a16z 的 David George 谈 a16z 最具争议的押注：AI 时代，利润率和收入还重要吗？

Our best-performing fund in the history of the firm is actually a $1 billion fund.

David George is a general partner at Andreessen Horowitz, where he leads the firm’s growth investing. His team has backed some incredible, defining companies of this era. He’s now investing behind a new generation of AI startups.

If you overweight the fear of future theoretical competition, you can always talk yourself out of making an investment. The number one way to measure a company is ultimately return on invested capital. On the gross margin point, today I’ll say this: We give a little bit more of a pass than we used to.

At what point does the entry price, even for OpenAI, become not a good use of dollars? We have to constantly reassess this. What I just don’t understand, and I would love to understand, is Flow. Can you help me understand Flow? I think the world kind of scratched its head: Why did it make sense to you when it didn’t make sense to anyone else?

You remember what I said earlier about investing behind strength of strengths?

Yeah. Ready to go? David, dude, I am so excited for this. I’ve been looking forward to this one for a while, and I feel like I’m extra prepped now. I’ve just listened to you on Invest Like the Best, so I’m ready to go, dude.

Let’s do it.

1. Why Everyone is Wrong: Mega Funds Does Not Reduce Returns

Okay. I actually spoke to most of your partners beforehand, and they said to me that I had to start with a show that we did with Everett Randle. Everett Randle said on the show that you cannot look LPs in the face and tell them you’ll do a 5x with the fund sizes you have. How do you think about responding to the notion that one can’t say to their LPs, “You’ll do a 5x with large funds”?

Well, Harry, it is great to be back with you. I love hanging out with you, so I’m glad we’re diving right in. As it relates to fund sizes, our funds consistently beat small, large, diversified, and concentrated venture funds. Our larger funds have outperformed our smaller ones, and our larger ones actually have similar multiples of money to our smaller ones across strategies.

I would start by just saying this: In venture, we have 2 customers. We’ve got the LPs, and we have founders. On the LP side, money is going to flow to where the highest returns and best risk-reward are, and so I think our fund sizes are a reflection of that. Our best-performing fund in the history of the firm is actually a $1 billion fund, so it’s a large fund.

In that fund, Databricks has returned 7x the fund so far. Coinbase has already returned 5x the fund on a DPI basis. In that fund, we also had GitHub, DigitalOcean, Lyft, and many other things. To me, you can see it in the data and in our returns already. It’s about the number of winners you capture, and if the big ones are great, that can really work out.

I think the idea that large funds can’t have great returns is just not true in our experience. Private markets have changed, and tech waves create bigger opportunities. Let me just talk about each.

The private markets have grown 10x over 10 years, so it’s over $5 trillion in market cap now in our market. We actually just looked at the 50 top IPOs from 2017 to 2025. If you disaggregate where the dollars of return come from, 47% of the dollars of gain happens between the seed and the Series B, and 53% of the dollars of gain happens from Series C onward.

There are actually a lot of dollars. I was surprised when we looked at this, but there’s a tremendous amount of dollars of gain that happens at the later stage. That’s 2017 to 2025 IPOs, and it actually skews a little bit heavily toward when companies were still going public, when they were smaller.

The size of outcomes is huge. Again, we’ve got $5 trillion of private market cap. If you look at our LSV funds, the aggregate market cap in those funds has ranged between $700 billion and $1.5 trillion. It’s just large companies, and if you apply ownership assumptions to that relative to generating 3x or 5x returns, it’s pretty manageable.

That’s the private market and the conditions that have changed, and we can talk a bunch about that. Tech waves tend to create massively different value. This is very well covered, but the big story of mobile, social, SaaS, cloud, and e-commerce all at once was $20–25 trillion of market cap creation.

If that started from scratch today, given the public-private market dynamic that I just described, so much of that value creation would take place in the private markets. We’re in inning 1 of this new big tech wave. I never would have expected in the last wave that companies like Salesforce would be worth $230 billion, ServiceNow would be worth $175 billion, CrowdStrike $130 billion, or DoorDash $100 billion. But here we are.

If you look at what’s happening in the private versus public markets now, the size of the winners from a new tech wave is going to happen in the private markets.

With the extension of private markets, are you worried that companies are not going out for so long that they’re getting competed by new private companies before they get a chance to get out? You can look at the dynamic between Axon and Flock Safety as a good example of that. Axon is eating away at part of Flock Safety’s business, where they replaced them in Atlanta, a core part of Flock’s business. Both are private, and they’re eating away at each other in a world where one of them would have gone public in that time, in a traditional world.

I don’t think that whether it’s public or private has much to do with the competitive dynamics, to be honest.

But it does in terms of liquidity for venture investors.

We’ve led 3 rounds in Flock Safety. We led their last round, too, and so we’re still quite bullish about Flock Safety. You could talk about the increasing competition with Axon. The real story of that one is that the market is actually embracing technology now, finally.

Historically, selling into law enforcement was a terrible category. Now it turns out that it’s a wonderful category. If you actually have the most compelling products, you can get tremendous amounts of market share. I don’t worry about that dynamic at all.

Frankly, I think the more some of those companies have stayed private, the more it’s been to our benefit because we’ve been able to increase our ownership over time.

Are you able to take money off the table with the extension of private markets, given how big a name you are and how big a position you often have? You’re just a big piece of a cap table. For someone like me, it’s much easier to sell out in a later round. Are you able to? Do you have that discussion internally of, “Hey, we should take chips off the table now”?

We could, but historically we have not. For the most part, for the companies that have decided to stay private, we’ve been really excited to stay in them, keep backing them, and that’s probably the strategy that we’ll continue to have.

2. The Biggest Advantage of Staying Private for Longer

I think this staying-private dynamic is a little bit overblown because there are some idiosyncratic reasons why certain companies have stayed private. Many companies and many CEOs that I talk to are very happy to be public, or they’re excited to go public.

I tell our CEOs all the time—I’ve been fortunate to work with a bunch of public companies—not one of them has said, “I regret going public.” I think for most of the companies that we’re talking about, they’ll wait longer than they had historically, but they will still end up going public.

Seriously?

Yeah. I don’t mean that horribly. I don’t meet many public CEOs who don’t tell me they wish they were private.

