry and Sergey recruited him out of academia to come join as employee number 8, and his initial job title was search engine mechanic because, quote, “everything was broken.” So that’s how he built all this incredible infrastructure.

Jeff Dean, whom they also recruited around the same time from his role as a senior engineer at DEC—

Yes. Yes.

And Jeff is basically Google’s Dave Cutler. So today Jeff runs AI at Google. He also implemented the first version of AdWords, built AdSense, rewrote the core search pipeline 5 times, co-invented and implemented Bigtable, MapReduce, TensorFlow, and Gemini.

He actually keeps his résumé up to date online. We’ll link to it in the show notes. It’s incredible. We’re burying the lede here: we spoke with Jeff to prep for this episode, and I watched a handful of talks he’s given. Delightful human, and God, what a great engineer. Just generational talent.

But this is that early nucleus of engineers that Google recruited. It’s amazing that they attracted them because prospects were not good that all of this would work and scale. And it was only because of these guys that it did.

Well, and here’s the crazy thing. Later, there’s an easy point to make, which is Google got to hoover up all the best talent because they were a solid business after the dot-com crash. But this is in 1998; we’re in the go-go times. The dot-com bubble hadn’t burst yet, and Larry and Sergey managed to recruit this talent.

I think this is like history turning on a knife point, or like a make-or-break-the-company thing. The fact that they were able to get these guys in a hot talent market really speaks to Larry and Sergey’s vision, the excitement around the idea, how novel their approach was, everything.

Yep.

And part of the reason why this talent was attracted to Google—sure, some of it was, “Oh, the product’s really good, and people are using it, and so that makes the company interesting.” The other part of it, though, was that the technical challenges and the architecture coming out of Stanford were super unique and novel. This was a really interesting thing to work on.

And why was that? So the Google index that they needed to build and operate for the search engine, for PageRank to work, was so much bigger than any other index out there. Google needed the entire page to compute all the rankings and find the links, find the backlinks. They needed to architect Google with this huge distributed computing system.

So the index was so big that it wouldn’t fit on a single machine or a single server, no matter how big or how expensive. So what they did to store the index and operate on it with this distributed file system was break the giant index into tons and tons and tons of little chunks, as they’re called, of individual 64-megabyte files.

Small files, tractable files. And they get stored on lots and lots of different disks and lots of different machines and lots of different servers, and ultimately in different data centers all over the world. And then there’s a separate server that keeps a master mapping of all the chunks, including where the chunks physically are.

And so when a query comes in and needs to operate on the index data, the master server just returns only the chunks that it needs, not the whole index. And that makes the whole thing possible.

So basically, that one server you're talking about can kind of just say, “Oh, all the chunks are here on all these different machines that are distributed throughout my data center. Just look at those chunks,” and that way it can kind of just pull, in a parallel way, from all those different chunks concurrently.

Yeah. And I think that's even abstracted from a compute perspective. They see the master map; they feel like they have access to the whole file, but then what's actually getting returned to them to operate on is only just the chunk data that they need.

Hmm. So Google was sort of forced to do distributed computing because their index file was too large to store on any one machine, no matter how big or fancy it could be.

Yep, I think that's right.

That sort of enables the whole thing in the first place and is technically extremely interesting. But now, the physical infrastructure side: you have all these chunks, and they can live anywhere, and Larry and Sergey already had to grab commodity hardware—hard drives and motherboards—directly back at Stanford.

Well, Urs comes in and he's like, “Whoa, we can just keep going with this. Let's keep using cheap commodity components and hardware.” And, yeah, they'll suck and they'll fail a lot, and things will burn out, but that's okay, because we've got this distributed file system. We'll just replicate everything, like, 3 or 5 times, right?

And we can cleverly design software to account for the fact that we have commodity hardware or commodity RAM. These systems were not assembled with the notion of being enterprise-grade; we can sort of design Google with the idea of taking into account the fact that the hardware is not enterprise-grade. And that means we can get cheaper hardware and run in a distributed computing way.

And frankly, I think this makes it interesting to a lot of engineers who kind of want to work on hard problems. How do I design a system when I can't count on a whole bunch of stuff from the underlying hardware that I would get to count on if it was a fancy DEC server?

So, I read that the industry-average server hardware failure rate at the time was around 3% to 4% per year. Google's hardware failure was over 10% per year, but the whole system was designed so that it didn't matter. It was all just replicated. Super interesting.

So this keeps scaling up and up and up over the years with Google pretty quickly. Maybe even while Google is still a private company, they technically become the world's largest computer manufacturer.

Oh, wow. Because they're not buying fully baked servers. They're just buying components and assembling them into this sea of components—a proto-data center in their data center.

And then, in their early data centers, they're never really putting them in PC housing, right?

Yes. So these early, quote-unquote, machines they're building, they're not even putting PC cases on them. They just mount the motherboards directly on corkboard, and then they put the RAM in there and they put hard drives in there, and then they just stuff them in their data center racks.

The photos of these early Google, quote-unquote, server racks are crazy, because the way that their agreements worked in the colocated data center facility is they would lease by square footage—not by energy consumed, not by number of machines, but by square footage.

When you give a computer scientist a constraint, they will optimize for it. And the goal is, how much of Google can I power in this square footage? And the way that you optimize around that is, well, incredible density of hardware. So we're not putting cases on these computers; we're putting corkboards in.

And imagine just a sheet of cork, which is an insulator. So you know that these electrical components that you don't want to conduct between each other are not going to conduct between each other. And you just stuff a server rack full of corkboards with all this commodity hardware sort of strewn about it. And it looks unbelievably messy. It's extremely economical, and then you just kind of handle it all in software.

The net of this is that Google can scale, period, but also can scale way more cheaply as search traffic rises and as the index keeps growing and getting bigger than anyone else out there on the market. So once the business model kicks in, this is why Google Search has like an 87% gross margin on it.

Yep. There's this incredible story of Google's first data center. It was a colocation data center facility, kind of a shared physical space, in Santa Clara called Exodus. And the data center cage—the space that Google had allocated—was right next to the cage for Inktomi, which was a competing search engine out there that we'll talk about in a minute.

And Google folks talk about the Inktomi cage having all these gleaming Sun machines and lots of space and lots of airflow and all this incredible cable management, and then you had this Frankenstein Google thing next to it. And to your point, Ben, they were only paying by square footage. They weren't paying for power, so they were sucking up all the power of the data center.

I heard a story that they actually, at one point, may or may not have stolen a power circuit from the Inktomi cage next door.

Borrowed.

Borrowed. Borrowed. Yes.

So one fun illustration of what it means to be on commodity hardware versus enterprise-grade hardware: Jeff Dean shared a fun story with us. On enterprise-grade hardware, you would have something in RAM called a parity bit, and in consumer-grade hardware you don't.

And what is a parity bit?

A parity bit adds 1 extra bit to memory that is basically for error checking. It looks at the rest of the data in the byte, and if it's even, it'll set the parity bit to 1, and if it's odd, it'll set the parity bit to 0.

The con of this is now you need an extra bit, so it makes the overall machine more expensive because you're losing 1/8 of the RAM to this parity checking. But the benefit is you know that nothing ever got corrupted, because the likelihood that one of your bits got flipped and it also flipped your parity bit is very low. You can kind of check and see, “Wait, it's supposed to be an odd number according to the parity bit, but it's an even number,” so you know likely something went wrong.

And when I say something went wrong, this is from random radiation that's just flying around the universe at any given time. Stuff goes wrong all the time.

Well, because Google is using this commodity hardware, they then have to do these crazy things in software and build all these layers themselves to say, “We're running on crap hardware. We don't know for sure that the value in memory is correct. Can we have a second way of verifying that it's correct?”

So it's that sort of—I don't know—cool software engineering, but also another layer of systems that you have to build when you're on commodity hardware.

So the net of these constraints and the incredible technical team Google has is that they design everything from the ground up: the computing systems, the file systems, the data centers, the racks, the hardware, everything. And they built stuff like GFS, the Google File System, and MapReduce.

Yahoo would eventually feel like they needed to copy MapReduce to be competitive, and they would open-source that as Hadoop. So if people know Apache Hadoop, that is a Yahoo copy of Google's MapReduce.

Yep. Of course, shepherded and stewarded mostly by people outside of Yahoo eventually, but that's where it came from.

Yep. And then ultimately, because Google is building their own hardware, racks, and data centers, and they can do it cheaply, they put data centers all over the world. And then that means they can deliver search results and ad results instantly to users all over the globe.

Yeah. So this speed really starts to become a bragging point for Google, where whenever you do a query, it'll show you how long the query took. It's usually like a quarter second. And they used to brag about the index size. Now they say, “I'm returning a gajillion results to you.”

But this was a really big flex for a long time: we searched a huge index, billions of pages; we found a huge number of results; and we did it really fast, and we're going to show all those numbers to you because, A, we're engineers and we're awesome, but B, we know it's the best stats that anyone out there could report to you.

Yep. It's a stake in the ground.

So all this grows out of the constraints: they don't have any money, first at Stanford and then this little angel round they raised. They don't have a way to generate any money. And, yeah, they don't have any way to generate any money.

So we cannot overstate how important Google's infrastructure innovations were. All this comes out of these constraints that Google the company has. But none of this would have mattered if they didn't figure out the business model.

7. Google Finds Its Business Model

All right, David. So, going from having no business to the greatest business model humankind has ever discovered—not exactly a straight line, huh?

No, not at all. How does it all start? Early 1999: even despite the incredible infrastructure work and being able to scale cheaply, Google is still running out of money from the angel round. They hire a recent Stanford undergrad named Salar Kamangar, who joins Google as employee number 9.

Like many Stanford students at the time, he started using the search product while he was an undergrad, was blown away, and thought, “I’ve got to go work for this company.” He basically bangs down Google’s door, tries to get hired, and finally they’re like, “Okay, okay, come on in.”

The first thing that Larry and Sergey give him to do is, you might argue, the most important thing in the company. They’re like, “Well, we’re running out of money, so we need to go raise venture capital. We don’t want to write the business plan and the pitch deck. You write the business plan and the pitch deck.”

Salar goes off and writes the pitch deck for the Google Series A, in collaboration with Larry and Sergey. They come up with a 3-pronged business model that they’re going to present to VCs—3 ways that they’re going to make revenue—and it’s hand-wavy as hell.

You know what this kind of feels like to me? It feels like they’re saying, “Google.com is so precious and amazing and special. We don’t want to risk it by having to make money on it. Can we make money doing something else with our technology?”

Yes. That will fund what we really want to do, which is Google.com.

Now, it’s pretty funny to talk about this in retrospect. They did have a couple of reasons why they thought this might work. Number 1 was that all the way back at Stanford, with BackRub, and then with Google, they had actually used Google for this use case. You could use Google to search internal Stanford intranet stuff, and people did. It was a great experience for Stanford students.

Also, before the Series A, somehow Larry and Sergey had managed to sell one of these deals to Red Hat. That was their first revenue, right? Red Hat?

That’s right. That was their first revenue, right? Red Hat?

Yeah, the open-source Linux company. They sold this enterprise search deal to Red Hat for $20,000. They were like, “Oh, great. There’s a market here.” So, that was going to be the main business driver.

There were going to be 2 other business lines in the company, too. One was going to be, “Well, sure, okay, VCs, you make us. We’ll sell ads—CPM banner ads—the same way everybody else does. We’re not going to like it, but we’ll put it in the business plan.” It seems like they didn’t think through it any more than that, because I couldn’t find anything about whether that was going to appear on the search results page or next to the search box on Google.com. It seems like it was never real enough to actually have a plan for it.

Indeed, the Series A pitch deck and business plan were intentionally vague. I think part of the reason Larry and Sergey were like, “Okay, Salar, new guy, you go do this,” was that they didn’t actually want to tell the VCs that much.

The third thing was that they were going to license Google organic search results to portals and directories as essentially OEM search, to backfill results like we were talking about with Yahoo. Other search engines were doing this. Inktomi had gotten started at this point in time. Inktomi was the next-door neighbor at the Exodus data center in Santa Clara, and they built a sizable business selling white-labeled organic search results to other portals.

Exactly. This was Inktomi’s whole business: they just sold organic search results white-label to other portals. But there were 5 big customers and 50 total customers out there for this business.

Yep. So, this is the business plan. This is the pitch deck. They go out. Remember, we’re in spring 1999 here, so even though this is a little harebrained, it’s still the dot-com bubble. Money is still flowing.

There’s a great Michael Moritz quote in Steven Levy’s book talking about this particular moment in time. He says, “Nobody’s feet were on the ground.” This is a very Sir Michael way of putting things.

A deliciously Michael quote.

Yes. Google’s got all this usage and engagement and growth numbers. Internet companies trade on eyeballs, so of course this is going to be a hot deal.

Famously, Kleiner and Sequoia end up splitting the deal. Michael Moritz and John Doerr, the 2 most legendary VCs in the world, team up, join forces, split the deal, and both join the board of Google at the Series A, which is unheard of.

The fact that Larry and Sergey were able to say, “You both only get 12.5% of this company, and you have to do it together”—they both must have really, really, really wanted to do the deal.

That is true. And I love that even today, even you have that opinion.

We heard from folks in the research that this really was a Google PR masterstroke to seed this narrative.

Well, they held a press conference in person with both Sir Michael and John Doerr. That was the first Google press conference. Larry and Sergey were there in Google-branded shirts.

Yes, they made a big deal about this.

The reality is Sequoia and Kleiner split tons of deals. This was not the first one. It may have been the first one that Michael and John split together, but they had been on boards together before. Maybe they had done one round or the other, and Sequoia and Kleiner split deals all the time. But it was the dot-com era, and everybody needed a PR strategy. That one worked well.

Fascinating. But the point is, this was a hot deal. There are all sorts of stories out there about other investors coming into the round or trying to get in, or trying to get in later.

Yeah, I think in the middle of negotiations they got another term sheet at a $150 million valuation instead of $100 million. But I think they were sort of already pot-committed to Sequoia and Kleiner Perkins. I actually don’t know who the $150 million offer came from, but I do know it was someone who Ram Shriram set the meeting up with.

