Anthropic募资130亿美元，OpenAI以全股票11亿美元收购Statsig

1. Anthropic Raises $13BN: The Analysis?

Now, I just want to dive straight in this week. Rory, as always, I'm diving in with a late edition because you love late editions, but we had to: Anthropic raises $13 billion. It started with $5 billion, moved to $10 billion, and now it's $13 billion at a $183 billion post-money valuation. Wow, it's a lot of money. That's a high price. Rory, going to you first because you prepped for this one.

Guest

Absolutely. Well, funny enough, I did because it was a little bit anticipated. I think, to some extent, it's a high absolute number because $13 billion is a lot of money, but is it a high price? It's interesting, and we talked about this a little last time, relying on reported numbers.

If the growth trajectory is really $100 million two years ago to a $1 billion run rate starting this year to somewhere around $5 billion now, maybe $8 billion or $9 billion by year-end, let's just take that as roughly true. Let's say $1 billion and $9 billion, which typically means GAAP revenue is roughly the average of the opening and closing ARR. That means GAAP revenue could be, again subject to the numbers, around $4.5 billion this year.

What are they going to do next year? If they go from $1 billion to $9 billion—and these are the facts. Up until now, all I've talked about are facts, things we know today, right? The million-dollar question is: What does today's trajectory say about the next year?

You throw a rock forward and it falls down, but it moves forward and falls down at the same time. How much does 9x revenue momentum this year persist into next year? Do the math here. Even if they go from $9 billion to $30 billion, which is 3x growth, down from 10x growth, then GAAP revenue—the average of $30 billion at the close and $9 billion at the opening—is around $20 billion. This is 8x NTM revenue and FY26 revenues.

The stunning thing with this growth rate—big underline—is that you're only buying in at 8x to 9x next year's revenues if the growth lasts. Now, will it? That's the billion-dollar question. But high absolute number? Absolutely. Does it make sense? If you think the growth rate's there, then it's not crazy. I did not go into that math expecting that answer. I went in saying, “Oh, aren't those guys so silly paying so much?” And you look at the numbers and go, “Maybe those guys are being quite smart.”

This round was also super-oversubscribed. I know folks in the industry have been clamoring over this, and it was hard to get into. They could have raised, I think, 10x that—10x the amount of funds that they put in.

Guest

You're exactly right. If you're a growth-stage investor, a big growth fund, it's hard to imagine a world where you say, “We're growth-stage investors, but the 2 largest market-cap companies in growth in the last 3 to 5 years—the 2 LLM models—we don't have a piece of that.”

So there's probably some kind of huge corporate imperative at every growth-stage firm saying, “We're either taking a big-balls call that this isn't going to work, or we need to get 1 of those.” So, yeah, I imagine you're right, Cliff. I'd say there was huge demand.

As insane as, on its surface, Anthropic at $160 billion, Databricks at $100 billion, and Canva at $42 billion sounds at first, if you do the math, there's a multiple component.

That's barely 10x for us. We sound very cheap.

That's the point. It is cheap. Literally, if you're using forward ARR multiples, these are not especially high, especially if you use forward ones. To Rory's point, if you use next year's Canva, Databricks, and Anthropic, they seem reasonable, right, as long as the growth can persist.

Canva, again, I'm a superfan since the old days. I wouldn't have thought it would be this big, but the growth at scale is epic, right? It's crazy, Canva's growth. It's not growing 8% today, right? It's growing 5 to 6 times that. I don't know if you predicted that in the old days, Cliff, but it breaks your rules of TAM, right? And it's not even Anthropic.

My question is: We mentioned the oversubscribed element there, Cliff. When you literally have a 5x oversubscription—I'm sorry if I'm being naive here—but in the same way that Dario did, how do you literally choose which dollars you take?

It's very tough. For us, we priced our round before Figma went out and had all those conversations and relationships, and so the Figma IPO kind of threw a cat amongst the pigeons in regards to proving that we will close the year very close, if not at, $4 billion, growing close to a 40% growth rate and reaccelerating growth. We are compounding growth at scale, which is a good place to be.

When it comes to investors, we really need to think long term. We have a lot of long-term partners, and you want to pay a lot of loyalty to the people that have supported you along the way. But we're also thinking through what an IPO looks like. Who are going to be the cornerstones of that IPO? How do we see this not as a point-in-time deal, but as a relationship-building exercise through the next 18-to-24-month period with these really long-holding investors? How do we instill trust in them and for them to trust us as a leadership team that can take this through IPO and beyond and make good decisions?

We're not looking to ratchet up the price. We're not looking at playing any silly games. We're really seeing this as: How do we build these long-term relationships that are going to be with us a while?