No, I mean, look, I think there are tremendous benefits to being public. There are huge benefits to being private as well, which we can talk about. But you could look at many of the public companies that are out there that had difficult paths, and they would say they wouldn’t trade it.

3. The Most Controversial Decision in Andreessen Horowitz History

Can you genuinely tell me what those benefits would be? It’s easier access to capital in some cases. There are select few private companies that have very easy access to capital in the private markets. I think there’s a trade-off in the private markets where you actually have a more expensive cost of capital, even if you have access to a lot of it. So I think you can get a cheaper cost of capital in the public markets.

Do you think you can still get a cheaper cost of capital in public markets? Public markets seem more expensive to me today. In private markets, we’ve given more elasticity on price today.

No, I don’t think so. The companies that we’ve invested in, I’m very excited about them in the private markets, and I think if they were—

When you look at likely Ramp or Lovable at the price where they are, they’re priced at the same price as Wix, and Wix is doing $2 billion in revenue.

I’m not close enough to those to know. I don’t follow those companies. We’re not close enough, either.

The comps are very sharply contrasting what we’re saying: that, actually, public is harder and private has a cheaper cost of capital.

We’re not close to those companies. I’m not close enough to know how they’re valued relative to their performance. I can say that, in our portfolio, the companies that we have invested in over the last year or so, I’m pretty confident that if they were in the public markets, they’d probably have access to capital at a cheaper cost. I always remember watching John Collison say, “Oh, why I’m not going to do the accent”—I’m not going to do it because I’m terrible at accents, which is why I’m not an actor.

You don’t need to mess with it.

Stop it. Sorry. You’re too kind. I don’t want John to unfriend me because I’ll sound like a Russian. But he was like, “I don’t understand why I would go public. I don’t need some 25-year-old associate to tell me that I need to plan more efficiently.”

Yeah. For certain companies, it’s a huge benefit. For somebody like Stripe that can get a pretty liquid market in the private markets, I get it. For them, I think the biggest benefit is not so much that, because I think in the fullness of time, if you’re transparent, tell a good story, and share with the public markets, they’ll understand your business.

I think the biggest advantage is the avoidance of volatility in your stock price and employee management. If you can steadily grow or control your stock price in the private markets, even if it’s a slight discount to where you would be in the public markets, I get the benefit of that for sure. We’ve seen some of our companies that have been able to do that, right? Stripe, SpaceX, Databricks—it’s worked to their advantage for sure.

4. Is Public Market Capital Actually Cheaper Than Private Capital?

Is there anything else that you think is completely misunderstood, or that people don’t see, about the extension of private markets and the opportunity that’s opened up for fund sizes like yours with this extension?

I think the biggest thing that’s missing is just the change in what that means for asset classes. It used to be that you could get access to great companies in the public markets that were small-cap. It turns out that’s fewer and farther between now.

We just did an analysis on this, and it turns out that the number of public companies has been cut in half over the last 20 years. Many of the companies that we’re talking about would already be in the public markets, and they’re not. If you look at where the returns are getting generated, the returns are actually getting generated in the private markets before they go to the public markets.

Now, if you look at what remains in small-cap land in the public markets, there are definitely some high-quality companies, but the quality has deteriorated. A friend of mine just shared this analysis with me that showed the return on invested capital of the Russell 2500 over the last 30 years. If you look at ROIC, which to me is the easiest measure of the quality of a company, the ROIC of the Russell 2500 over the last 30 years has gone steadily from 7.5% down to 3%—more than cut in half.

That’s a pretty steady decline. It ebbs and flows with economic cycles. I think the biggest thing that’s missing—and it’s probably a reality that we have to adapt to in how we run our business, but it’s also a reality for institutional investors and the LP community—is that the asset class is no longer a bespoke, small thing. It’s the grown-up leagues; it’s the big leagues.

If you just look at the size of private technology and high-quality companies, it dwarfs the size of private-equity technology in the US. That’s a major shift. We’ve had to adapt our business to it in a big way. If the companies stay private longer, we have to give them new things. They have to be multiproduct, multichannel, and international. With AI, it’s happening much, much faster, so we’ve changed our business as well.

I think the market reality is that historical views of what the asset classes are do not reflect what they actually are today.

Completely agree. I’m an institutional investor with a $10 billion endowment fund. How should I change my asset allocation between private, venture, and public, given that blurriness, merging, and lack of clarity that you just mentioned? What would you genuinely advise me?

I’m heavily biased. I recognize that many of the endowments have a starting position, which is that many have probably found themselves a little bit overallocated to privates. I don’t know how to assess that relative to the future outlook. But if I take the future outlook only, where I think the most attractive opportunities are is this: If you just start with where the 10 most valuable companies in the world are today versus 25 years ago, 8 of the top 10 are US West Coast-based technology companies, and they were venture-backed.

If you assume that the future is likely to be something similar to what’s happened in the last 20 years, I think the most interesting place to be is this asset class, which has exposure to what those next-generation, dominant companies can be. I think the allocation should reflect this sort of melding of what used to be part of the public markets that no longer is. That’s a newer asset class.

That’s one piece of it. My friends in private equity do an amazing job. They have incredible returns.

Do they have better returns than you?

My compliance guy doesn’t let us talk about returns, but if you were to look at our returns—or the top-performing venture funds, let’s just call it that—relative to top-performing private-equity funds, the top-performing venture funds outperform. That’s historical, but I think it’s going to be more extreme in the future because AI and the effective implementation of AI are going to be the most important things for companies over the next 10 years.

There are so many things that I want to talk about. You said there that 8 out of the 10 are US-based. Candidly, would you say, “Don’t worry about Europe. If you have the US covered, you’ve got 8 out of the 10 and dominant market share”? Silicon Valley’s retained the title as the AI center. I’m obviously in London, so I’m not going to be offended, but is that what you would say?

No, not at all. There are great entrepreneurs in Europe, and we’ve backed a bunch. We backed Mati from ElevenLabs, and he’s doing an extraordinary job building what we think is a generational, market-leading company. You’re shaking your head.

Yeah, I turned it down at seed.

You can’t bat a thousand.

Dude, another one of yours I turned down at seed that keeps me up every day. Every day. Alex—

There are amazing entrepreneurs.

Deel. Deel.

Oh, Deel.

Oh, two on 12.