Interesting.

And remember, Larry’s older brother had co-founded eGroups, which Moritz had funded at this point. They knew they wanted to go with Sequoia and Kleiner. That was the goal all along.

Regardless, the round gets done: a $25 million total raise at a $100 million post-money valuation. A $100 million valuation was genuinely wild for the time.

Yeah. Even in the heyday of the dot-com craziness, for a Series A at a $100 million post-money valuation, that was newsworthy.

So funny thinking about this today. I know. Today it’s so quaint.

The comp in today’s world is: imagine reading a headline that a Series A, after a company just had a few angel investors, got done at a multibillion-dollar valuation. That’s kind of the way it would have felt in tech at the time.

Yes, totally. But despite that, and despite all the hype around the Series A, there’s still an urgent imperative to make revenue. Yes, there’s $25 million in the bank now, but you’ve got VCs involved. The playbook back in the dot-com days was: invest in the company, get quick revenue, go public.

Right around the same time as the Series A happens, Larry and Sergey meet a guy from Netscape named Omid Kordestani. The first time they meet, I think Omid is thinking about it in the context of, “I’m wearing my Netscape hat. I’ll evaluate whether there’s some partnership here.”

I think he quickly gets the sense that maybe there is, but Netscape just got bought by AOL, and it’s getting a lot less fun there. Omid was the VP of sales and business development at Netscape, so he’s very familiar with building an internet-based business. What these Google guys are doing is very interesting.

Of course, Larry and Sergey, as the great recruiters that they are, say, “Hey, why don’t you come work here?” Omid joins Google essentially as chief revenue officer, and he’s tasked with, “Okay, take this business plan, take these 3 areas, and make them a reality.”

We should say, too, Omid is an awesome guy.

We talked to him in research.

Yeah. So he goes out, and of course, the first innovative business model that they pitch the VCs—the enterprise search business model—he starts trying to sell it. But as you might imagine, especially at the time, there wasn't a lot of customer pull, shall we say, for this.

So, for the first 6-plus months of the company, basically the rest of 1999 after the venture funding, things are not looking good on the revenue front. David, do you remember—this is way back in Acquired History—what Doug Leone told us in February of 2020 when we were recording our episode with him? You give the quote.

Yes. I know exactly what you're going to say, and I've got some more flavor on that quote. The quote is from Sir Michael Moritz. He comes to Doug, and they're running Sequoia Capital together at the time, and he says, “Doug, we've never paid so much for so little.”

That's the lore. I got a little more behind-the-scenes flavor on the quote. Apparently, it was not Sir Michael who said it first. He might have just been repeating it back.

Apparently, was it John Doerr?

It was Vinod.

Oh, in the Kleiner partnership. That's what I heard.

This is like the hot potato of quotes. Nobody wants to actually take credit for this.

Either way, though, the sentiment is right. You can understand why Kleiner and Sequoia would feel this way. They kind of have egg on their faces. They just paid a $100 million post-money valuation for a Series A company with no revenue. The revenue is not materializing. We're now into the year 2000, the bubble is starting to burst, and Google still basically has no business.

Yeah, no business. But growing market share, fervent, loving fandom among the people using it, and providing real value to people. There's got to be something here.

Yes, exactly. So the revenue imperative is becoming—well, imperative. More of an imperative, shall we say? Omid's a smart guy. He's like, “I'm not going to just keep banging my head against enterprises here. We're going to pursue the other 2 business lines that are obvious. Who knows how big they'll be, but at least we'll make some money.”

So, set up just regular ads, the same way as everybody else does it. Omid goes and hires Tim Armstrong in New York to set up an ad sales force. Google does start selling ads at the top of search result pages. And what do these ads look like, David?

Importantly, Larry and Sergey insisted that if we have to have ads on here, they need to be text only. We can't serve images and banner ads like everybody else does because that'll slow down the page.

It's like they always talk about it for taste, which is true, but yes, it's a performance thing.

Yes, it's a page-performance thing.

There are these great stories about Tim in New York, Omid, and the ad sales force that they're building up at Google. They're going to ad agencies. They're going to advertisers directly. They're trying to sell these ads: no images, just text. “Trust us, it's going to work.” Not very exciting to these Madison Avenue guys.

How are they paying for them right now?

Still CPM. So, you buy a keyword, and then you're promised that you're going to be an ad that appears on the page whenever that keyword is searched, and you're going to pay per thousand impressions of that keyword.

Is that right? Yes. Okay. Not self-serve. No web tools for this. This is negotiated over the phone, and then they manually hand it into Google: when this keyword is searched, you need to display a text ad for this person and track how many times you displayed it, because we then need to invoice them for how many times the page loads.

Not even on the phone. That would be really technologically advanced for Madison Avenue at the time. All ad insertions were done by fax at this time, so Google had to install fax machines at its headquarters to take these insertion orders for the ads that they were selling. Awesome.

Now, what was the pitch to Madison Avenue about why this would work?

Intent, baby.

Exactly. So, the very first project that Jeff Dean did when he came over from DEC is Larry, Sergey, and Omid told him, “All right, the VCs say we've got to sell ads. Go figure out the tech to serve ads on Google.com, but don't do anything to degrade search or the user experience.”

Jeff Dean works with Marissa Mayer, who just joined from Stanford undergrad, and they're like, “Okay, this is going to have to just be text. What can we do as a test to see if we can engineer something that'll work with text ads at scale and run it against a bunch of queries?”

“Well, what about Amazon affiliate links? We know a guy at Amazon who happens to own a good chunk of this company.”

So, Google goes and signs up as an Amazon affiliate. How crazy is this? The Google business model was validated by doing customer development—startup-idea validation—using Amazon affiliates as the mechanism.

Yes. Jeff Dean codes it up so that dynamically, as users are searching, if there's a query that is related at all to any book in the Amazon library catalog, Google will dynamically generate a text ad saying, “Go buy this book at Amazon,” insert an Amazon affiliate link, and drive traffic over to Amazon.

And the amazing thing is, of course, it actually works. Whatever the Amazon affiliate commission—if it's 4% or 5% of the revenue of a $10 book—obviously isn't going to change Google's fortune in the amount of money that they'll generate from this. But it's a test. It's a test and it's proof that they can then take to advertisers: “We can capture intent, and we can send highly monetizing traffic to you based on the keywords that you are buying.”

And so, if you think about the funnel steps, there's the impression of an ad: they saw the ad. Then, 2, there's the click: they click through the page. Then, 3, there's the on-page conversion.

It would be one thing to just test click-through rate, which would have shown a great result here, because click-through rates on search ads are higher than ads that are just randomly around the internet. Someone is intending to buy something. They have high intent. That's a great place to show an ad, and the click-through rate is going to be higher.

But what they also know, because it was Amazon affiliate links, is the down-funnel number too. They know conversion is actually higher from this traffic, because they got to know how many books on Amazon were sold from the number of impressions that they served.

So they can say, “This intent-based ad system has a high click-through rate and high conversion. Yeah, they're just text ads, but you're going to like the numbers.”

This is ultimately a math problem.

Yeah, it was an absolutely brilliant test and bootstrap of the first step of the Google ad model.

Yep. So that's sort of Gen 1 of the Google Ads business. They get it set up, it's going, it's making some money, and they're winning some clients on Madison Avenue. Good, great.

The more interesting piece for the next year, year and a half, is the OEM search portal deals. There are some big deals to come: Netscape, Yahoo, AOL. This is white-labeling Google search to be the search powering other places' search activity that has tons of traffic—which, oh, by the way, is also going to train huge portions of the internet to use Google search.

Yes, it is. And specifically, what were the portal deals at this period of time before Google had a real, functioning paid-search business?

What a portal deal represented was just letting a portal use Google search to power its organic search, and in exchange just getting paid a fee for that.

Yep. It was the Inktomi business model.

It was exactly that. We're selling our search results for you to use on a third-party page, and it's effectively like a B2B supplier. They're a vendor to a portal, more or less.

Yes. It turns out, though, that we're still in the era of the internet where portals are pretty big. A ton of traffic.

And they're not just any old vendor. They're a vendor that at the bottom of every page says, “Powered by Google.”

So, pretty quickly after Omid joins from Netscape back in 1999, he goes back to his old colleagues at Netscape and AOL and gets essentially a proof-of-concept deal done with them. That's for Google to backfill organic search results on Netscape's own directory service that they just launched to compete with Yahoo.

So, it's interesting. We always think about Netscape, the browser, but this was presumably the Netscape homepage. Whenever you opened up, it would go to Netscape.com or something, and there'd be all these Netscape services available, one of which was search.

Yep. And I presume, having been acquired by AOL, this was now more of a strategic priority for Netscape, because this is AOL's whole business model at this point besides the monthly dial-up fees, right?

Worth remembering: AOL was way, way bigger at this point. Tens of millions of people were using AOL, and way fewer were going through Netscape.com to deliver traffic to this Google search. This is sort of the small part of the organization they're working with right now.

Yep, exactly. But relative to how small Google actually is at this point in time, it's huge.

Still big enough that when they flip the switch and Google on Netscape goes live, with Google powering the organic search results, there's so much traffic that it blows out Google's infrastructure.

Well, it comes close. Basically, they're watching the analytics like a hawk. Omid gets an urgent call from Sergey saying the traffic is about to tip over. And this is potentially company-killing, because this is their big strategic priority.

If they prove that they are untrustworthy to Netscape and can't deliver, then how are they going to keep Netscape's business, let alone get any other portal deals or be able to go sell to enterprises that they're thinking they're going to do at this point in time?

They cannot tip over. And so, they have this pretty tough decision to make.

But actually, it’s not even a decision at all. This is obvious: we are shutting off Google.com for today, and we are going to prioritize all traffic from Netscape for our servers until we can stand up more machines.

Yeah, just think about this for a minute. Think about everything that Google is today: never goes down, universally available basically everywhere in the world on every device. It’s freaking Google. In 1999, they shut Google down so that they could serve Netscape’s users.

Yes. I mean, look, the revenue is very material coming from this, and the reputational impact is very material. Like I said, it sounds like a hard decision. It’s actually not a decision at all. It also ends up being the right strategic decision because of what you mentioned a minute ago: the Powered by Google logo at the bottom.

Sure, you shut off Google.com for a day, and your own Google users, of course, don’t like that, but you’re training millions of new Google users who are going to see “Powered by Google” at the bottom.

And they got trained. I mean, the Netscape deal brought in 3 million total searchers per day. At first, Google was sitting there begging these early portals, “Hey, please put Powered by Google on,” and really trying to get that inserted in the deal. Later on, Google got so well known for having quality, fast search results on a big index that it was a value proposition to show your users, “Oh, our search is powered by Google.”

It becomes the Intel Inside of search.

Yes. It’s the ingredient brand. That’s exactly right. So now they’ve got some distribution, millions of users, but still very little revenue in June 1999.

Yep.

So over the next year or so, obviously they’re working on the other business models too, but Omid keeps signing up some smaller portals, some international portals, on the success of the Netscape deal, getting more of these OEM portal deals for Google. And then they start working on the big kahuna, Yahoo.

Yes.

All right, David. So, how did Google get big?

In June of 2000, they sign a deal with Yahoo right as the whole world is falling apart. The dot-com bubble is bursting. The peak of the Nasdaq was March 2000. In June of 2000, Google signs the deal with Yahoo that they are going to take over all organic search-result backfills on Yahoo.com, with the Powered by Google branding. And Yahoo is going to invest $10 million in Google as part of this deal.

What a deal.

This totally saves Google. Between the revenue that they got from Yahoo for this and the $10 million investment, it keeps the company going through the next couple years of the dot-com winter until they figure out the AdWords business model.

Yep. So traffic doubled to 14 million searchers per day on day 1 of this deal, June of 2000. We’re now a year later than the Netscape deal, so they started to get a material portion of web traffic here, with 14 million searchers per day.

Yep. And the next year, in 2001—the first full year of this Yahoo portal search deal—Yahoo pays Google $7.2 million for organic search results. So it’s material. And again, to underscore, between the $10 million investment, this revenue, the other portal deals—Netscape and others—and then others that they’re able to get on the back of Yahoo, this revenue really bridges the company through the dot-com winter.

That’s such a good point, and something that’s often pretty overlooked: there was no potential for Google to raise more money here. The venture capital gravy train was over. And so we’re sitting here saying, “Oh, they really need to make money. When are they going to turn the revenue switch on?” And we’re sort of hand-wringing over here. We’re only 2 years into the company’s life. Think about startups today. You don’t have expectations of profitability within a couple years of founding.

Yeah, exactly. Within a couple years of founding, but Google’s got a very expensive business to run between the people and the infrastructure, and there’s no more ability to finance it. So revenue really was the only option.

So in the midst of all of this, as Yahoo’s coming online, the board is also pushing Larry and Sergey to hire a CEO.

Yes. So as part of the Series A process, John Doerr had very begrudgingly extracted a promise from Larry and Sergey to hire a “professional CEO.” Larry was CEO for the Series A and then CEO for the next couple years after the Series A. They really didn’t want to do it. They were dragging their feet. It took 16 months to find a CEO. I don’t think that was entirely because it was hard to find someone. I think some of that was, “Let’s see how long we can get away without one.”

My favorite story from the whole Google CEO hiring process was the standard playbook here that John Doerr, Mike Moritz, Sequoia, and Kleiner Perkins would run with founders when convincing them to hire a CEO: take them around the Valley, take them on the tour, have them meet the CEOs of the great companies in the Valley, of the public companies, and say, “Look, see what a great CEO can do for your business.” So they do this with Larry and Sergey. They go around, they meet everybody in the Valley, and they’re unimpressed. They don’t like any of them. And finally, after months of this, they come back and they tell Kleiner Perkins and Sequoia, “All right, there’s 1 person that we met in this whole process who we think meets our bar, who we would be willing to come in and hire as our CEO here at Google.”

Oh, God. Who is it?

Steve Jobs.

Really? Wasn’t he an idol of theirs?