Do you have Fidelity?

Yep.

It was super interesting. We had Brian Halligan on from HubSpot, and he was talking about the central role that Fidelity played. I actually wasn't quite as aware as I'm sure Jason and Rory were of how important and strategic that was in terms of aligning them for when you are public.

I'm not sure how much I can say, but, yeah, they're the anchor of the round. They're the largest check in this round.

I thought you promised that to me, though, Cliff.

I said I can shake some trees. So, the problem now is that this is all secondary. We've got over $1 billion in cash in the bank. We don't need to raise money in primary funds. We've been a profitable company for 8 years.

When we go out and do this employee secondary, and also we have some investors that want to sell, then they see Figma go out and a lot of that sell-side demand dries up. So, even though we're already oversubscribed, we've got this supply-and-demand imbalance at the moment. It's an interesting dynamic to work with. It's a high-class problem to have, but you don't want to disappoint people.

How do you coach employees on that? Do you stay out of it? Because, especially when you see Figma go out at that multiple, it's tough for employees to process the decision, isn't it?

Yeah, I like to be very transparent, and I think Figma is an absolutely incredible company. They don't have a massive float. There are dynamics to any float that can make things go higher or lower.

What we do is show them a spread of public companies. We show them their growth rates. We show them how we really think about it. If you compare us to some companies, we're undervalued. If you compare us to some, we're sort of on par. We give them that spectrum so they're not just taking 1 single point and referencing all their marks to that.

We're also talking through the long-game nature of this. We have been through trials and tribulations ourselves. In 2021, we were worth $40 billion, which was, I think, a 50x multiple on our revenue at the time. In 2022, everything came crashing down. The market came crashing down, and that took us down to $26 billion. That was a tough pill to swallow when we thought we were riding high.

We've slowly just compounded that growth. The company hasn't stopped growing. We were still profitable. All of the foundations were right. That's what we really focus on and educate the team on. The markets will do what they're going to do, but as a company, we can compound growth, we can compound margins and increase margins, and we can deliver value to our customers first and foremost.

And you're exactly right. Just talking about that, even the little example you gave makes 2 things clear: Very smart people with MBAs swore blind you were worth 50x ARR in 2021 and 20x ARR in 2022, and now you're at 10x ARR. So, I—

Funny point: It took discipline to take the 50x, because—I won't name names—we had people coming in at higher multiples, and even we were like, “This is batshit crazy.”

Totally. I did a post years ago, “Beware the madman in the back seat,” which is: Wall Street is the madman in the back seat.

Finance—we all do. You change your mind so drastically and so quickly, right? As you said, all you can say to the team is, “They’re going to do what they’re going to do. They’re going to 50x, they’re going to 20x, they’re going to 10x.” What you can do—and it’s very impressive—is recognize that finance can be wrong literally by 5x. We were wrong from 50 to 10, and you’ve been able to recover that valuation by just working hard and growing for 5 years.

In the end, the thing that bails out our incompetence is your growth rate. That’s the dirty little secret of the positives.

That’s what worries me about some of the AI companies now, because I’ve seen this time and time again with people trying to copy Canva or be Canva for this or Canva for that. There is an early-adopter syndrome that pulls forward a lot of revenue. I think one thing we’ve done well with Canva is cross the chasm to the mainstream—middle America, people all through Europe—not the Twitter sphere or the X-sphere, whatever you call it these days, that are always using the latest and greatest products and paying for that because they’re happy. Then there’s consolidation of those early-adopter products.

A lot of those products struggled across the mainstream because then it comes to distribution, and distribution at scale is a lot harder to reach those people in middle America, all through the globe, than people who are actively tracking what AI is doing, actively on Product Hunt or whatever, and on X seeing all that sort of stuff. So crossing from $50 million to $100 million to $1 billion in revenue—that’s a big leap, and I’m interested to see how some of the companies can navigate it.

Yes, there are 2 separate things embedded in that, though. One is just the sheer fact of going from $100 million to $1 billion is a grind. But the second thing—and I thought you were going to go there, and I want to ask you about it—is a separate comment: Our nagging suspicion is that some of the year-2 ARR renewal rates for some of these AI products will be pretty low. In other words, the AI product-market fit is getting covered over a little bit by AI enthusiasm early on.

I’d be curious: How are you guys thinking about your AI products? Are you getting the usage you want? Do you think users are growing them, or where in the AI adoption curve are you and your users?