Deel. My favorite thing about Deel—I mean, Alex is just absolutely relentless. I recently had a post, I think it was an announcement of something, that I posted on LinkedIn. Somebody had commented on it, a CFO of a growth-stage company, and I immediately got a screenshot from Alex circling the comment. He said, “Can you introduce me to this guy? He looks like a great Deel customer.”

I’m like, “Man, this guy is always selling.” In a market like that, that is exactly what you need. I love it.

Did I ping him on a Sunday morning and say, “Hey, a Project Europe company—very young founders under the age of 25 with no employees—wants to be a Deel customer. Who’s the lowest person on your team I should introduce them to?”

5. Quick-Fire Round

He said, “You can do it now, please. I’ll take the call today.” I was like, “Dude, it’s like this one person.” He said, “I’ll do it. It’s cool to meet him.”

It’s actually amazing. He’s relentless. This is very much the kind of founder that I love.

One thing that I do worry about when we look at this stage of the market, especially when it comes to this price, is that we’re taking venture risk in terms of the probability and stage of the company, but at prices that were previously for very mature companies. How do you respond to and think about taking venture risk at super-high, mature-company prices?

I think there are certain instances where it makes sense. I would agree with you that there are many instances in the market where it doesn’t make sense. I think there are certain instances where some degree of likelihood of success is very, very high despite a very early stage.

As an example, my partner Sarah led a round in Character.AI. It was extremely early-stage, and we invested at what you would call a growth-stage price. But we knew that the likelihood of some degree of success in backing Noam was extremely high. It worked out that way.

For extremely special people like that, we’re comfortable stepping into those situations.

So would you argue that, for deals like that, the risk is not actually the entry price because you’ve got the liquidation preference, which means someone like Noam is always going to get bought for whatever the liquidation preference is—$100 million or $200 million, obviously?

Yeah. We almost never make an investment saying, “We’ve got the liquidation preference.” But there are certain situations like that where we feel like it’s pretty asymmetric. Backing Noam, you feel like there’s a pretty safe downside and an extremely high upside.

I think the kinds of people—I say people because some of these are earlier-stage people—that warrant an investment decision and a thought process like that are extremely small.

I mean, the list is 5 people, I think. Love that. I spoke to Brian Kim on your team, and he asked me, “Do you see it as part of the growth fund’s charter to fix the errors of omission from your venture team?”

Very much so, but we do it in partnership with the early-stage team. This is our whole model, right? We talk about mistakes we make all the time, and I have very painful errors of omission at the growth stage, too.

If you think about what our business is, we’re never going to have 100% market share of all the best deals at the early stage. By having a growth fund, we can come later and—we call it the “fix-the-mistake fund” internally when we’re joking around—but we do that in close partnership with our early-stage team.

6. The #1 Investing Rule for a16z: Always Invest in the Founder's Strength of Strengths

We always join team meetings. We’re always talking to each other, asking the early-stage team, “What Series A’s do you wish you had done that you passed on? Which seeds do you feel like you passed on?” When you have a situation like what you described with Mati, and you’re pulling your hair out that you didn’t do the seed, that’s okay. Come back and fix the mistake at the B or the C.

It’s a huge part of our charter. By the numbers, about half of what we do is follow-ons from existing venture companies. From a dollar standpoint, another 15% is follow-ons from existing growth-stage companies, and about a third or so is fully net-new companies. When we’re doing the fully net-new companies, we have a pre-existing relationship with those founders from the early stage every time.

Can you just tell me, on the 50% and 15%: 50% is follow-on, but 15% is what? Follow-on of a different kind?

Of an originated growth-fund investment. The thing that’s important about that is, when we invest, I don’t know, 2/3 of the time or so, it’s into a company that we have a pre-existing relationship with, either at the early stage or in the growth fund.

The 50% is, we did the ElevenLabs growth round, and thankfully Jennifer and Bryan did the early-stage round. The 15% would be that we led 2 more rounds in Flock Safety, or we led another round into Figma, or we put more money into SpaceX or Waymo—something that was originated out of the growth fund.

What did the venture fund do that you didn’t double down into, where, with the benefit of hindsight, you’re like, “We should have done that”?

Oh man, there are many of these. We don’t get it right all the time. I think the most relevant are when we passed and then ended up fixing our own mistake. For example, with Deel, there was a round in between when Anish led the Series A and then we co-led the Series C, and we obviously wish that we had done that.

What did you learn from that? I have this, too. I actively ask myself, “What do I learn from missing ElevenLabs, from missing Deel?” My takeaway is very simple: I thought I was smarter than markets. I thought I could forecast what OpenAI’s product roadmap would be in the case of ElevenLabs, and actually, I should have 100% backed up the truck on an amazing founder.

Same with Alex at Deel: payroll, ADP, Paychex. Alex is amazing. Just back. What was your takeaway from missing that B, which is a mistake?

I think often the takeaway is that when we make an investment, we should always be investing in strengths as opposed to a lack of weaknesses. This is a philosophy that comes from Ben: if you have spiky strengths in a founder and a company, it’s okay if there are weaknesses or concerns.

Often, the mistake will manifest itself as the fear of future competition—the fear of theoretical competition, right? That’s the perfect articulation of what you just had for ElevenLabs: “Oh my gosh, aren’t the labs going to do it?” It’s the old VC trope of, “Well, isn’t Google going to do it?” or, “What happens if Facebook does this?”

If you overweight the fear of future theoretical competition, you can always talk yourself out of making an investment. We try really, really hard not to do that.

Other mistakes—if we pass on great companies, it’s not because they’re the market leader. It’s not because they have a good business model. It’s because we think the market might be too small. Those are mistakes, too. We always underestimate the size of a market, and we have fun stories about that all over the place.

We do. I just did a show where the guest talked about the TAM trap: SaaS is like Japan, with a shrinking population, shrinking seats, and TAMs actually being smaller than we thought.

Whether it’s your Dropboxes, your Twilios, or your PagerDutys, I think many of the incumbents—I call them the new incumbents—are in a much better position, I would say, than the legacy incumbents when SaaS came along. If I were to rank-order the level of disruption that’s coming for these companies, business-model shift is number 1. We can talk about examples where that’s most in practice today.