Yeah. Well, he had just come back to Apple from NeXT. Whether they really meant it or not, I’m sure if Steve had been willing to come be CEO, they probably would have said yes, of course. But I think it was more like a little thumbing of their noses at the VCs: “We’re keeping our bar high. It’s Steve Jobs or nothing.”

Wow. So great. Also deeply ironic, given what was to come between Apple and Google 10 years later.

Yeah, but that is for the next episode. So anyway, it was a pretty contentious process through all of it. 16, 17, 18 months in, finally Eric Schmidt emerges as probably the only viable candidate out there. And I think Eric was acceptable to both sides, both because he was an actual engineer who had been at Sun and because he was CEO of Novell. He was a business person, too. He’d been a CEO, been a CEO of a public company, and he famously hit the Venn diagram of everything.

He also went to Burning Man, as did Larry and Sergey, right?

And so Eric joined in March of 2001. Again, I think Larry and Sergey were still kind of resentful of the process. I think it did work pretty well, and Larry especially realized, “Hey, there are parts of being a CEO, especially as we’re getting bigger, that I don’t really like, and Eric can do those things. I don’t really want to run a finance org. I can have Eric do those things.” And it ended up working really well at a critical moment for the company, where they needed revenue, they needed to build a business, and they needed to scale.

Yeah. And the 3 of them kind of ran the company together. I think they had a daily standing meeting. So it wasn’t like there was a CEO that took over and put the founders out to pasture. It was CEO, and then Larry was president of products and Sergey was president of technology. But really, it was like there are 3 people running this company together.

Yep. And a trusted relationship between 3 people is just more manpower than 2 people.

Yeah. And the organization at this point was so uniquely Googley that there would have been organ rejection if Eric tried to take a heavier hand. We haven’t talked about Googleyness yet.

Yes. Let’s talk about Googleyness.

He really came in with a lens toward learning and understanding. There was somebody who decided very early in his tenure, maybe even on his first day, to move into his office with him because there wasn't enough space anywhere else. So he was camped out like an engineer at Google. He had an officemate, an engineer, for many months, and it's sort of this Googleyness. It was a little bit of an acid test for him.

Yeah. Right. But this giant worldview: let's solve big problems together. Can we think bigger? No matter how crazy the solution, if it sounds like a good idea, it's worth running down. Googleyness is kind of utopian in a way that makes all other companies look almost like an evil empire. It feels like a university in a lot of ways.

Yeah, that was the culture there. They wanted the mentality of a campus, too, where they wanted inexperienced people who didn't know what they didn't know, so they tried novel approaches to problems. They collaborated more than they otherwise would have, but they were really high horsepower.

Yeah. Everyone there was ludicrously high IQ from the very beginning. But I think that sort of collaborative, utopian thing went along with the IQ.

The phrase that I heard a lot in the research, from talking to folks who were early at Google, was “a healthy disregard for the impossible.” That was the modus operandi there. It's this culture that comes up with a mission statement to organize the world's information and make it universally accessible and useful. This was in 1999, in their very first press release after the financing, and that has been the mission statement.

It's also amazing how much that mission statement scaled.

Yes, I was going to save this for way later in the analysis, but “organize the world's information”: not too broad, not too narrow, in many ways altruistic to attract the right type of talent that you want, but also one that lends itself to tremendous monetization. If you're going to organize the world's information and you have a bunch of smart people, you are going to be able to create a money-printing machine based on organizing the world's information.

They actually have a great quote in their IPO prospectus: “We believe that the most effective and ultimately the most profitable way to accomplish our mission is to put the needs of our users first.” So there's this almost trifecta of a wonderfully altruistic-sounding mission: A, that B lends itself to this incredible monetization model, and C, as long as we're putting the needs of our users first.

Yep.

I think Eric really bought into this because it was a risk. Even though he joined Google in 2001 and a lot of these portal deals were already underway—the Yahoo portal deal had already happened—it wasn't clear that Google was going to be a smash-hit home run. I think it was clear that it was going to survive, and they had enough revenue and they could be profitable. But we're talking about somebody who was the CEO of a public company, Novell, and taking a risk to come back to a private company that, yes, had a lot of usage, but startups were out of favor now. I think it was really him making a bet, too: no, this is what I want, and I'm going to buy into this.

Yep. Totally agree. All right. So, what was Eric walking into here with Google in, call it, spring of 2001? We've now gotten the Yahoo portal deal done. Basically, the ship has been stabilized and the company saved. Google's going to survive the dot-com crash, between the $10 million investment from Yahoo and the revenue from that portal deal.

Eric hasn't started yet, but Larry and Sergey now turn their attention back to ads, and they're really not happy with the current state of play with ads. Even though it's working to a certain extent and advertisers are happy, there are a bunch of problems with it.

1. It's all still hand-sold on Madison Avenue. The pool of potential advertisers is nowhere near as big as the pool of potential searchers and the intent that's happening on Google, right? They can't really scale this business, and so it would require getting an enormous amount of spend from each of the small number of customers they already have.

Yep. And then scale, too. Another reason it's not going to scale well is that it's all sold by hand. So as you scale the business and the number of advertisers, you're going to need to scale the number of people you need to sell to by hand. That sucks. Then you end up looking just like Yahoo.

Yeah. So that's on the scaling side. Then on the experience side, the user-experience side, there's no notion of ad quality here. Google, as a value proposition to its users, is: we give you the highest-quality, most efficient, best search results possible. We help your needs the best, and the ads aren't really lining up with that. There's no way to make sure that they're good.

Yes, exactly. So that's a problem. And then 4, Google's just flat-out leaving money on the table. They're giving advertisers this great product of, hey, we have the intent of people searching for these keywords, but Google is just getting paid on a straight CPM basis for what they're selling. They're not participating in the economic value, and they're pricing a little bit finger-in-the-air on what the price of any given keyword should be.

Exactly. So now, fall of 2000, they're like, “Okay, let's address this.”

Yes. So all 4 of those issues are things that Google is going to address in this next evolution of AdWords. But there's a whole part of the world that heavily inspired AdWords v2.

8. Overture Reinvents Paid Search

Yeah. AdWords v2, you might say, is Google's Instagram Stories moment.

Yes. There was an innovator in the space called Overture, or its original name, GoTo.com. We should tell you that story now.

So GoTo.com—Bill Gross started the company out of his startup incubator, Idealab, and he did it with quite a bit of flair, coming to the world from the TED conference in February of 1998. So, same time as Google was about to launch, right? At the time, existing search engines, as you'll remember, had a problem. This is the same exact problem that Larry and Sergey recognized: quality was going down.

In the old world, keyword-matching algorithms were fine. There was no one gaming the algorithms, there wasn't a lot of real commercial activity yet, and search engines weren't well understood yet. So the old, “Hey, go search for dogs, and the most relevant website is probably the one that says dogs the most,” still kind of worked.

Yeah. So now you're starting to get, in 1998, all this stuff like keyword stuffing, white text on a white background, people getting porn sites to appear in search results no matter what you're searching for, hijacking traffic, all that sort of stuff.

So Bill had this very radical idea: the best search results should be determined by the free market, with dollars. Whoever is willing to pay the most is probably the very best search result for your given query. Spammers who aren't relevant to your search can't afford to pay because there's not going to be super-high conversion, but super-legitimate businesses that would actually solve the pain point that you're searching for could.

It's just like how the Yellow Pages in the phone book had paid inclusion as a philosophy that would lead to only the most relevant listings for any given category, right? At the time, this was a completely crazy idea, but when you think about it, it actually does make sense.

If I have a product or service that can solve the need you're expressing through your intent in the search, I should be willing to pay more than anybody else to meet your needs.

Absolutely. It's just a different way of solving ranking and relevance than Larry did. Larry and Sergey sort of figured it out on the organic side, and Bill sort of figured it out on the paid side.

So Bill went so far as to say that GoTo didn't actually develop any organic search technology on their own. It was all paid.

Yes, only on paid listings. If you kept scrolling, they actually did show organic results, but they would license them from Inktomi and others, like you were saying, David, as a backfill.

All right. So the net of all this: on the TED stage, Bill gets wildly criticized for this. Some people even booed the idea when he was on stage at TED, but crazily, Bill's idea was basically right, and it had a ton of ideas that would become a part of Google that we'll talk about here in a minute.

Here's how it worked. When you searched, GoTo would show you a list of the paid results exactly in order of who paid the most, with no fanciness at all beyond that. They would show you the price that someone was willing to pay for your click. So right there on the page, you could see 21 cents, 23 cents, 24 cents.

Yeah, it was fully transparent.

Yes. So insight number 1: paid ads on keywords, auctioned off to the highest bidder, showing up first.

Insight number 2—and again, this is way back in 1998—was that this whole cost-per-thousand-impressions thing was wrong, and that eventually he thought the whole world was going to move beyond this to a cost-per-click, or pay-per-click, pricing. So he thought, why not just do it today? GoTo advertisers only had to pay when a user actually clicked.

And the origin of this is, since Bill had a bunch of companies at Idealab, he could uniquely feel this pain point. He sort of hated the fact that he was getting billed for all these impressions at his companies when he just wanted to pay for the actual clicks.

I mean, this is how advertising worked throughout all of human history to this point. There's that famous John Wanamaker quote: “Half the money I spend on advertising is wasted. The problem is, I just don't know which half.”

This new model of performance-based advertising wasn't possible until the internet, when you could track clicks and conversions. But now, all of a sudden, as an advertiser, you don't have to worry anymore about what's wasted. You know it's all performing.

And the nuance is CPM actually works fine in brand-building situations, but on conversion, you actually care about the click.

So basically, with cost per click, you're getting free exposure every time your ad shows up but nobody clicks on it. But in a high-intent environment, you're not trying to get exposure. You're trying to actually capture the clicks. So it's kind of reasonable the way that it shook out that a lot of brand-based advertising is still CPM-based, but on search engines, it totally should be CPC.

How did it go? Well, it worked insanely well out of the gate. GoTo did $100 million in revenue in 1 year—way more than Google. Yes, this is a good business model. They had found, by the way, that this was also self-serve. There was a website where you, as an advertiser, could log in and place a bid. There was an auction that happened, a real-time auction, where the person with the highest bid, again placed through the website, was at the very top. Does this sound familiar to anyone who's used Google's advertising tools? So the $100 million happened in year 1. By mid-1999, they had 8,000 advertisers.

Compare that against what AdWords had when it launched in October 2000: 350 advertisers in the beta program. To your comment about scale, David, this can just scale to so many more advertisers.

GoTo went public within a year.

Yeah. Isn't this crazy? You're probably sitting there thinking, how are they not the dominant player?

So one thing that did not happen was patents. They did not patent the idea of the auction or of pay-per-click. And I got the chance to talk to Bill when we were prepping for this episode. He's very direct about all this, very reflective, and also a brilliant guy. He just thought they were obvious. He just thought, this is the way it should be done. Of course it should be billed per click. Of course there should be an auction, and the highest bidder is the one that wins.

The nuts and bolts of it are that right before going public, the lawyers flagged, "Hey, you really should patent some of this." But they were just outside the window of what was patentable because he had shared them more than 1 year earlier onstage at TED. So the ideas were no longer eligible.

The TED conference. That's amazing.

There's some really interesting background to all this, too. You would think, of course, Bill was right. This stuff is obvious. Why had nobody tried this until 1998? Somebody actually had tried this earlier. There was a search engine called OpenText that did try paid search results in 1996, but the internet was still enough of a utopian community—small enough, and sort of an outgrowth of academia—that people booed Bill Gross and GoTo onstage at TED in 1998. In 1996, when OpenText tried to do this, it was like they got kneecapped right away.

Heresy.

Yeah, it was heresy. And because that happened, everybody else had a hangover from it, like, "Oh, that's a third rail. You can't touch that. Internet users will never tolerate paid search," right? So it's funny. Maybe they wouldn't have gotten the patents anyway, since OpenText was doing it before, but that was the ethos of that early web: how dare you litter our organic results with your paid inclusion, putting these ads front and center. Look, originally Larry and Sergey were thinking this, too, right?

Yeah.

The great irony is all this criticism. We're going to flash forward for a second. When Google does launch AdWords v2, there's a sidebar with a separate color. It looks super different. The word "Sponsored" is very clear. You are very aware that you're looking at a whole separate pane over there. That's the paid, icky world relative to my beautiful, clean Google search organic results.

Anyone who's used Google in the last few years knows the world basically ended up exactly the way Bill Gross envisioned. It's 1 column of results. The first few are sponsored. In Google's case, they label them even less than Bill was labeling them at GoTo, and then it's followed by the organic results after that. So what was once criticized as absolutely heretical has come to become basically the dominant model of search and search monetization today.

Yeah. But the interesting thing is the timing was not right in the mid-'90s for this.

Yes. By the time the bubble was sort of fully inflated, the internet had become commercial enough that, hey, it was okay for Bill to try this. And then he and GoTo—Overture—set the example of, like, "Oh, this is how you're going to monetize search. This is really how you're going to monetize the internet." And then Google can look and see, "Oh, maybe we should do that, too."

So, a couple of quick things. Overture did file some smaller patents on the self-serve tools. Google did eventually end up owing them $360 million for infringing. But these big ideas—CPC, auction—are now out there, given to the world for free.

Within the next 2 years, they realize that they can take this paid-search model they have and bring it to portals, too. So just like Google started doing organic portal deals, GoTo starts doing paid portal deals. This goes so well, they become a B2B company. They rebrand. This is when they switch from GoTo to Overture. They start powering the ads for Dogpile and MetaCrawler. Then they get to the big boys with AOL and MSN, and eventually they get Yahoo. Yahoo alone was a $100 million deal.

Google's playing over here in fun, pennies-on-the-ground land, where they're saying, "Please, sir, give me some money" for the organic results, and meanwhile Overture has it figured out: these paid results. We are doing massive, massive white-label rev-share deals. Once they get to Yahoo, some huge percentage of Yahoo's overall company revenue becomes paid-search ads powered by Overture.

Right. Yes, I think it's like 75%.