Sarah and Kimberly from our side led investments in Decagon. Customer service is the most obvious one, where you can certainly price based on completion of a task, and it’s a better, faster, cheaper value prop for the customer. So, if you’re going to compete with a seat-based customer-service thing, look out—that’s hard.

That’s a business-model shift, so that’s the most disruptive piece. The 2nd most disruptive piece, I would argue, is UI and workflow. The 3rd most disruptive piece is access to data: what data do you actually access?

If all 3 of those undergo major change at the same time, I think you’ve got a really good chance for a startup to come and beat the incumbent—the new incumbent, if you will.

At the same time, I just never in a million years would have thought that the big software companies could be as large as they are. I have to think that this next wave probably presents the opportunity for this next generation to be much larger than the previous generation.

It doesn’t mean they have to go eat all labor, which we have on slides, too. I don’t think, in practice, that’s actually what happens. I think, in practice, what actually happens is massive surplus gets delivered to end customers, and you can still create much bigger companies than the previous generation.

I do think it goes back to this great question, though: we have to see the transition of spend from technology budgets—or, sorry, from human-labor budgets—to technology budgets. If we don’t, then the TAM for technology spend just stays the same, and we’ve all just overpaid a ton.

But the problem is that this has to be product-driven, not top-down-driven. That needs to be pulled from the market—it needs to be slapping the customers in the face that there’s the value prop for them to go do that—as opposed to CIOs or CEOs saying, “We need to do AI stuff,” and so let’s shift labor spend.

Yeah. I think it will happen. I’m saying you’re seeing green shoots. I don’t think it necessarily means that every SaaS company is doomed, but even Microsoft has reduced its headcount by 6% over the last year or so.

There are a couple of companies that have started to show signs of actually running their business differently and showing really high ROI from AI. Have you heard of C.H. Robinson?

No.

Yeah, it’s a truck brokerage. They take customers who need to ship stuff and trucking companies, and they broker deals between the 2 so that they can ship things.

Most of the industry in the U.S. is actually intermediated. It’s not direct; the trucking industry is very fragmented. This is a large business, and historically they’ve had football-field-sized call centers of people making phone calls and connecting dots.

They just disclosed in their last earnings that they saw a 40% productivity increase—measured in shipments per person per day—in their core business since the end of 2022. A 40% increase. It’s incredible, and it’s AI-driven. What’s actually happened is their operating margin has gone up 680 basis points.

That’s a very effective implementation of AI. People always ask, “Is there real usage? Are we in a bubble?” All this stuff.

But that just proves what I said to be true, though: the transition of human labor to technology is fundamentally necessary for us to have a great business.

Yeah, and I think it will happen. I’m saying you’re seeing green shoots. I don’t think it necessarily means that every SaaS company is doomed, but even Microsoft has reduced its headcount by 6% over the last year or so.

I do think it means you’re going to tap out, though. Sadly, I’m a big shareholder in Monday.com and Duolingo. One of our recent guests, who’s a dear friend of mine, was like, “Yeah, but that’s exactly the problem: there’s no human-labor replacement there.” Unless you have a human-labor replacement story in public markets today, you’re not going to get the premium.

Yeah. I think we'll see that, and I think it will come with a business-model shift. You're talking about the public markets: in the public markets today, you are guilty until proven innocent. It's the full flip side of our criminal justice system, where you are assumed to be doomed by AI unless proven otherwise.

7. Does Revenue Matter as Much in a World of AI?

I think there's probably an opportunity. You can see the way the stock prices have gone. Fortunately, that's not our world. I don't have to play in that world; we get to bet on the next thing. But I do think there's going to be a huge opportunity to shift that.

Speaking of huge opportunities, some companies are taking advantage of them, and the revenue scaling, dude, is just so much faster than any of us have ever seen before. We see the race to $100 million ARR. I think you guys just did Gamma, which is a product—an awesome product—and Grant scaled very fast to $100 million. Does revenue mean as much as it used to when it's gained so quickly and also seems so transient?

8. Does Kingmaking Still Exist in Venture Capital Today?

Okay, so this is a great question, because I think this is where you have to be really discerning in the market. It does mean the same as it has before if it is high retention and high engagement. The bar has actually gone up significantly for us when we look at AI companies because they have grown so fast. You can't actually look at years of renewal behavior, but you can look at shorter cycles of retention.

Most importantly, you can look at engagement. If people are using the product a lot and getting a lot of value out of it, that's a really good leading indicator, and we can take comfort in that. But we have spent way more time focused on that than we did in the previous generation.

So, what makes companies like Gamma so special? Again, this is one of Sarah's deals: 1) heavily organic customer acquisition, and 2) really high engagement and retention. We talked about the engagement and retention piece. It's magic when you have ease of customer acquisition.

You and I have talked about this before, but this is one of the most impressive things that we're seeing in AI companies. ElevenLabs has this, ChatGPT has this, xAI has this, Abridge, Harvey—companies where the market is just absolutely starving for their product. That's a really good sign. Just because it grows really fast doesn't mean it's going to end up transient or lower quality, but the bar for assessing that is way higher than it used to be.

Totally get that. The bar for other companies is also way higher, it seems. My question to that is, dude, I'm sitting on a lot of great enterprise software companies. We were—I was always taught, dude, that you're going to get great funding if you triple, triple, double, double. Is triple, triple, double, double dead in this new world?

I don't think it's dead in this new world. I tend to think that the number-one way to measure a company is ultimately return on invested capital. The way you do that with an early-stage company, mostly, is efficiency of customer acquisition. Not every company needs to go from $0 to $100 million; it depends on what market they're in.

But I do think that with AI companies, if there are very starving end customers, momentum gives you a chance to build a moat. I think that's the most important thing about the debate about how high of growth is good enough. It depends on the market you're in. In some markets, they're not going to move as fast, but in the markets that are moving really fast, if you're not moving really fast, that's a risky place to be.

I think the most important thing about momentum is just that it's relative to your peer set. If your peer set is growing really fast and your direct competitors are growing really fast, and it has high retention and customer acquisition is relatively easy, you need to be growing really fast, too.

But the opportunity cost of cash is so real. We were talking about this the other day with the company internally. I completely agree: it could be a very good way to build a very solid business over a long period of time. But the opportunity cost of my cash is that I could be in the next Gamma, Harvey, Lovable—you name it. So, yeah, it's good, but is it the best place for my precious dollars and for my LPs' precious dollars?

Yeah. I think we spend all of our time thinking about where there is market pull, right? Those are the best places where you can build a company. All those companies that you just described have extreme market pull. The reason they've grown really fast is not because they've poured tons of money into hiring sales reps. The reason they're going very fast is because there's tremendous customer pull for them. So we look for those markets.

I believe that kingmaking does exist. Kingmaking, for those that don't know, is when a financier is able to invest so much that they are able to anoint a winner in a category, and that then leads to moats and everything that comes with it, and ultimately winning. Do you believe that kingmaking exists, or do you disagree that it exists?

As we think about investing in companies, we always seek to invest in the winner. If the investment thesis is, “Our investment is going to make them a winner,” it's probably a pretty flimsy investment thesis. An investment that we make in a company that is already attracting resources, hiring really well, able to raise capital well, and able to deploy more money into go-to-market and more money into R&D, it can generally help.

This is the whole theory of preferential attachment, which is why increasing returns to scale is a concept, right? Even if you're not a network-effect-driven business, if you're Salesforce.com, Workday, ServiceNow, or CrowdStrike, the more you become the leader, the more resources come your way and the easier things get for you, potentially. So we look for situations like that.

I would contrast it with situations like the original SoftBank Vision Fund. They did a lot of really good things. Honestly, they did a bunch of really good things.

I genuinely want to be educated here, because I immediately shivered.

They were early to figuring out that there would be a huge opportunity in AI. They famously had NVIDIA in that fund. They did some really good investments like Slack and Guardant. The one piece that was missing, in my opinion, was that capital as a weapon was a viable strategy.

Capital as a weapon in enterprise is really, really hard to do because you physically have to hire people. You have to hire sales reps, you have to hire marketing people, et cetera. Capital as a weapon in consumer, most of the time, doesn't really work. TikTok is maybe the exception, maybe Uber.

The thing that maybe was wrong about it was the idea that we can kingmake if we just put the capital into the companies, and then that will allow them to win. But that's a bit of an adverse-selection machine, where the companies that opt into that as their winning strategy are the ones that maybe don't have as good of a reason to win or competitive advantage in the first place.

If that money is going to go back to consumers or drivers, or whatever it is, in that case, and just get funneled back to Google and Facebook, I don't think that kingmaking for that is necessarily a good strategy. But investing a lot of capital and having a brand that gives a seal of approval can definitely help make a company succeed.

I think Mark and Ben have described it well in the past. What are we giving to our founders? Partially, what we're giving to our founders is a loan on our brand, a seal of approval, and, often, especially for early-stage companies, it really can help with hiring.

Have you heard that talk track of, like, “What store do you want to be?”

We can talk about that. There's sort of a barbelling of the retail market, and there's basically Amazon and Walmart on the one end. On the other end, there's extremely high-end retail. So—

We call that Chanel, but, yeah.

Yeah, Chanel, Zegna—this is where Europe really thrives. There are scale players, and then there are specialists. We're obviously a scale player. I think the risk is everything in between, right? Department stores that have general merchandise but don't have scale, for example, and that's a very risky place to be.

Our strategy is very much scale—build scale—and the reason we do that is because it gives a huge advantage from a resources standpoint to our portfolio companies.

Totally get that and understand. Do you mind being, like, cool Walmart then? I don't mean that rudely, but I love you, dude, and that looks like a beautiful Loro Piana. There's nothing about you that screams Walmart.

We're happy to call ourselves Amazon.com. Customers love it. You're very well served.

Oh, we're Amazon. We're not Walmart. Okay, gotcha. Yeah, yeah, yeah. We mentioned making competitive categories there. One I just can't get over, dude.

And I’ve tweeted this: the customer support category has, like, 50 companies. There are so many, and Bret at Sierra is obviously the OG of OGs of SaaS. Can you help me? Why am I wrong to be so confused by this space? There’s something for every vertical.

Yeah. Well, I think there’s a good reason why there’s excitement in the space. It’s better, faster, cheaper already today, with today’s model quality, the reasoning capabilities, and the cost of the models. You don’t need to believe in any future state of a different product or a different model capability. The functionality is there.

I think there’s good reason why we put on EBCs for our portfolio companies. Every time Decagon appears in one of these EBCs, there’s extremely high interest and, most of the time, conversion to a deal. I think the market pull and the market size are what’s most interesting about that space.

Jesse and Ashwin—again, this is not my deal. It’s Sarah and Kimberly’s. But they are special founders. They’re really, really good. They’re relentless. They’re the kind of founders that we really love to back.

If you look at SaaS and cloud markets, about half of them are winner-take-vast-majority—like, the overwhelming majority. In about half of them, there’s sort of a breakup of market share. For example, you mentioned Deel. The payroll market is not a winner-take-vast-majority market. There are many markets like this in SaaS and cloud, and so it’s possible that Decagon is the winner. They move really fast on product, and they win the market based on having the best product, the best distribution, and all the things that we talk about.

It’s also possible that it’s a more distributed market, sort of like payroll. Either way, the growth is staggering, and the market pull is staggering. Decagon for us is a great company. Love Jesse, love Ashwin.

But how do you think about the willingness to pay up ahead of time? Because that’s kind of where you’re going, which is that you’re just paying so far ahead of time. $10 billion for Sierra is pretty amazing, but you legitimately are paying a shit ton ahead of time.

Yeah. I don’t know. We’ve not been close to that. Obviously, we’re existing investors in Decagon, so it’s hard for me.

But you were with Decagon. I’m like, how do you get there? Do you just say, “Okay, well, if the market continues in this way…”? Because if you map out expected growth rates, you have to map it out with a freaking Excel sheet to see where this lands.

I’m not sure I would agree with that, actually.

Okay. I’m taking a company—not Decagon—at $50 million in ARR. You have to expect that it will 5× to get to $250 million, then 4× to get to $1 billion, and then 3× to get to $3 billion, which are all pretty optimistic growth rates. And then with a 6× in the public markets, or 7×, we’re looking at, what, a 3× on the cash at the price that we’re paying today? Wow, that’s not a good opportunity-cost dollar spent.