So let's just flash all the way forward to this. Yahoo ends up buying Overture for $1.6 billion. There's a little bit of a bidding war back and forth with Microsoft, but that's the final price. Yahoo basically says, "We have to own this thing." I mean, it is our revenue. And Yahoo's market cap had gotten decimated when the bubble popped. So this was a large portion of Yahoo's market cap that they spent for Overture.

Yep. But what choice did they have? They were over a barrel. The majority of their revenue was coming from this vendor who was revenue-sharing with them.

Yep.

Okay, so, David, to end the Overture story before we go over to what Google learned from all this and started implementing, some fun trivia: did you know that GoTo tried to acquire Google?

I did not know that. So here's how it went down. I asked Bill about this. Bill thought it was a match made in heaven. Google's got the best way to bring relevant organic search with PageRank—really amazing for informational, noncommercial searches. And GoTo has this amazing paid system for the commercial searches. You should totally have 1 system that marries informational queries and commercial queries together. It's got the 2 best ways to surface relevant things to you: 1 paid, 1 organic. And Larry and Sergey, before they raised the Sequoia and Kleiner round, came to Bill and said, "What about $200 million?"

Wow.

Bill thinks, "Actually, seems fine. This seems fair. You guys are really onto something."

They had a chance of getting acquired for 2× the valuation of that entire fundraise, right? Wow. Was Overture already public at this point?

Yes. So then Bill goes to the rest of the Overture board, and Overture at the time is worth $2 billion. And the board—their conclusion is basically, how could we give up 10% of our very important, valuable, revenue-generating company to this little company with 0 revenue? It'd be a dilutive transaction, and so no deal.

Well, there's almost 0 chance that Google becomes Google if that deal had happened.

Yeah, exactly. That's the thing with these what-would-have-happened-otherwise acquisitions.

Yeah, that's amazing.

Well, okay. So back to fall of 2000, when Larry and Sergey and Google can now finally focus on improving their ads product. I think we heard this from folks in the research: once they saw how well the GoTo and Overture model was working, I think Larry and Sergey were kicking themselves, like, "Ah, we should have just done this from the beginning. Why did we waste time doing this the other way?" And, like, yeah, it's going to take a lot of technology to build this out, and, like, yeah, we're going to have to focus on it, but it's obviously the better business.

Obviously, Larry and Sergey—geniuses from their childhood through their undergrad research projects, the way that they conceptualized the original PageRank algorithm, everything—truly geniuses. But the second superpower on top of that is it doesn't always need to be their idea. They're very good at hearing the best idea, whether it's from outside of Google or someone else inside Google, and adopting that and making that the thing that they run with.

October 2000, they put Salar Kamangar on the project to improve AdWords. The first obvious thing that they need to do is they need to build a self-serve system. As long as Google is still taking manual orders for ads, they're not going to be able to implement any of the technology to bill people per click or anything like that, or let in smaller advertisers and expand the pool.

Exactly. Which clearly Overture had shown there was a market for this. If they had 8,000 advertisers against Google's couple hundred that they were selling by hand.

Yes. Famously, by the way, Tim Draper was looking to invest in Overture and eventually did lead their round. And to test it out, he actually opened up his computer and bid on the keyword "VC" when they were pitching him.

That's amazing. That's the historical proof that I have that Overture had self-serve, and then he got outbid. It was like 1 penny, 2 pennies, and then he started a bidding war over the term "VC." Ah, that's such a great story. I love it.

Okay, so obviously they need self-serve. That's the first thing to work on. But Salar and Larry behind the scenes, too, are clearly thinking, like, "Okay, how do we do this in a Googley way?" Yes, we're going to borrow a lot from Overture, but I think they had a Spidey sense already that that wasn't Googley, you know? We're so pure over here. But also that it wasn't quite right—that Overture had gotten three-quarters of the way there on cracking the business model.

Yep. And so the thing that Salar and the team really start noodling on is: we've got this beautiful algorithm in PageRank that can deliver highly relevant organic results. Is there a way that we could incorporate something like that into our ad system as well and ensure ad quality? Yes, the paid system in and of itself goes a long way toward ensuring ad quality, Ben, as you were talking about earlier, but there's still potential for abuse here. What can we do to really make sure that these things are good?

Well, okay, if we're an online self-serve system, we're measuring clicks, we're going to ultimately switch to pay-per-click, and we could track those click-through rates. What if we made that a signal to the ranking of how we show the ads? I mean, it's not just fully pay-to-play, where if you pay the most, you get placed at the top, but actually we incorporate, as part of our ad-ranking system, how effective your ads are in terms of click-through rate. That might solve the problem.

Such is the birth of Ad Rank. You've got PageRank, which uses all the clever things we talked about earlier—the number of people linking to you and how authoritative those sources are—for the organic results, to make sure that the most relevant results are being surfaced to you. Now we have a way, over in the paid side of the house, with Ad Rank, to take all the great stuff that we just talked about with Overture—the self-serve model, the auction, the cost-per-click-based system—and add in click-through rate and feed it back into the algorithm.

Ad Rank is really the main 2 things that go into where your ad is going to be positioned in the ranking. It's both how much you're willing to bid and how often users are actually clicking through, so they know that it's the right ad to be showing at that right moment. Click-through rates are a proxy for relevance. And by the way, as a really nice side benefit of that, if your formula for placing ads is a combination of the price that an advertiser is willing to pay per click and the click-through rate of the ad, well, that's actually the mathematically optimal formula for maximizing your own revenue as Google.

Oh, that's interesting. The highest price paid per click and then the highest likelihood to click—that is the ad that you should show to maximize your own revenue. Oh, it's basically an expected-value calculation.

Exactly. Oh, that's funny. But it's also perfect for advertisers, because it means that if you're a better advertiser for that keyword, then you actually get to pay a lower price. If people are more likely to click through to your service and transact on your product, you get the privilege of bidding lower prices and still winning the auction. All incentives are aligned for the user.

Oh, and for the user, because then it means that the user is only ever seeing products that are the most relevant.

Yep. So, it's funny how all of this gets rolled out. It's fall of 2000. They start working on this. The 1st version they launch includes self-serve and includes ad quality, but it doesn't yet include CPC or the auction, which is funny. It's sort of like they did the hardest technical stuff first.

Interesting.

So this 1st rev in the fall of 2000 attracts a ton of advertisers. You've now opened the floodgates to the long tail of advertisers, and you've introduced this click-through-rate element—this ad-quality element, Ad Rank—to how ads are going to get served.

Advertisers pretty quickly figure out they're still paying on a CPM basis, not on a per-click basis. They figure out that they can game the system by clicking on their own ads, because that will boost the click-through rate and then their ads will get shown more.

That's so funny. And they're not paying per click, so they're not costing themselves money when they're clicking on the ads.

Right. It's actually an efficient use of impressions to use them internally to boost click-through rate. Exactly. So, for a set of months, the greatest arbitrage in the history of the internet was to click on your own ads.

That's so funny—on Google.

But it proves that, okay, this is going to work. So then they fully borrow the rest of the model from Overture. We're now into 2001. They borrow the rest of the model with the cost-per-click payment basis and the auction model.

Now, interestingly, they one-up Overture on the auction model. They go to the second-price auction.

This is such a genius mechanic. There are 8 genius mechanics we've talked about so far in the episode, but this one really sings.

If you're the winner of an auction, they make it so you never have to pay anything more than 1 penny above 2nd place. So, let's say I bid 20 cents, you bid 30 cents, and then some other guy bids 50 cents. Well, that other guy is going to win, but he's only going to have to pay 31 cents.

And you might say, well, that's silly for Google. They're leaving money on the table. But what Google is thinking is with a much longer lens and saying, "Well, we'd rather have our advertisers, A, trust us and feel like we're not gouging and, B, not feel like they have to constantly check and fiddle and look to see who the other bidders are and whether they want to adjust their price." It's actually a long-term value-maximizing thing to do, even though in the short run, of course, you're leaving pennies on the table each click.

It's a version of this theory you have, Ben, that you and I have been talking about: that every great company has a stored potential energy of value maximization that it doesn't fully maximize. Costco is the extreme example of this, but Google has this too.

That's a great point.

Yeah, the second-price auction is storing potential energy in a way. So, there's a fun story around this. As you can imagine, this is a little hard to explain to advertisers when they roll it out—how this works, why it's going to be good for them, et cetera.

They used to just write 1 check and send a fax, and that check bought them a big batch of impressions. Now you have this confusing morass, right? So, there's a fun little story about how to educate advertisers around how this model works.

All of this—the full version of AdWords that we know today—had launched at the very beginning of 2002. Sheryl Sandberg had joined the company right around that same time, too. Part of her job was to pitch to advertisers what this new model was and explain it to them. She's banging her head against the wall. It's so hard to do.

And so she calls up her mentor, Larry Summers, who had previously been the U.S. Treasury Secretary, and Sheryl had been his chief of staff there. She's like, "Larry, I'm having a hard time explaining this to advertisers. You know how this model works, the second-price auction. It's weird." And Larry's like, "Oh, this is what's called a Vickrey second-price auction. There's a lot of economic literature about this. This is the optimal way to do auctions, and this is actually how the Federal Reserve sells its Treasury bonds. You should just tell advertisers that."

So she does it. I don't think it sinks in right away, but eventually advertisers get the message.

That's funny. I do know this was very painful for Google to do—to transition all their advertisers over to this new model. They actually called it Project Sunset, where they had to sunset them off the old model and bring them onto the new pricing. Even though it was better for everyone, it's miserable along the way.

I have a funny coda to the whole Overture thing.

Great. Go for it.

So, I think it is correct that Google, and particularly Salar, led the charge on the insight that click-through rate is really important and factoring it back in is relevant. Overture did also figure it out earlier.

The problem was that after the team implemented it, advertisers, as you would imagine, no longer knew how much to bid. Overture's whole thing was, "We have this transparent thing on the page where we show the prices," and whenever you loaded up a page, now the prices were out of order. You're like, "Well, what am I supposed to bid to get the top spot?" Also, it makes it plain and clear—you guys are supposed to be transparent, and now you're this confusing black box.

Overture did not take the pain of transitioning over to this new model, and they just abandoned it and said, "We're not going to mess with this click-through-rate thing," which ended up being crucially important to the model.

This also highlights something that we would be remiss not to say as well. The technical infrastructure to dynamically execute second-price auctions every single time a user is making a search query was incredible, especially at Google scale in 2001, with the technology available then.

Yes. I totally believe that Overture did try to implement it, had the same idea, and also that Overture had really good technology too. I think that is true. But again, back to Google's infrastructure—the commodity hardware and the scaling out of data centers, the distributed file system, and distributed computing—to really scale this, you needed special infrastructure that only they had.

So, okay, we're talking about the pain of transitioning the Google ad model and ultimately the whole business over to this new, beautiful AdWords model. To put some numbers on it, in 2001, the year that the Yahoo deal saved the company, Google ultimately did $86 million in revenue that year and $10 million in profit. So, it turned profitable in 2001.

Great numbers by any metric for a startup, and especially in the middle of the dot-com winter, but almost all of that revenue was the combination of these portal deals and the old ad system that was in place. So, Project Sunset and transitioning over to the new ad system put a large portion of that revenue at risk.

Oh, yeah. It was not necessarily an easy decision. I mean, it was an easy decision because the performance was so clearly better, and eventually economic incentives would kick in, but it was still a little tenuous there in Googleland for a while.

The other thing that's happening at this exact same time is Eric Schmidt arrives and discovers that 50% of Google's searches are outside the US, but they have no international ad sales. There's no international business.

So he almost jokingly tells Omid, "Just go get on a plane. Go to the airport Monday morning, and I'll call you and tell you what market to buy a ticket to. We'll just kind of go from there. Or just pick a country, and we'll figure it out."

They basically build these little startup teams in a bunch of different geographies that act as their own companies, selling Google ads on the new system. It basically works year 1 because Omid spent the whole year on a plane.

18% of revenue is now international. In 2002, it grew to 22%. In 2003, it grew to 29%. Today, it's half of Google's business.

And they even had a business in China for a long time, until famously they got into a fight with China. That started in—when was that? 2002, 2003, 2004, something like that.

Yeah. Eventually, Google finally just withdrew in 2010 rather than censoring results.

But basically, everywhere that is not China, starting in 2001 with Schmidt being like, "We need an international business," they grew themselves a just-fine international business.

Yep. Yep. Yep.

So 2002 is this year of transition to AdWords for the company. Spoiler alert: it worked. In 2002, the company did $440 million of revenue, so up like—whatever that is, 5 or 6×—from the $86 million they did the year before, while transitioning all of that revenue to the new model.

And when I say all of that revenue, I really do mean all of it, because a lot of that $86 million, remember, was the portal partnership deals. By midway through 2002, Google is realizing that paid search and AdWords is working so well, we should stop having portals pay us for organic search. We should start paying them to do paid search on their sites and share the ad revenue.

And that leads to the landmark summer 2002 deal with AOL.

9. Google Bets the Company

Okay, so the AOL deal—and one context-setting thing, just to see how fast Google's world changed in 2002 with the new AdWords system going phenomenally well—as recently as late 2001, just months before, there's a dinner with Terry Semel, who's the CEO of Yahoo, a big media executive guy who comes in and takes over as the big-media CEO. And he sits down with the founders and says, "So, guys, I think we're your biggest customer, right? Us paying you for those organic results in the portal deal?"

And they say, "Yep." And he says, "That's less than $10 million that we're paying you. So you don't have a business, do you? The business can't be that big if we're your biggest customer at less than $10 million." And everyone's all excited about them, and they're like, "We're excited about some of the things we have in the works." They're clearly thinking about this new AdWords system that they're going to launch on the spot.

He offers to buy Google for $1 billion.

Wow.

Even knowing you aren't doing much revenue. And this is at a time when Google was so secretive pre-IPO. I mean, no one else in the business knew about them. Terry Semel sits down; he didn't have a real sense of their revenue. He just knew that he was the biggest customer.