I’ve been historically surprised at how good the best companies can be and how fast they can grow, especially in markets that are early innings with a big technology shift. So I’m very optimistic. Those are abstract numbers. I also don’t think that every great, high-growth company will end up trading for 6× in the public markets. There are some that are going to trade higher based on very high growth rates or high cash flow, and so it’s hard to debate an abstract financial case.

For most of these companies that we’ve backed—these winning apps—they’re growing 3× faster than predecessor SaaS and cloud companies. Sure, high valuations from the outside, I think, in many of those cases are warranted.

9. Do Margins Matter Less Than Ever in an AI-First World?

Everyone shits on them for margins. Do you think that’s a really weak argument to shit on AI apps? And do you think we’ll just see the transformation of those margins pretty quickly over the next 2 to 5 years?

The history of technology inputs would suggest that the margins will rationalize and go up. There’s a high amount of uncertainty today, so it’s possible that this next generation of companies has 50% gross margins. If they’re delivering a ton of value and growing really fast, that’s totally fine.

Today, the input costs per token have gone down massively, but token usage has also gone up massively with the introduction of reasoning. So in the last year and a half or so, it’s been a bit of a muddy picture on the input costs. I think over time that will rationalize and go down.

I think the market structure will end up sort of like cloud for the models, where cloud costs for the average end customer are fine, and cloud is an oligopoly that makes high profits. I think the model companies that serve APIs will be relatively oligopolistic. They’ll probably have reasonably high margins, and the end customers will be pretty well served.

On the gross-margin point today, I’ll say this: we give a little bit more of a pass than we used to. If we ever see a company that pitches us as an AI company and has SaaS gross margins, we ask a lot of questions, because it probably means that people aren’t actually using the AI features.

I do want to ask this, dude, because I did listen to the show with Patrick, and there was something that struck me. You said you look for greatness lying where others don’t, and kind of the art of the pick—more like determining beauty where it’s not obvious. I thought that was kind of interesting. Again, you can shit on me for this, but your biggest positions in Stripe and OpenAI struck me as not exactly diamonds in the rough.

What I mean by finding beauty or opportunity is that most of the time, it’s seeing a magnitude of greatness that isn’t totally obvious on the surface. When we’ve made original investments in some of those companies, we invested in Anduril in the growth fund when they had 1 program of record, and it was border towers. Now they have many, many programs of record and some of the coolest products in the market.

10. My Biggest Miss: Anthropic and What I Learn From it?

We invested in OpenAI before they had ChatGPT. Often, there’s an opportunity where we see things that may be great in the future, even if the companies themselves are already great or hot.

We’ve talked about errors of omission. What error of omission lingers on your mind? What company are you not in that you would most like to be in, and why? For me, it’s Revolut. It actually upsets me every day that I’m not in Revolut. I use it, I love it, and it upsets me. I have a lot of errors of omission. I have many dating back deep in my career.

The ones that really linger for me are Revolut and Deel.

Yeah.

For current companies on the model side, Anthropic has done a really great job. We’re not investors in Anthropic, and they’ve done a really good job. I think it’s one of those cases where, similar to cloud, if you could own all of AWS, Azure, and GCP as independent companies, that would suit you pretty well. Again, that’s one of those markets that was not winner-take-all, even though it’s a scale market. It’s sort of oligopolistic.

11. Has OpenAI Won Consumer AI? Will Anthropic Win Enterprise?

Do you think the market will evolve with OpenAI winning consumer and Anthropic winning developers and B2B?

Yeah, I think they actually will diverge in pretty meaningful ways. This is sort of what we’ve seen in historical technology markets, but I think each will try to remain competitive in their spaces. In B2B, Anthropic is certainly putting more resources after it today. OpenAI is going to have a really good B2B business; they already do. So I think that market is going to be pretty competitive—not just coding, but general B2B API usage and moving up into the application stack. Both of them are obviously trying to do that.

I think Google will play some part in that market, but the big head-to-head competition will come between OpenAI and Anthropic. On the consumer side, I think it’s ChatGPT. Ask my family in Kentucky what they use. They know what AI is; they know ChatGPT. They use ChatGPT extensively.

Google is going to take a crack at it, and they already are, trying to compete in that market. But I think brand and the best product in the market can take you a really, really long way. So as we have underwritten future rounds of OpenAI, or later rounds of OpenAI, it’s very much with consumer in mind.

At what point does the entry price for OpenAI, do you think, become not a good use of dollars? This is one thing where I’m permanently reflecting on it myself. If you think it’s a $2 trillion company, you can still see a 4× from here.

At what point does the opportunity cost no longer make it worth it? We have to constantly reassess this. So, again, you should look back at our investment case for investing in Databricks in 2019. We did an investment out of our growth fund. It was one of our first investments, the largest growth fund investment in Fund I, at $6 billion.

Our investment case never would have predicted what they became. And so, we have to constantly push ourselves and think about how big they can become. I’ve been surprised at how big, in absolute dollar terms, the companies can be and how good they can be. So, we constantly have to push ourselves on this.

The example I always use is Google and Facebook. 10 years ago, Google and Facebook were monetizing their users at like 1/7th of what they are today. It’s hard to forecast that. It’s hard to model that.

But it would be limiting to think you’re ever at a steady state of productivity or a steady state of new products. So, we’ve been surprised. We like to invest in the ones where there’s a theory that the core market can be bigger than we would expect or others would expect.

Stripe is an example of this. SpaceX with Starlink is an example of this. Waymo, when we invested, is an example of this. We also like to invest in the ones where we feel like the founders have an advantage in figuring out the next product.

Anduril is a perfect example of this. We knew border towers would be a huge product line. But with the team, we were also pretty high-confidence that they were going to figure out a bunch of other stuff. I wouldn’t have predicted that they’d figure out autonomous fighter jets, which is pretty awesome.

The best ones—the best ones who know their markets the best, who have market leadership, who are product people and tech people—they tend to find the next product areas. That’s what we want to find at scale.

At the scale you are, do you just say, “Hey, we have to invest in competitors”? You can’t not.

No, we don’t. When we invest, we try to avoid conflicts as best we can, especially if we’re on the board. That’s the trickiest part of the scale of our business, and we don’t always get it right there.

You delicately do it between funds and say, “Oh, that’s in the early fund.”

We try not to do that. I mean, look, the thing that we see more often is that companies diverge more often than they converge. The perception of what a conflict can be in the future often doesn’t come into play.

There are also examples in the opposite direction, where we funded a company and then they pivoted into a different space. We try to help the founders as much as we can, even if that’s the case.

Can I ask you what decision you, Mark, and Ben most disagreed on, and what was the outcome? Where were your views very divergent, and how did that resolve itself?

The biggest one was our original investment in Waymo. We invested in Waymo in early 2020, so we were the only VC fund that invested in Waymo in early 2020. It was extraordinary. The product was magic even at the time.

We did demo rides. This was obviously well before they were everywhere on the road. It could drive smoother than a human. It could do unprotected lefts. It could avoid construction sites. It could do all these really special things that you wouldn’t think an autonomous car could do at the time.

But at the time, they didn’t have a product in the market, and I thought the valuation was really high. So, I said, “Here’s all this analysis.” Our team produced all this analysis that showed that the price was really high, and Mark and Ben were like, “It’s autonomous driving. What are you talking about? This is the endless market size. This can be the biggest company in consumer technology, and they’re the market leader.”

The way we did it was that we invested a smaller amount at the time, given our conflicting points of view on it. But that served us well because we kept a close relationship with the team, and we wrote a much larger check into their most recent round. I’m really excited about it. They have a very exciting future.

I’m going to San Francisco after this, and I’m going to take a Waymo on the freeway up to our office in San Francisco from Palo Alto. That’s sort of a magical product experience.

This is one of those cases—we talked earlier about potential future competition—where there’s going to be tremendous potential future competition, but the product in the market today is magical.

I’m always quite annoyed about it because I always see it on social, and we don’t have it in London. I’ve never been in one.

They’ll try to get to London soon. London’s a tough market to enter. You remember what it was like for Uber to enter London in the first place. It got brought into the market kicking and screaming.

12. Why Did You Invest $300M into Adam Neumann and Flow?

But London and Tokyo will be some of the best international markets possible for autonomous driving.

That I understand. What I just don’t understand, and I would love to, is Flow. Can you help me understand Flow? I think the world kind of scratched its head: Why did it make sense to you when it didn’t make sense to anyone else? You remember what I said earlier about investing behind strength of strengths?

Yeah. Adam has extraordinary strengths. He has some of the strongest strengths of anybody—any entrepreneur—in the market. It doesn’t mean that he has no weaknesses, but he absolutely spikes in the areas that are most important for the business he’s trying to build.

What would you say those are? I’m not in those meetings, and no one is, and I’m fascinated by product and hiring. When I say that to the world, those sound like things that are maybe a little fuzzy.

But they’re not. I mean, they’re the most important ingredients for early-stage company building. He’s surrounded himself with an extraordinary team.

He’s got an incredible insight, which I think is fascinating. Consumers in the US—obviously, homeownership is declining rapidly, and people aren’t able to buy homes. There’s a whole political and social issue with that, but it’s the reality of the case.

The average renter in the US spends 30% of their disposable income on rent. It’s the highest amount of spend of any category, and yet it’s the only unbranded experience in anyone’s life.

If you think about the food you eat, the clothes you wear, the car you drive, the places you go, all of those are branded experiences, and consumers pay a premium for that branded, better experience. His idea was, what if you actually brought brand and a better product experience to a renter’s life?

There’s a huge market opportunity for it, and there’s a great business model that goes with it. If there’s anybody who can do that, given the intersection of real estate and brand, I think it’s Adam.

If you think about the average entrepreneur who walks in off the street and pitches us an idea, what is the likelihood that Adam can build a humongous company versus the average entrepreneur? It’s extremely high. It doesn’t mean that it’s without risk, but he has extremely strong strengths. So, that’s the theory behind it.

The founder of Calm, who I walk with every week, always asks me one question. He goes, “How often do you meet a founder like this? Once a month? Don’t write the fucking check. Once every 6 months? Probably write the check. How often do you meet a founder like Adam?”

I don’t know. You can answer me on your data set, but it’s probably quite rare. In which case, you’re like, “Well, then write the fucking check.”

Yeah, yeah, yeah. It’s extremely rare. Adam is a learner. He is a deep student of the game that he’s in.

So, I’m really excited about Flow. Mark, Ben, and I are all involved. Justin from our team is involved as well. They’ve sort of proven out the value proposition of the product, and now it’s just about scaling.

Dude, can I do a quick-fire round with you? What have you changed your mind on in the last 12 months? Mine was Anduril.

Oh, that’s good. I like it. Anduril and YC. YC is the single biggest buy, I think, in venture.

Every great European company is a YC company. They've crushed it internationally, for what it's worth. I mean, they're really, really good in the US.

Yeah, they're really good.

And I'm a big fan of Garry.

I don't know if it's in the last 12 months, but if you think about the moment that all of the models started to demonstrate their capabilities, I would say there was a moment in time where we thought that the models would eat everything in consumer and enterprise software. And I think maybe there's a bit of a shift back toward this in public markets, at least—that the models are going to eat all these application-software categories.

We fully changed our mind. I think there are going to be application-software companies built on top of models in pretty much every direction. If you look at our investing behavior, it obviously reflects that.

That's probably a little bit further back. That's probably more like 18 to 24 months ago. We all thought at first that the models would just do everything and subsume everything. It turns out there's tons of stuff you have to do around the tasks that humans do in order to build a viable product.

The example I like to give—I know it's a lightning round—is radiology. AI has been able to do a better job than human radiologists prior to this whole wave. Neural nets were able to do a better job than human radiologists at looking at scans, and yet, since the proliferation of AI, the number of radiologists has actually gone up; it hasn't declined.

So why is that the case? It turns out that radiologists only spend 30 to 40% of their time looking at the scans. There's another 60 to 70% of their time doing all the other stuff, and the model companies aren't going to do the work to figure out how to automate the other stuff—the 60 to 70%. But that's what the opportunity would represent for an independent company in that space. Does that make sense?

I know it does, and I 100% agree with that. We've got a business called Solve Intelligence, which is patent-law AI. No way they're going there. Agreed. But OpenAI are doing customer support. Gemini and Google have just released Firebase Studio, or whatever the fucking Lovable competitor is called. They are moving into the application layer in ways that we didn't know they would.

Yeah. But it's one of the 30 things in their AI divisions that they're trying to do. It's sort of like how AWS and the cloud have service offerings for basically everything that you could possibly have. And yet there are still tons of infrastructure companies that are independent.