So I think this kind of illustrates just what an insane 12 months it was to go from being in that position to the numbers that you just shared. What is it, 5× revenue in a year?

Yep. They more than 5×ed revenue from 2001 to 2002, from $86 million to $440 million in revenue.

All right. So they're feeling real good about this, and they decide to go to AOL, the big source of traffic. I mean, AOL has 34 million users at this point. It's funny imagining this: AOL was still a big and important company, but if they could figure out how to be the search results provider and, more importantly, the ad provider for AOL, that is a potentially company-making event.

Yes. I mean, hell, in 2002, I was probably just transitioning from using AOL as my way to access the internet as a senior in high school. I think we had maybe just gotten broadband that year, maybe the year before.

Funny how sometimes things feel like a lifetime ago, and sometimes they feel like just yesterday.

Right. Right.

So, yeah. Okay. Summer 2002, the transition to the new AdWords model is blowing the doors off. Google goes to AOL, which, by the way, Yahoo sees, and they come back and say, "How about $3 billion?" I had heard about that $3 billion number. Google comes back and says, "How about $5 billion?" And, of course, no deal gets done. This would be the last time that they seriously try to acquire the company.

Yeah. But then Yahoo would buy Overture. But, as I pointed out a minute ago, when Yahoo bought Overture for $1.6 billion, that was a huge portion of Yahoo's market cap. If they had actually done the $5 billion price that the Google founders floated at them, it would have been a reverse takeover. It would have been Google taking over Yahoo.

Oh, that's a good point.

In fact, Google throwing out $5 billion is almost a farce. It's like, "How about we buy you?" It's like it's a counter-offering.

Yeah, we'll buy you. That's so funny.

So, this AOL deal. The current state of things is that the 34 million AOL users' search experience is powered by Inktomi on the organic side since 1999, so the last 3 years, and Overture for the last 2 years. It's sort of a bake-off: do we want those 2, or do we want just 1 to take over all of it, since Google seems to have kind of the whole package?

Now, Google wins the deal. Here's the shape of the deal, and then we can kind of talk about the philosophy behind it, but it works all ways. So Google can sign them up, or AOL can say, "Hey, advertiser, I know we have a long-standing relationship. You can buy ads on our properties other than search, but for search, here's the URL you go to to place your order with Google."

Okay. Go through the rest of the deal points, but this is huge.

That is huge. Google will then share back 85 cents on the dollar to AOL for all of that revenue.

AOL wants 2 things in exchange for turning over their entire business in the search advertising world to Google. One is warrant coverage. What they end up getting granted is the option to buy 7.4 million shares of Google at $3 per share, so a total of a $22 million investment. They get that as part of the deal.

Two is a $100 million revenue guarantee. We want to make sure that, hey, even if this whole thing falls apart, you're going to pay us at least $100 million, and hopefully more if these ads perform well and we're getting 85%.

Yep.

So here's the crazy thing: Google doesn't have $100 million. When they're negotiating this deal in May of 2002, it's just starting to work.

Sergey Brin has a quote. He says, "We could have gone bankrupt." This is quite literally Google betting the company.

And the way to think about it is financial leverage. They took on a fixed-dollar obligation with that revenue guarantee to AOL. If there's upside for Google, it would have been a huge, huge win.

Yeah, if they couldn't make it work on AOL, right?

They obviously have very high confidence that it would. But if something happens and they're like, "Oh, shoot, we actually can't sell these things at the rates that we thought," it would have gone from "Oh, shucks, bummer" to "Now that we've signed this deal, we're bankrupt."

Yeah, this was a really contentious decision. And I think it ultimately came down to Larry and Sergey pushing for this.

You are absolutely right. There's a great quote in Ken Auletta's book about this, where Omid says, "You're betting the company if you do that." And Larry Page responds, "We should be able to monetize the pages. If not, we deserve to go out of business."

That's great.

So, yeah, those are the deal terms, but that first one—advertisers are going to use Google's system—this is why it's worth betting the company.

This is when Google discovered what we talked about on our Meta episode, when Boz—Andrew Bosworth, the CTO—had the insight that more ads equals better ads and then argued to Zuck and Sheryl, “Then we need to show more ads in feeds, and then they’ll get better.” The more ad inventory you have in your system, if you’re serving ads dynamically based on a ranking and targeting them, you want to have as much inventory as possible to give you as many candidates to choose the best ad to serve. And so onboarding all of AOL’s search ad inventory into Google’s system was hugely strategically valuable.

Yeah, it’s a market liquidity thing.

Yes, exactly. Basically, the more volume you have in your market, the more deeply traded the market, the more likely you are to have an ad that has perfect product-market fit with the query. If you have a thin pool of advertisers—and several folks I talked to at Google made this point to me—if Google had come out of the gates with the AdWords business model in all its glory, which it ultimately became, it would have been very hard to bootstrap from a cold start because you would have had this inventory problem. You wouldn’t have been able to deliver the magical, high-quality ad experience because you would have had a very thin inventory of advertisers. You almost had to bootstrap it up how they did and then onboard this other supply into the marketplace to get deep liquidity.

It’s funny. I just want to pause for one second. When you say “magical,” there’s nothing magical about search ads, but I think you’re right that they’re the least offensive. They’re the most likely to be what I’m looking for without giving me any delight whatsoever.

Sorry, I meant magical in an economic sense. It is the most magical economic transaction I think ever known to man. Right? I mean, for an advertiser, you are reaching the exact right person at the exact right time, when they have the most intent possible to find your service. It is actually a pretty magical economic lever.

Well, to your point at the beginning of the episode, Google, with this business model, makes more profits than any other company.

Yeah. Ergo, logically, it’s the most magical business model ever discovered.

Right.

So, the interesting comment on this was that they commented to The Wall Street Journal on AOL’s decision: “They’ll learn over time that Google takes your users. It doesn’t help you build your property,” which wasn’t wrong. I mean, how many people use any portal today versus how many people use Google directly today?

So, how did this actually go? They made this huge bet. They put $100 million on the line. You better be really, really sure that you can come through when you’re betting your company.

It worked. It worked. AOL made $35 million in 2002—the first half-year of the deal alone—and then in 2003 made $200 million.

Wow.

Yeah. Yeah, they blew through the guarantee there. Absolutely. This made Google a major player in the paid listings market almost overnight, and they weren’t a player at all before this. Overture dominated this market before. So, this is an absolute bet-the-company move that couldn’t have gone better, and they were taking all that inventory effectively from Overture, who was AOL’s partner before this.

That’s exactly right.

This is also where Sheryl Sandberg really makes her mark. She joined Google, as you mentioned, sometime in the last year, right around when Eric Schmidt joined or right after that, and she was looking for the right job to do. She was poking around the company. I think she was a business-unit manager or something like that, but Google didn’t really have business units. And so Omid sits down with her and says, “You’re looking for a big job, right?” And she says, “Yes.” And he says, “We have this huge AOL deal that we just signed. We have a ton of new advertisers in a bunch of categories that we have no idea how to service. We need an army of people to handle these thousands of new advertisers, and they have to be smart and adaptable because we have no systems built for this yet.”

Then, over time, they’re going to have to figure out how to scale themselves so that we’re not constantly hiring more people. They need to feed ideas into our technology organizations to make it so that we get more leverage off of the people that we hire. And she basically hired all these great people, built out the entire AdWords sales function to service this monster AOL deal. That was what she did at Google before going on to become COO of Facebook. We’ll talk about that on the next Google episode here.

Yes.

This also, I think, as Google is digesting this deal and realizing the huge strategic value of it, gives them license to then go play offense on traffic acquisition everywhere. Basically, the light bulb now goes off: if we can have Google Search, paid and organic, be part of the user experience anywhere on the internet, we should do everything possible to do that because it will build our liquidity pool and our business, and we will just monetize the internet.

Oh boy, will it ever. David, you want to do distribution? Do you want to go there right now?

Oh yeah, let’s go there. Great. All right. I’ve been chomping at the bit. So, we’re going to talk about all the crazy stuff they did for distribution. But before that, it’s worth a discussion of the business model of search. We’ve been talking about it all episode, but there’s a very particular, unique characteristic that, once you realize it, completely changes how you should think about distribution.

So, search is a winner-take-all market. And not just because it’s large and consumer-facing and it’s horizontal across industries. There’s something more to it than that—a second layer. They’ve got all the traditional economies of scale that you’d expect. If you make 1,000 of a widget, you get cheaper pricing than if you make 10 of that widget. So, just like everyone else, they amortize the fixed costs of their infrastructure and their employees, and they have a better infrastructure model, as we were talking about earlier, et cetera, et cetera.

Exactly.

But there’s this crazy thing that happens with Google where, at scale, not only do their costs decrease on a unit basis, their revenue actually increases per unit. So, here’s what I mean by that. When you have more bidders on every keyword, you have better price discovery in that little market, and the winning bid is a higher price than it would be if they had fewer bidders.

Ah, this is another reason why you want a deeper market liquidity pool.

Yes, exactly. The second thing, too, is you have bidders on keywords that are less common. So, let’s say you’ve just got the 100 biggest advertisers in the world. You only get to monetize some of your searches. But if you’ve got a big, long tail of advertisers—or just a lot of advertisers—then you get to monetize more of your searches. So, it’s more likely that any given search results in revenue. Having marketplace liquidity means you always generate the most revenue per search versus other smaller search engines. So, it’s not just that unit costs go down; it’s that, as they scale, their revenue per search actually goes up due to the auction system.

Yeah, there’s a third statement that you need to add to Boz’s insight from the Facebook days of “more ads equals better ads.” It’s “more ads equals better ads equals better business.”

Absolutely. It is crazy that you make more money per search the bigger you scale.

Yeah. In this auction-based marketplace system, it’s increasing returns to scale.

Exactly. So then, keep following the logic tree. Because each search is worth more, each user is worth more over their lifetime, which means you can pay more than other search engines can to acquire a new user. And once you realize this and get a little bit ahead—which is where Google is right now in history, in that 2002 era, a little bit ahead—you can start pressing your advantage. And once you start doing that, it’s really hard for anyone to catch up.

So, the cycle is: get distribution. We haven’t yet talked about how, but somehow we’ve talked about it a little bit in the portal deals, which drives volume of searches. More searches drives keyword bids. Keyword bids drive up price in auctions. The price creates more revenue for Google. More revenue for Google means they can pay more for distribution. The virtuous cycle obviously goes on. So, the obvious lesson: do not just sit back and let organic growth do its thing, even though they’ve got great organic growth and the best brand in the world. Here in 2001 and 2002, you want to be aggressive and gobble up this market as fast as you can because someone else is going to have this insight, too.

Mm-hmm.

So then, the tactics: what do you do? One: pay massive revenue share to your distribution partners, in some cases up to 100% of the revenues generated. We’ll talk about who the distribution partners are in a second, but even earlier we heard with that AOL deal they were willing to give AOL 85%. That’s a huge split.

Yep.

Yes. So, with some partners, they were incredibly aggressive. They were like, “We’re going to give you all the revenue for a while just to get you on.”

Yes.

Exactly. If Google monetizes each search the most, then their rev share to distributors is going to be better than anyone else’s. Let’s say they give away the same percentage as other people. “Oh, we’re only giving away 70%.” Well, that’s more than someone else’s 70%. So, press that advantage; go 100%. I even heard 1 example where they gave more than 100%. They realized the payback on this was so strong because this property was so valuable to get this distribution.

Yes. Eventually, it just doesn’t make economic sense for a competitor to match your pricing. They literally will run out of money trying to spend the way that you can spend because your monetization per user is so high. Realizing this is a secret weapon.

This is also where being private was nice. Some of the other search engines were public by this point, and so they were reporting very consistent metrics that they wanted to continue reporting.

Google could irrationally do things like, “Eh, we’re going to overpay for distribution in this case,” and they could potentially risk having a worse quarter. Ultimately, Google discovered this property and they had a belief that no one else had, which is that search is going to be really, really big. Not like a $1 billion-big or $10 billion-big market. Search is currently a half-trillion-dollar annual revenue market.

This is a market worth betting everything on, and they had the stomach to invest very, very, very heavily, where others thought, “Geez, is the final payoff actually going to be worth investing in this market?” Google thought it was literally worth any amount of money that they could invest in this, especially in being first and being biggest.

So, we’ve been talking so far about distribution deals in terms of these search deals with portals. Tell us about some of the other crazy stuff they end up doing, because once you realize this, the game just becomes: get users and advertisers at all costs.

That’s exactly right. Currently, people need to know how to type in Google.com. That sucks. It would be really nice if you could get users without having to hear that from a friend, load up a web page to start searching, and hopefully bookmark it. You don’t own a browser.

So, how do you get a Google search box to actually appear in the browser instead of on Google.com? Which, by the way, Microsoft owns the browser right now. If there’s anybody you need to be afraid of figuring out this secret, it’s Microsoft. And Microsoft definitely did figure out this secret.

Google was already starting to pull away from the rest of the market. It would have taken a boatload of money to try to compete with Google even a year or two into this, which almost no one except Microsoft had.

And then, spoiler alert for Part II, who would a few years later try to spend a boatload of money to compete with Google?

Microsoft.

Yes. Okay, so Microsoft has Internet Explorer. Google doesn’t have a browser. What do you do?

It’s December of 2000. We are not talking Chrome territory here. Google Toolbar, baby.

Google Toolbar. Man, when this came up in the research, it was the biggest blast from the past. Man, I love that thing. I had not thought about that in about 15 years.

And holy crap, everybody thought this was just this gift that Google, the benevolent Google gods, bestowed upon the internet ecosystem.

No way. It was a hugely strategic business-model piece for them.

Here’s how it worked. They shipped it super early, in December of 2000. This is 2.5 years after the company was founded, before they had figured out AdWords v2. They had just launched AdWords v1.

It’s both the offense we’re talking about here—go be aggressive, get users—but also defense. Google is paranoid about Microsoft entering and using Internet Explorer as a weapon. If Microsoft owns the browser, they can direct the traffic wherever they want.