Dude, you've met many great founders. What's the one first founder meeting that was most memorable? I'm not asking for the best founder or anything like that. I'm just saying, the most memorable first founder meeting.

Okay, so there are more extreme-success versions of founders that I've backed and gotten to know over time. One of the ones that struck me recently was the first meeting I had. I had dinner with one of my partners, Santiago, with Shiv from Abridge, and I didn't know what necessarily to expect.

I knew he was a doctor, a practicing cardiologist. I knew that he was making a lot of progress in his market, but he was one of these perfect archetypes where he knows his end market, he knows his product, he knows the technology, and yet he's a total, total killer. He's got great bedside manner as a cardiologist, but he's an absolute killer.

I love when I have those first meetings and you can already feel that.

Dude, he actually reminds me of Winston at Harvey, which is like, you feel the authenticity to the core domain, but then it's not the elegance of that domain—it's the aggression of a tech founder with the academic nature of the core domain. Do you know what I mean?

Yeah, of course. This is actually a really good archetype in a lot of the vertical-software categories. You can definitely see it with those folks. Honestly, speed of execution and aggression are a huge part of success in those categories.

Totally get you. You've got a seed firm, a Series A firm, and a growth firm that you have to invest in. Other than a16z, which do you put your money into?

Obviously, 20VC. Very sweet, thank you. So I can help you out. For me, I put my seed in Hummingbird, my Series A in Benchmark, and growth in either you or Pat. I'm not just saying that, but I think scale is super important and brand is super important. Or Napoleon at Founders Fund. I think those guys are all great. I have tons of respect for all those guys. We end up doing rounds together. We're in companies together. I think they're all great.

Who's not in a16z who you would most like to work with?

I think the best would be Nat and Daniel. We partnered a lot with them when they were investing. They got back on the field to do real jobs now. But if they were to come back off the field, I think it would be fun to work with them.

Mine would be Lee Fixel. The guy's ability to predict and forecast markets, like a 10-year vision plan, I think is really amazing. Or Fenton's clarity of thought. Fenton could make a fucking plastic bag seem like it was made by Jesus. Seriously, it's amazing. Anything just sounds poetic. Who's the best picker in a16z?

There are a bunch of really, really talented people at the early stage. I love that I get to learn from these people all the time. I think the people at the early stage who have developed the most clarity of thought on their approach to early-stage investing—I think it's Dixon.

He obviously runs our crypto funds now, but he's got a generalist background as well. He's been doing this for a really long time, and I think he has the clearest articulation of what our early-stage strategy is, which has been adopted, I would say, across the firm. But I think he has the clearest view on it.

When you need to win something internally at a16z, who's the savage that you bring in to win?

Mark and Ben.

You can choose one.

They're both exceptional. It depends on what the founder wants.

How does that differ? I'd love to know.

I'll tell you what the spikes on both of them are, from my vantage point. They're both exceptional at every element of the job.

Mark can see the future. If you ask Marc for any 10-year prediction, they're very often right. Most of the time, they're right. He's often high on magnitude, and it ends up being justified in the future. Things that may seem too high or too crazy—in the fullness of time, he's generally right. He knows consumer internet extremely well. He spikes there.

Ben is probably the best management coach and has the best understanding of executive dynamics and problems that I've ever encountered. He also is a futuristic thinker, but he spikes in that way. Mark spikes in seeing the future.

Who's the most helpful post-investment? I think about the pillars of venture, which are finding, winning, and helping internally. Who's the one the founders just love? Who helps so well? Who internally is the best helper? Obviously, you have portfolio services—I get that.

I can't pick one of these. There are too many. It would be unfair to pick one.

It's funny. I wouldn't mind. I'm not a helper.

Yeah. No, neither am I.

But ultimately, if you could change anything inside a16z, what would you change?

I wouldn't change this, but one of the elements about us scaling has been that we've had to decentralize the way we run our business. When I first joined the firm, we used to sit around in partner meetings all day on Mondays and hear all the pitches from all the various sectors, and then on Fridays too. Obviously, that's not a scalable approach to doing venture, especially given that we're in a bunch of different sectors now.

But selfishly, on the growth-fund side, that was extremely high-signal: great information, tons of soak time with all the best thinkers.

Did it not make you a better investor, seeing that and having that view?

I think we have to go out of our way to go get that information and signal now. It actually has made us a better business at the early stage, and then, as long as we're coordinating right from early stage through growth, it'll make us better. But we have to seek it out and do a little bit more work to get all that information.

Final one for you, dude: What are you most excited about for today?

I look at the world and I'm very excited for the first time. I'm like, I can tell my mom that there's hope for MS sufferers, that there might actually be treatment. I like tons of optimism. I like happiness. I think there's not enough of it in the world, despite my cynical disposition most of the time.

What are you most excited for?

On the personal side, I'm really excited. By the way, these are 2 areas that I think over the next 10 years are going to be really exciting and really investable, but they're kind of early today. One is personal health. It's a little bit related to your point, but more health management.

I was with a really talented entrepreneur. He's a former large-company executive, and he's thinking about starting a company. His extreme version of it was tracking and AI coaching that happens for you and explains the trade-offs of every decision you make. That's a little bit too extreme, but more proactive, more involved management of personal health is something that's going to happen.

It's one of these large consumer categories that hasn't really hit yet, but I think it's going to happen.

Can you imagine wearing a bracelet and every time you picked up a cookie, it's like, “Heart disease. Heart disease”?

You just took off 17 minutes of your life.

I think that's too extreme, but I do think there's a positive version of that that could be super valuable. It would be good for society, but I would love it as a consumer, and I think the technology capabilities are going to be there pretty shortly.

The other is robotics. We have not made a large investment in robotics, but I think it's going to be the largest category in AI—B2C, B2B. There's still debate on what the right form factors are, whether it's at-home help, whether it's industrial—all these things.

I do think 10 years down the road, we're all going to have really helpful robotics assistants in B2C and B2B. I think it's going to be super exciting as a consumer, but I also think as an investor, it's going to present some awesome opportunities.

I'm going to be honest. I feel pretty guilty because I freaking love you. You're such a lovely, wonderful dude. You really are. I feel like I just battered you with hard questions.

You went hard.