So, once Google Toolbar is installed by a user—

Maybe we should, for younger people, explain what toolbars were—what Google Toolbar was.

Yeah. Toolbars were a plugin, the equivalent of a browser extension, that would basically create a bar underneath your bookmarks bar—or, I don’t know if bookmarks bars were even a thing yet—kind of where the bookmarks bar is, at the top of the window. It had a little Google search box in it, along with some other functionality, and you could just search right from the toolbar without having to go to the website.

Right. Nowadays, every modern browser lets you search from the bar at the top of the browser. There used to be 2 different things: there was a URL bar, and at first that was all there was. Eventually, they put in a search field inspired by the Google Toolbar.

Here’s the economics on how it all works. Once Google Toolbar was installed, a user averaged 7 times the number of searches, which obviously makes them 7 times more valuable. That means you could pay a lot of money to get someone to install it.

Yes. So, how did they pay money to get users to install the Google Toolbar? They weren’t paying users.

The average annual revenue generated by a Google user was $2. But with Toolbar, it was $10-plus, even if you’re being conservative.

So, that difference—somewhere around $8 a user, call it—is your budget to play with. Estimates are that Google ended up spending, on average, way less than this for a Google Toolbar install.

But you can understand the amount of lift that they get from a Google Toolbar when you understand, “Wow, it’s worth $8 more per user in this year.” And, by the way, average revenue per user—ARPU—is skyrocketing. It’s growing very quickly.

So, this $8 is just this year. It’s going to become $20, $50, $100.

Yeah. So, Google just paid everyone they possibly could to bundle Google Toolbar with their installer for an application. This includes Adobe. You’re downloading an Adobe app: “Hey, congratulations. You have Google Toolbar.” You don’t know it, but Google just paid Adobe a bunch of money. RealNetworks, same thing. WinZip, same thing. They were hyperaggressive.

And just to be really clear about what’s happening here: when users are downloading programs—what apps used to be called—to run on their computers, Google is paying the maker of that program to include a Trojan-horse payload of the Google Toolbar, which will then become a Trojan horse that lives in your internet browser, in Internet Explorer, and Google will make a lot more money from you.

The most horrible way to describe this is that it’s adware, it’s spyware, it’s a Trojan. But users loved it.

I loved the Google Toolbar.

Totally. But the technique itself—are you going to get into pop-up blocking?

No. Lay it on me.

As they were building the toolbar and thinking, “Users are going to love this because users love Google, and being able to access search is a value proposition in and of itself for this thing,” pop-up ads were a problem on the internet at this point in time. One of the most popular plugins for web browsers was pop-up blockers.

Google decided, “Well, hell, why don’t we make the Google Toolbar also a pop-up blocker?” Just one more incentive to install it and become a sticky Google user.

Yes. Genius.

They even did a deal with Dell directly to make sure that when new PCs shipped with Windows, they shipped with Google Toolbar preinstalled on Internet Explorer. They famously did a deal to become the default search engine in Firefox just as it was becoming popular, which served as Mozilla’s main revenue source for decades.

And this is a great one that’ll be real close to home. David, you remember the Google Earth acquisition?

Oh, yes, I do. Classic Acquired episode.

So, Google’s an ad-based business. They buy Earth. There’s a conversation in Google: “How do we put ads inside Google Earth?” Instead of doing that, they realized that Google Earth was going viral. People were downloading this thing like crazy because it was really cool to just play with a globe on your computer.

And the original Google Earth was a program that ran on your computer. It wasn’t baked into Google Maps, the web app.

Yep. And Google Maps was very lame compared to Google Earth. They did very different functions. Maps was clearly for driving directions. Earth was, “Oh my God, I can zoom in on my house, and it’s all 3D. It’s super cool.”

They just bundled Google Toolbar with the installs of Google Earth, and then it more than paid for itself. They didn’t need to do ads. It was way more than paid for itself.

Yeah, it’s crazy. No need to put ads in Google Earth.

In the mid-2000s, ARPU would eventually grow to $10, $20, $30, and a Google Toolbar user was always stickier than a non-Toolbar user. So, they just had more and more and more budget to play with in acquiring users.

And, not to spoil too much, but obviously this still plays out all the way to the much-debated Apple Safari deal today and the tens of billions of dollars that Google still pays to Apple in traffic acquisition costs.

Yep. So, the takeaway here is: yes, they had the best product; yes, it was fast; yes, it was technically the most competent; and yes, they had this amazing culture. But everyone kind of forgets about the fact that they were so aggressive in distribution deals. They didn’t just let people magically find their way to Google.

Yeah. And it was all so strategic. As I’ve been talking to people over the last month about making this episode and talking to friends, thinking about how to position it, I’ve been like, “This first Google episode feels like when we made the Costco episode.” It’s the same beautiful ballet where every piece of the Google business model works together and reinforces the other pieces, and in concert it creates the best business model of all time.

Yep. And it’s funny: Toolbar happened to be the one that worked, but it wasn’t the only thing they tried. They tried so many desktop applications. They even had one called Google Desktop that would search your desktop and then incorporate those results privately into your web searches—the original Google Enterprise Search application, reincarnated as Google Desktop.

But that’s basically the strategy for all these applications and clients: how do we make you a stickier Google user?

So, do you know the final chapter of this story? In 2004, a new product manager is hired by Google. They come in and take over this applications and clients team that includes Google Toolbar. That product manager is Sundar Pichai.

That’s right.

And that is all we will talk about for Sundar on this episode, but obviously he will come into play much more in the future.

Well, it really highlights how important Google Toolbar was and how secretive the company was about this really being the strategic linchpin of what they were doing—a strategic linchpin.

So, there’s one more business-building story to tell here before we get to the IPO and the end of this episode. It’s another version of this extension of the Google business model, and that is AdSense, which is Google’s second big business line after AdWords on search pages.

Yep. And it’s another brilliant insight into how to extend the strategic Google business model. Before AdSense, Google was limited to making money when a search happened. That was the atomic unit of the Google business model: a search query, which actually doesn’t happen that often. It’s a really valuable thing when it happens because it’s high intent.

But most of the time, someone loads up a page, it’s a website that is not google.com, right? You’re consuming content on the internet much more often, or for a higher share of time, than you are running queries and searches. Queries and searches have high intent, so they’re really, really valuable. But there’s all this other time and content on the internet that Google can’t monetize.

Because of everything they built for PageRank and organic search and understanding what’s on a page, then serving the page as search results, and also for the ad system—for ad targeting and ad quality and predicting click-through rate—they realized that they didn’t really need a query to happen to serve effective ads against what a user is consuming. What if we essentially run a version of the same algorithms on static pages, on publishers’ web pages, and then reverse-serve the keywords that we would have served for a search query that would have landed on that page?

It’s absolutely brilliant. It’s a little bit different because they’re matching the ads to content instead of to intent. It’s sort of trying to fit in with the content around it. But if you’re consuming content, you likely have some future intent around that content, or maybe even loose intent right now.

Yep. And we’ve got this existing pool of advertisers in our system and their ads. We could literally just run the same ads on web pages. So in February of 2003, they have this idea, and Google is still massively inventory-constrained for serving ads. There’s way more demand from advertisers to be serving their ads against queries than there is query supply in the system, so to speak.

Which is why the auction works. If you had way more searches, then everybody would just be bidding 1 and 2 cents on everything all the time and winning. Right. Right. Right. You always want to be supply-constrained as a business.

Yes. So legendary Google engineer Jeff Dean builds AdSense in 6 weeks.

Of course he does. They’re great stories.

They launch it first on Google Groups, and then they want to test it on true third-party websites to see how this works. As they’re testing it, what they decide to do is, “Oh, we’ll just buy display ad space on these other publishers and, rather than running display ads, we’ll serve it however they need to serve it, but we’ll effectively serve a display ad that is just a window into the text ads of AdWords that we’re serving onto that page.”

I remember seeing these for the longest time. When you, a publisher, enabled Google AdSense, you’d get what looked like search results, just 3 across in a banner.

Exactly. That’s what AdSense was in the beginning. Their favorite website for testing this was HowStuffWorks.com, because all of the pages on HowStuffWorks, it turned out, were high-intent, highly commercializable AdWords pages for AdWords queries. If you reverse-engineered the search queries that would have run AdWords against them, they were great pages.

Susan Wojcicki came in as the product manager for this, and it becomes another big business for Google—much lower margin than their own first-party AdWords business, but it adds hundreds of millions of dollars of revenue to Google off the bat.

Yeah, I’ve seen different estimates from different points in time. Sometimes that they share 67% of revenue, sometimes that they share 80% of revenue. But the right way to think about it is that most of the revenue on a click in Google AdSense actually goes to the website publisher, and Google takes the smaller part as their split.

Yes, that’s right. Whereas if you actually own the search results page and you are running first-party ads, you get 100%.

Yeah, but it’s very similar to the portal ad deals that they were doing with AOL and others, sharing revenue with the publisher. It’s the same model, right?

Yeah, it’s a great point. It’s someone else’s traffic. The net result of the AdSense launch is that Google has had a troubled dance over the years, back and forth, with publishers. Are they good for publishers? Are they bad for publishers? What does it mean for the news industry? Blah, blah, blah.

Right in this moment when they launch AdSense, publishers, and especially small ones, love them. “I’m making content on the internet, and all I have to do is drop in some HTML and Google just starts depositing money in my bank account. This is amazing.”

Think about this, too. This is the precursor to the YouTube business model for creators. “Oh, wait. You mean I get to just create content and put it on YouTube and Google deposits money in my account?” This is the first version of that. An ad network running on a thing that you make is a beautiful thing for a small business owner.

Totally. And with all the liquidity of Google’s advertisers. So they launch this thing. Jeff codes it up at the beginning of 2003—call it the end of Q1 in 2003—and they launch AdSense. By the end of the year, just a few months later, it’s doing over $1 million a day in revenue.

That’s crazy.

It’s just crazy how fast this grew. We’re zoomed in on AdSense right now, but we’re still pre-IPO here. The rest of the business is still very much developing. It’s worth bouncing around to a few different parts of Google to share some updates that didn’t fit into the story arc along the way.

One is that 20% time is happening. People are launching all sorts of fun side projects, and there’s Google Labs, which is this really great way that they’re starting to surface this stuff to users. Google News comes out of this.

Google News is legitimately someone’s actual 20% time, their own personal motivations. Particularly after September 11th, there’s this super-strong hunger for people to have a way to get rapidly updating news on a topic.

Yeah, you couldn’t get that before. You had to go to CNN.com, right?

So Google News is starting to really get some traction. And speaking of disputes with publishers, that is starting to heat up as well.

The second thing is that their organic search ranking is really developing. What started as just PageRank, plus using the anchor text, is becoming all sorts of other things. By 2007, they were using 200 different pieces of information to determine the ranking on a query. Early on, I know some of these signals included data that they were actually feeding back from observing traffic.

There’s this data network effect that’s starting to happen, where more people use Google and that makes Google better. On the whole, they can understand, “Oh, if someone keeps bouncing off that page every time this query is searched, then clearly that thing doesn’t belong. We’re doing a bad job on this query.”

Yeah, yeah.

Or B, personalization: what can we learn about you from a whole bunch of things that you’ve done in the past? Now, without spoiling too much of the future, there’s not a strong reason to be logged into Google yet. Personalization only works so well, but once Google Accounts become a thing, then that’ll really take off.

So it’s worth knowing—and I really didn’t know this—PageRank really is the thing that got Google going, but that part of the algorithm isn’t really the thing that’s the main differentiable asset today. It was just the start.

And I think this is also a major difference in mindset between Larry and Sergey and Google versus the other search providers. Everybody else just said, “Oh, we’ve got our insight. Okay, great. We’re done. Yes, we solved search.”

Google has never said, “We solved search.” They just keep investing and investing and investing. I think, A, they believe search is going to be bigger than anyone could have realized, but B, they sort of had the insight that the exponential curve of content on the internet is growing faster than Google will ever be able to index it all or come up with clever strategies to sort it all. It’s so dynamically changing that we’ll actually never catch up. And so we need to constantly be investing to approximate good results, because we’ll actually never have optimal results.

Yep, totally. And then the third big thing that kept developing is their infrastructure.

They put so much into building out all the hardware that we were talking about.

A ton of software stuff.

Yeah, and that hardware shifted from, “We’re being really entrepreneurial, shall we say, in how we use commodity hardware cheaply,” to, “Oh, we’re designing our own data centers.”

Yep, absolutely. And we’re designing our own file systems with GFS. I think we’ll talk about a lot of it in the next episode, but it’s very real system-level software that is pioneering.

Yep. Bigtable, MapReduce, et cetera, et cetera.

Yep. That’s on top of the clever search software that they’re writing for things like synonyms. People don’t pay that much attention to this, but synonym matching is actually a crucial part of getting search right. If you’re searching for “cat” and there’s a whole incredibly relevant page about kittens, and you don’t have a good way to understand synonyms, then you’re never going to surface it, even though it might be incredibly relevant.

Google had all these little tricks, like realizing, “Oh, wait. When users are searching for ‘cool cat pictures’ and then they change ‘cat’ to ‘kitten,’ and that happens a lot, we can infer that that is actually a synonym.” Then we can develop our own constantly updating synonym dictionary in real time, which will make our search results better. They have a thousand of these things, and so they’re just pushing and pushing and pushing.

So we're finishing out 2003 here. They're effectively never capital-constrained from this point on. They can always fund every idea that they have. The business has flipped from one where, pre-June of 2002-ish, they had to make trade-offs to one where, after—call it—the end of 2002, there are no more trade-offs ever. They always have the cash for every single thing they want to do.

Yes, we talked about 2002 and the $440 million of revenue and the $185 million of profits. In 2003, that explodes to $1.5 billion in revenue and almost $350 million in operating income. What was operating income the year before?

$185 million. So it went from $185 million to $350 million in 1 year.

Yep. In 1 year. Wow.

Some of you might be listening and be like, “Well, wait a minute. Sounds like their margins got a lot worse.”

And yeah, depending on your accounting, they did. You get to keep all the money from AdWords, and you only get to keep 20% of the money from AdSense.

Yep. AdSense is the answer there. AdSense added like half a billion dollars in 2003 at much lower margin, but was awesome.

Yeah. And as best we can understand it, Google AdWords stayed an 85% gross-margin business.

10. Google Goes Public

Yep. Incredible.

Okay, so that takes us to the last chapter of our story today: 2004, which is the—I think today thought of as—famous Google IPO, and today thought of as a successful IPO, and today thought of as, yes, a successful IPO. At the time, it was the infamous and horribly unsuccessful Google IPO of 2004.

I don't think, other than Microsoft, there had ever been another company like this where there was no good reason for Google to go public. It was wildly profitable, generating plenty of cash, and did not need the investment money. I assume they've never spent their IPO proceeds.

No, of course not. They've never not been wildly, wildly, wildly profitable.

And actually, even more so than Microsoft, Google had a really, really good reason not to go public, which was Microsoft. As we alluded to, there was desperate paranoia in the company: We can't let Microsoft know. They are the actual front door to the internet for all of our users through Internet Explorer. They actually have the capital to fight us, and they have no idea what a good business this is. We can't let them know how good this is.

Fortunately, I guess for the investing public and unfortunately for Google, the JOBS Act had not been passed yet. The 500-shareholder rule was still in effect for companies in the US, which was that if you crossed the threshold of having 500 distinct shareholders for your company, you had to report your financials publicly as if you were a public company.

You know, David, I read all this, too. They had venture capital backers. They had to go public.

Sequoia and Kleiner Perkins are not going to be just sitting there on their hands like, “We love being private shareholders forever. We like dividends.”

Yeah. No. In 2003, these funds had just gone through the dot-com crash, so most of their other companies got wiped out. And it's not like they're the venture capital funds of today that come up with all these clever strategies to look like more permanent vehicles and offer liquidity. These were closed-end freaking funds that needed to get their money out.

Well, I think the question is: Did Larry and Sergey care about that?

Whether they did or didn't, the 500-shareholder rule was a forcing function. But obviously, the VCs—but to your point, I guess the VCs didn't control the board.

Yeah, exactly. Just like Bill and Paul at Microsoft, it was extremely rare that, because Google never needed to raise VC money after the Series A, Larry and Sergey, together with the employees in the option pool, controlled a majority of the votes in the company.

Hmm. Meta, Google, Microsoft. There's something correlated between founder control and incredibly good capital efficiency in a business.

Interesting.

Regardless, this is all academic because truly the 500-shareholder rule would have required them to disclose their financials anyway. So, like, might as well go public and make the VCs happy, I guess. And employees, too. There was clearly pent-up employee demand, and there weren't the same kind of liquidity markets that there are today.

David, put 4 underlines under that. This IPO made half of the 2,000 people who worked at Google millionaires.

Yes. A lot of interest in this thing going public.

So, as we get to the end of 2003 and the beginning of 2004, they know they're going to cross the threshold during 2004. They're going to have to go public. They start interviewing investment banks.

And Larry and Sergey really don't want to do this. They don't like anything about the process. They don't like the IPO pops. They don't like how much money the banks make, et cetera, et cetera, et cetera.

And we should say, this IPO pop thing—it sounds good, right? It's like a phrase that the investment banking community invented to make it sound like a good thing.

A pop is a bad thing for existing shareholders. It means that, in the IPO, you incorrectly priced at too low a price. Then, within 1 day, upside that should have—or really, that is—yours, because it happened and all the intrinsic value was built over these years, goes to the people who had access to buy your IPO shares: the investment bank's clients. Then they get this nice little pop on day 1, and you were mispriced. That is the problem that a lot of people try to solve for in different ways.

So, during the IPO process, they learn from Bill Hambrecht of WR Hambrecht, a boutique investment bank in San Francisco, who really doesn't have the same incentives that the big banks in New York have with their clients, that there is actually an alternative way to price your IPO: something called a Dutch-auction IPO process, which is this arcane thing that had been done before.

How do you do good price discovery? The optimal way to do this is a reverse auction, where you start the bidding high and come down in price incrementally until you reach a clearing price where the entire offering size is spoken for with bids at that price.

And you can imagine just how much this appeals to Larry and Sergey and Google. “Oh my God, it sounds perfect.” They're like, “This is the whole business. This is what we do anyway. This is delightful.”

There's this legend that Eric Schmidt talks about, that they also got a letter from a little old lady who, hearing that Google was about to go public, was really hoping she could get in, and that the small retail investor would have access to it. That pulled at their heartstrings, and I'm sure that's true, too. But you can see why this appeals in a vacuum. This sounds perfect, and like everyone should do it.

Yes. And theoretically, this is also kind of what the investment banker—what algorithm are they actually running? It should be something like this, right? They're meeting with clients. They're picking up the phone: “Are you in at this price? Are you in? How much would you want at that price?” You kind of should be running this algorithm in a loose, human way anyway.

Now, they're worried about pricing and they're worried about the IPO mechanism. They're also really worried about losing control because once they go public, even though they have control of the company now as a private company, the employee shares are sort of captive. People are going to start selling. Now, all of a sudden, the public markets are going to control a lot more of the company. There's a risk that Larry and Sergey might collectively lose control of the company here.

So they do a thing that no one else in the technology industry does. Google has been swearing up and down that they're not a media company, and then they look to the media companies and go, “Wait a minute.” When the media companies need to separate editorial control and have sort of family stewardship of editorial control, but they want to let the shareholders come in and they don't want the business to be able to affect editorial too much, they've got this great dual-class structure for a tech company.

Yes. The families of The New York Times Company or Dow Jones back then, or basically all the major newspaper companies, the original family owners had super-voting shares that ensured that, collectively, the family would retain majority voting control over the company, even if they lost economic control.

So Larry and Sergey decide, “Oh, great. We're going to do a dual-class share structure for Google, too.”

Which today is super common.

Yes, but Google started it. Google was the first tech company to do this. Today, it's everybody: Facebook—Meta—Alibaba, Shopify, Spotify, Coinbase, Airbnb, Zoom, Datadog. Every major IPO since Google, every major tech IPO, has had this.

Famously, Snapchat even pushed the envelope so far. When they IPO'd, the public shares had no votes. So it's not even just super-voting. It's like, “Oh, you public—you get no votes whatsoever.”

That's like being a Green Bay Packers shareholder.

Yeah, exactly. Google pioneered all of this, which I think is why, in retrospect, this IPO is viewed as a famous success.

Okay, so they do the Dutch auction. In practice, it does not go well. But why? The numbers on this are pretty crazy. They are initially floating in the Dutch auction, using the software. And, by the way, Google software engineers wrote the software.

I know. Amazing. Isn't this crazy?

I don't think it's Google-owned. I think they were collaborating with the investment bank. It's this weird joint partnership that they're doing where it's Google engineers, but it's this investment bank running the process.

They're trying to figure out where, in the range between $108 a share and $135 a share, they should price it. Will it be fully subscribed? Well, the actual price where they end up filling the order is $85 a share.

Yep.

Which gives Google a $23 billion market cap at IPO, and they raise $1.7 billion.

You know, this is great, right? It's a $1.7 billion raise and a $23 billion market cap.

That looks like an astronomically high multiple that the company’s been given. So you should walk away and say they really maximized value there. And they probably were just wrong in that initial range that they were looking for: $108 a share to $135, and it only priced at $85.

Well, trust the mechanism. I guess it’s only actually worth $85 a share.

Nope. It pops to $100 on closing of trading the very same day: an 18% pop day one. Then, by the end of the very next year, 16 months later, it is almost a 5x.

Yeah, this thing was not at all priced correctly. So there’s a reason why, when you look at the legacy of the Google IPO, the dual-class share structure was a great idea. Everybody does it. The Dutch auction IPO was not a great idea. Nobody has done it since. It doubled within the first few months.

Yeah, I mean, it’s great PR, right? The stock’s doing well. People think highly of your company. That’s good for all sorts of reasons. But this did absolutely zero for making sure that the company doesn’t leave money on the table.

Yep. So funny.

It’s all kind of a footnote of history anyway, because what’s a few percentage points between friends when the company would go to over $2 trillion today, as we are recording this? That’s roughly 100x the market cap when it IPOed. And, David, you’re not counting dividends.

Not counting dividends, right?

Of course. If you reinvested dividends, you’d make significantly more than 100x since the IPO. Well, after our Steve Ballmer interview, we’re never going to not count dividends again. But to preview a little bit of the rest of the series, Google’s a $2.1 trillion market cap company today—roughly 100x since the IPO. Amazingly, I’m going to ask: Do you know—I know you know—what Google, or Alphabet’s, price-to-earnings ratio is right now?

Ooh, baby, I do know, because I was just looking this up. It is kind of an all-time low: 20x earnings, 6x revenue. So compare that to its peer companies. Amazon’s P/E is 35. Microsoft is 37. Nvidia is 46. Apple is 30. Meta is 27. And Google, or Alphabet, is down at 20.

It’s not like Alphabet’s not growing revenue. They’re growing revenue just as fast, if not faster, than all of those companies except Nvidia. Something is going on here. This price seems to reflect that, even though revenue is growing nicely and margins are quite high. Somebody—and that somebody is Mr. Market—thinks the future is a lot bleaker than they do for those other companies.

Never mind that Google invented AI and published the Transformer paper. We’ll get to all that. No spoilers.

Okay, no spoilers. That’s where we’re going to leave Google for part one. But one more little story to tease you with for part two next time.

In the same month, April 2004, when Google files its S-1 for its IPO, Google does an unexpected product launch on April Fool’s Day, which really wasn’t a good idea because Google had a history of fake April Fool’s joke announcements. If you have one that sounds ridiculous, don’t launch it on April Fool’s Day, because people will think it’s a joke.

The product they launch actually sounds way too good to be true: web-based email from Google with 1 gigabyte of free storage for every single user. To put that in context, Yahoo Mail and Hotmail at the time had around 2 megabytes of free storage per user. I think it was 20x the next best—that’s the statistic I read on Gmail.

Yeah. And it comes with Google Search baked in across all of your emails, and it’s entirely web-based. It runs in your browser anytime, anywhere. It’s like the greatest April Fool’s gift to Internet users everywhere that Google could provide.

So the question, though, is why did they do this, knowing what we now know about Google? Ben, wouldn’t it be great if there was a reason—a really compelling reason—for someone to be logged into Google? And wouldn’t it be great if we could just attach more things to a user’s life that could be entry points to Google Search and the greatest business of all time, search ads?

What if, then? What if?

Okay, David, we’re going to tell the whole Gmail story as part of chapter two, but I do have to give you one thing that’s specific to this episode.

Go for it.

The engineer who started Gmail, Paul Buchheit, is now, of course, a partner at Y Combinator and actually, with Bret Taylor, started FriendFeed.

That’s right. Paul Buchheit is awesome. He recently launched a new venture fund and was actually the original coiner of the term “Don’t Be Evil” at Google.

So he’s working on Gmail. It’s very early—it’s like 2001. He’s been working on this thing for two and a half to three years before it launches. So we’re at the very beginning of it, in his 20% time, right? It’s a 20% project.

It starts as, “I’m going to look in your Unix directory at your mail, and I’m just going to treat that like the web, just the same way that we treat web pages. I’m going to take a search box, point it at your mail folder, and let you search.” That’s it. That’s the only functionality of what would become Gmail. The search bar is actually the first feature of Gmail, and everything else came later.

As he’s playing around with this, he has an idea: “Well, if our core business is indexing organic results and showing some ads, maybe in addition to indexing and searching these organic results out of your mail folder, I should just go grab ads from our ad database, display them around, and see how well the content matches.”

He’s showing this off internally. Larry and Sergey see it, and they go, “Wait, does this work on websites, too?”

And so the thing that led to AdSense—this was the beginning of the idea for AdSense—was actually part of the prototyping process of Gmail.

Ah, amazing. I love it. And I love how you saved this to the end because you knew we were going to do a little Gmail foreshadowing.

Well, you texted me. You said, “I think we should do a little Gmail foreshadowing.”

Great. Yes. Thank you to Paul Buchheit for sharing the story with us. Amazing.

All right, bringing it home. That is the building of Google’s search business. Let’s bring this one home.

Yeah. So, David, this chapter, this episode, definitely feels like the building of the castle.

Yeah, and maybe next episode is going to be the building of the city around it, the state around it, the nation-state around it.

And depending on your metaphor, is it an entire property, a platform that they’re building around it? A city? Is it a moat?

Yeah, but it’s definitely—this one is building the castle. We’ll have to see.

All right, let’s go into playbook for part one. Ben, what do you got?

The way that I framed playbook for this one is I tried to itemize the bullet points of why Google worked. As I think through them, if I had to lay them out to someone, it starts with the best original algorithm insight. They had the best organic relevance out there, which created the best results in order: fast, delightful, clean, simple UX. And they were truly dedicated to organic search. The aversion to paid inclusion for as long as they had served them very well.

So this amazing original algorithm for organic search is one. Two: best execution of the search advertising model. Once you get all those puzzle pieces in place—the auction, the switch to cost per click, factoring in relevance with click-through rate—it really is this truly beautiful system.

Advertisers are incentivized to make their ads more relevant and only bid on the most relevant keywords, because it means they don’t have to pay as much. It’s the best ads to the right users at the right time. And to your point, David, it maximizes—it is literally the algorithm to maximize Google’s expected value. It is a harmonious system that they developed.

Three

clever infrastructure advantages. They just invented stuff, and they thought about problems differently. They reasoned from first principles.

Four

they hired the best people, truly only world-class people, for a very long time. And because when the dot-com crash happened, they could basically get anybody that they wanted there in the second half of this episode, Paul Buchheit had a great quote when I was talking with him. I don’t even think I realized it at the time that it was truly just the best people in the industry working around him.

So that’s four. Five: culture. A culture of thinking insanely big, mostly by inexperienced, untainted people, that helps you with creativity and helps you come up with new ideas. It was like the naivety of kids on a college campus who are dreaming, matched with the brain power of the very best PhDs.

And this hardcore belief that whatever our big ideas are, we always have to think with scale. Every little implementation detail has to be, “As this scales, will this work, or do we need to rearchitect the system?” That is very impressive.

So culture, which includes power-law dynamics, by the way—being willing to make big, bold bets because they could be these multibillion-dollar payoffs. That’s five.

Six

the self-reinforcing data network effects once it takes off. I think that is underappreciated about Google. A lot of people say, “Oh, the algorithm,” but the algorithm is so dependent on all the data that is generated.

And lastly, a mission that has stood the test of time: organize the world’s information. It’s not too broad. It’s not too narrow. It feels altruistic, but of course, the business behind it is actually the best business of all time.

Yep. I would add on to the second-to-last one you had there, the data network effects. It also has the flywheel effect of liquidity in the marketplace of users, queries, and advertisers. Everything that we just talked about—once Google had that realization of, “Our business gets better the more users and advertisers we have,” and thus we should be willing to spend basically anything to increase those two pools.

This is my quintessence.

Oh, okay. I'm stealing your quintessence. I love it. I feel like this is the most unique insight of this episode: whoa, these are economies of scale that don't just reduce your cost as you get bigger, but increase your revenue as you get bigger.

Yeah. Okay, great. Well, I didn't mean to steal your thunder with quintessence. Sorry about that. We did our quintessence earlier this episode.

Okay, great. Great. Great. All right, give me your playbook.

Great. I've got 2 other meta points that jumped out to me from this episode, in addition to what you just said about the incredible encapsulation of why Google worked. When you and I were talking about doing this and starting the Google series, the reason we decided now was the right time was because of everything going on in AI, and it feels like understanding Google has never been more relevant. If we're going to do Google on Acquired, we've got to start at the beginning and understand how Google was built, because that's what we do.

I thought this episode would set the stage to then get to today. Telling the story, though, and doing the research, I was like, today is exactly the same. The parallels—I had the exact same thought—between what happened between 1996 and 2002 feel like everything that we are living through right now, 2021 to today. Or even, let's start with the ChatGPT moment. Put more sharply, I thought this was going to be, well, we're going to have to eat a lot of vegetables to understand Google so that we can understand where the transformer came from, to get to the real great meat and what we can learn about AI by studying the present.

But I think by studying the way that search played out—how did monetization work? How did the value chains work? How did distribution work? How did monetization work that uniquely enabled distribution? Where did all the competitive dynamics come from?—this is a history that doesn't repeat, but it rhymes. And God, does this rhyme? Totally transferable lessons and dynamics.

Yep.

When you were telling the story of GoTo and Overture and the launch at TED, and how upset people were but how brilliant it was, I was thinking, well, what would the analogy be today? What if somebody made a chatbot—an LLM, a model—and what it told you was just what people paid it to tell you? People would go crazy if that happened. But, yes, is that worth trying? Should somebody try that? Well, let's see: product design here on Acquired by David Rosenthal.

Yeah. Right. Right. Right. Probably a bad idea, but it feels like such a similar moment that we're in. That's just what struck me over the head doing all of this. So, wow, history doesn't repeat itself, but it does rhyme.

Yep. And then the other big playbook theme I had was, God, did Google really come of age at exactly the right time? We talked about this earlier, but if Larry and Sergey had met and started working on this a few years earlier, it would have been Yahoo, because the web was just so much smaller. You didn't need a technology-based search engine to understand it, right? And then if they had started a few years later, it would have been too late. It would already have been too big. You would have needed too much technology and power to make it work. It was the perfect window. There was a very narrow window to start the Google of that era. And again, maybe this is a subpoint of my first playbook theme of just the parallels to today.

Yep. All right, powers. What of the 7 powers does Google have? And for new listeners to the show, this is based on a book called Seven Powers by Hamilton Helmer. It is the 7 factors that enable a business to achieve persistent differential returns, or basically how to be way more profitable than your closest competitor sustainably. The 7 are counter-positioning, scale economies, switching costs, network economies, process power, branding, and cornered resource.

Let's see. To start, I guess let's just go down the list. I actually don't think there was tremendous counter-positioning here. You could argue versus Yahoo. Well, it's interesting because it was a new industry. They were counter-positioned against the other search engines that existed at the time in that, as we talked about with Excite when they were trying to sell BackRub to Excite, the other search engines wanted you to stay on the page and Google didn't.

But I don't think that really counts, because it was a new industry and those were bit players. There was no incumbent already there, right? And Google was just better. Being better is not counter-positioning. The best argument for counter-positioning is versus the portals.

Yeah, versus Yahoo. It may have become clear at some point that search actually is important, but the portals couldn't really pivot to it because they couldn't give up all of their portal ad revenues.

Yeah, I think that's also right. And also, take Yahoo. Yahoo had constructed itself as a media company, even though it was started by 2 electrical engineering PhDs from Stanford, right, and just couldn't pivot. Famously, it tried to right the ship: it bought Overture, then it bought Inktomi, and then they spent years trying to put Overture and Inktomi together to create a packaged competitor to Google.

This was the ill-fated Project Panama at Yahoo, which, by the way, fun Easter egg: that is where Jan Koum and Brian Acton met at Yahoo, and then they would get so frustrated and leave and start WhatsApp. But yeah, I think there's counter-positioning against Yahoo there.

Okay, scale economies for sure. There's more powers here too, but the whole thing is scale economies. And it's more than just the traditional one. Think about Hamilton's traditional definition here: Netflix has scale economies because it can amortize the cost of buying a given piece of content across more users. This is more than that.

For a given piece of infrastructure or software or hardware investment that Google wants to make, or a user acquisition cost, they can amortize that across more users. But also, as they scale, they make more revenue per user. I don't know what that is. Is that a new power? Super-scale economies? Where does this come from? Auction-based businesses: whenever you have auctions to determine pricing, the more liquidity you have, the higher prices are. Are there other businesses that we can look at that are similar? Do scale economies ever explain why scale gets you more revenue? What is the thing where, with an increase in scale, their prices go up? They maximize their available take on any given micro-auction.

This is weird. I'm trying to think of another auction-based world. Christie's would have this, or Sotheby's: as you get more and more people into the auction house audience, any given sale is likely to go at a higher value. A real estate brokerage, if people were actually loyal clients of a real estate brokerage.

Oh, is this just network economies? The more queries you have, the more advertisers you'll have; the more advertisers you'll have, the more queries you'll have.

That is true from an advertiser to a searcher and a searcher to an advertiser. So, we should say this definitely has network economies. It's almost like there are negative network economies from advertiser to advertiser. You don't want your competitors to be on the platform, but Google does.

Yeah, maybe you're right. Maybe there's something unique to auction models here because the price is dynamic. Hamilton, if you're listening, we need to talk. Yeah, great. Okay, let's keep going.

Switching costs? Not yet. Talk about that in the next episode.

When you're not logged in and there's no personalization, no real switching costs yet.

Yep. And everything else you'd switch to is worse, honestly.

Branding. Yeah, to a certain extent. Absolutely. They built a brand of trust, speed, and fun. I had a Google shirt. I used to read Google blogs. From 2002 to 2006, I was as big a Google fan of things that were Googly as I was an Apple fan for that period of my time. And I think a lot of people were.

I think for those of us who weren't in Silicon Valley, that's what it meant to be successful in Silicon Valley: to become Google.

Yeah, I would agree with that too. I think it's weak branding power, though, because that's not a branding power like Hermès has branding power. It's gone away over time. Now it's just a brand. And, for the literal definition, are people willing to pay more for the brand? Are advertisers willing to spend more on Google than elsewhere? No, they're rational actors.

They had an employment brand, though. To your point, they absolutely had an employment brand. Smart people would be willing to do anything to work at Google.

Yep. And lastly, cornered resource.

Not really at this point in time. Not really. And process power—I also don't think there's much there.

Yeah, I don't think so. Okay.

Quintessence, we already talked about yours. Do you want to say another word on it?

No. The increasing returns to scale on the revenue side are unbelievable. And the fact that they can have that insight and then realize they need to go be super aggressive on spending—it makes total sense if you have a long view and think, our RPS are only going to go up. People are going to be sticky forever. It is worth investing heavily to win this race. It's amazing.

Yes, I love your quintessence. It's totally right. I will second and underline it. One not quite as good but alternative quintessence I want to put out there about Google is a quote from a page in Steven Levy's book In the Plex.

On June 8, 2007, Justin Rosenstein, who until recently had been a Google product manager, sent an email to his colleagues. “I am writing to spread good news,” the missive said. “Facebook really is that company.”

“Which company?”

“That one. The company that shows up once in a very long while.”

The Google of yesterday. The Microsoft of long ago. That company that's on the cusp of changing the world, that's still small enough where each employee has a huge impact on the organization, where you know you'll kick yourself in 3 years if you don't jump on the bandwagon now, even after someone had told you it was rolling toward the promised land. That was Google. Google was that company. Microsoft was the first—that company. Google was that company, and then Facebook was that company.

That's exactly right. And they are exceedingly, exceedingly rare.

Yes, they are. So, that about captures it. That's my quintessence: Google was that company.

All right, carve-outs. I have a 3-way tie of 3 excellent TV shows that I watched in the last 2 months because we didn't do carve-outs with Steve Ballmer. The first, and I think the most landmark of mine, is The Rehearsal with Nathan Fielder, season 2. Oh my God. I don't want to spoil anything for anyone, so if you're a person who believes that anything is a spoiler, stop. I'll tell you things that you'll learn in the first 10 minutes of the first episode.

Nathan, just to give you some quick background, did a show a decade ago called Nathan for You, where he went and helped small-business owners figure out how to make their businesses better and improve on a key area. But it's all kind of satirical. The way that he helps them accomplish their goals is very bad for their business in most other respects. He's an incredible comedian with a really dry sense of humor.

The thing that he did in Nathan for You was a pretty good commitment to the bit. The lengths that he would go to—for example, to get a coffee-shop owner to get more traffic in their store, they rebranded the store to Dumb Starbucks. He spent hundreds of thousands of dollars of the studio's money, or maybe millions of dollars, to commit to this rebrand. Obviously a bad idea, but great TV.

In The Rehearsal, he commits to the bit so unbelievably hard that it takes years of his life. He has a goal to reduce the number of plane crashes by doing a deep study of the thing that causes plane crashes, which he believes to be pilot communication. In this season of The Rehearsal, he builds elaborate sets and hires a bunch of actors to simulate different experiences to help these pilots feel more comfortable communicating with each other, so that fewer planes will crash. And I am telling you, David, this is the tip of the iceberg. It gets crazy.

Sounds amazing. Commitment to the bit at an all-time-great level. Are we committed to the bit?

We are. We're nowhere near as committed to the bit as Nathan is.

It sounds like we're not.

Yeah, it's inspiring. Okay, so that's one. Two is a much more casual, very enjoyable show called Your Friends & Neighbors on Apple TV+. It's Jon Hamm. It's beautifully shot. It's a little bit like following rich people around, like Succession. These are sort of fictional characters, but with an unexpected twist. Love Your Friends & Neighbors on Apple TV+.

Andor season 2 on Disney+ was excellent. It starts a little slow. The first 3 or 4 episodes are not as good as season 1, in my opinion, but the last 8 episodes are some of the best Star Wars canon that exists. I love it.

Ben, you are, in addition to being my best friend, also my smart friend who has great TV recommendations. So, I love that you get to be that smart TV-recommender friend for the internet, too.

I'm here for you, and I will continue dumping this on you, even though I know you don't watch TV. You'll never get to this. Two little kids. Tough. Tough.

This is for listeners.

My carve-outs. I've got 2. Well, I've got one standard carve-out: Gamecraft, season 3—the Gamecraft podcast. We did a crossover with Mitch and Blake years ago. They have really committed to the bit. Season 3 is excellent. I'm so glad they've kept up the podcast. It's really great.

If you like gaming and the gaming industry, Mitch and Blake are 2 of the best in the business. My next carve-out related to that is actually sort of a real-time dilemma carve-out. This is good because it could be a multipart series here on Google. I'll report back in the next episode which direction I went with this carve-out.

As I said before, I think I did a preemptive carve-out. I was so excited for Switch 2. Switch 2 finally launched. I haven't gotten one yet. I have a reservation for a shopping appointment at the new Nintendo Store here in San Francisco. I'm super excited. I can't wait for it. I can't wait to play it with my daughters someday as they're approaching that age.

All right. What's the decision?

As I was watching reviews on YouTube, I started getting into Steam Deck content, and now I'm desperately conflicted. Do I want to get the Switch 2, like I'd been planning, or do I want to go in a totally different direction and get a Steam Deck?

Well, Acquired listeners, tune in. If you weren't interested in Google Part II on its own merits, now you're going to be on pins and needles.

I would ask our listener base for help with this decision, and I would love all of your thoughts, but because of our editing process, the reality is I will have made my decision already by the time this episode goes live. So, tweet a picture, listeners.

I'll tweet a picture.

Yes. Yes. Yes.

All right, that's what I got.

We've got some thank-yous. We talked to a zillion people, as seems to be the new precedent when we do these large tech companies. First off, Arvin Navaratnam from Worldly Partners did an awesome write-up on the company, as usual.

And then, David, you've been maintaining the list.

Yes, I've been maintaining the list of some of the folks we should thank for helping us with this episode. In addition to the many other folks we talked to who we can't mention, thank you. You know who you are.

But specifically, thank you to Craig Silverstein, Google's first employee; Anna Patterson; Omid Kordestani; Alan Eustace; Clay Bavor; Bret Taylor; Jeff Dean; Jen Fitzpatrick; Danny Sullivan; Nick Fox; Paul Buchheit; Bill Gross; Wesley Chan; and Eisar Lipkovitz. Thank you so much for the conversations.

David, I feel like we got a dream team of people if we want to go start a tech company.

Yeah, all-star lineup.

I started joking by the end of the research process. I was like, I think we are sapping billions of dollars of market cap out of the economy by taking people's time to have these conversations, so we greatly appreciate it.

Yep.