Firstly, we’re all about creating workhorses, not gimmicks, in Canva. The mission of Canva was to empower the world to design, and that’s to take anyone’s idea and create a great piece of visual content, whether that be a video, a marketing material, a poster, a presentation, etc., etc. AI is just accelerating that massively for us, making it quicker, faster, and better for our customers to achieve their goals.

We already have a user base. AI is accelerating that. I think what you’re seeing now when it comes to companies applying AI into their products is that people threw a lot of shit at the wall and hoped it stuck. That was the right thing to do. Every company on Earth, hopefully, was running AI hackathons: What can we get into the product? Let’s test it.

Then it consolidates down to a small number of things that add true value, and there’s a lot of periphery stuff that is kind of neither here nor there. To answer your question in a different way, around AI consolidation and year-2 renewals, particularly around enterprise customers, if organizations were approaching AI the way Canva was approaching AI, the cost of implementing these tools—and a breadth of tools, as long as they meet our security requirements—is negligible.

Our approach was spray and pray. Use all the tools. I’m happy to open up an extra $10 million to $50 million budget to essentially hope we can drive employee efficiency and get more done with the same amount of people using all these tools. I’m not going to be the arbiter of “This tool’s better than that tool.” We’re running 4 coding tools at once, right? Cursor seems to be the one that’s kind of leading the pack.

The same goes for all the LLMs. We give everyone a choice. You can only have 2—we limited it. You can’t have Gemini, OpenAI, and Anthropic. You can kind of pick 2. So we’re like, use what you want as long as it meets our security requirements. Over the course of the next 12 to 24 months, we’ll start consolidating down as the clear winners take charge. To answer your question that way, 100% there’s going to be consolidation down from year-2 renewals.

Sorry, I do want to retain some semblance of structure because you mentioned clear winners there, and we spoke about Anthropic and the large raise. On the flip side, today OpenAI announced that it’s buying Statsig for $1.1 billion in stock. It was the same price as the last round that Iconiq led. It’s an incredible team with Vijaye Raji—super-obvious matchup, given that Vijaye is obviously joining OpenAI.

The business is doing $75 million in ARR. My response to the team WhatsApp group, where one of our partners is an angel, was, “That’s cheap,” and I wanted to know how you guys thought about that. Do you agree with me? $1.1 billion in stock for a $75 million ARR company with an amazing team. Guys, how did you feel?

Guest

Let’s take the perspective of the person who just did the last round in May of this year. I valued this thing at $1.1 billion. On the other hand, I’m probably the same investor—in fact, I think Iconiq did the round. They just did Anthropic at $170 billion, so they might be very happy: “Oh my God, I got me some OpenAI now,” right?

Effectively, within 4 or 5 months, you rolled forward into the next OpenAI round and the next OpenAI valuation. Maybe that feels just as good. I’m doing the math in my head, but the revenue multiple might not be that much different. You’re kind of like, “Yeah, I thought I was investing in Statsig. Now I’m investing in OpenAI. Worse things can happen.”

Maybe the angels don’t like it. I think when I saw that the round was apparently exactly the price of the growth round from Iconiq, it just makes you wonder: If you’re just getting your preference, you don’t really care as a late-stage investor. You don’t care whether it’s $1.1 billion, $1.06 billion, or $984 million, because you’re making the exact same amount. The fact that it was exactly the last round showed this was something that everyone wanted to roll into.

Obviously, the CEO wants it. What’s his role? He’s like number 3 at OpenAI, right? He’s going to run it with Fidji Simo. He pushed the CPO aside, and now he’s running the company. That’s big for me. I’d rather run my own company. I’m guessing Cliff would too, but for 95% of people, this might be a quick upgrade without risk, right? So he got what he wanted.

Iconiq rolls over $100 million into OpenAI. They’re not allowed to invest because they led the Anthropic round, right? So this is the only way they can put 9 figures into it because they’re soft-banned or hard-banned from the round. Maybe the angels don’t like it because they wanted to play another card, but everyone else—the fact that it’s the exact price of the last round, not 2x, not less—it’s just perfectly engineered to check everyone’s boxes, right?

The people we didn’t name who did the A and the B—Sequoia—are presumably, as always, happy and successful. So there you go.

Cliff, here’s $100 million of OpenAI stock.

2. Is Zuck’s $14BN Scale bet the biggest blunder in AI?

I’m not doing it, but we love OpenAI. We love them, but we’re cutting our own course. We’ve had acquisition offers.

Dude, I will always put you on the spot and ask questions like that. Rory knows that well. The tough one is poor old Zuck. You know what I love is the transience of Zuck, where it’s like, “Zuck’s a hero,” and then, “Zuck, what a fool for buying Scale.”

The wheels seem to be coming off the Scale acquisition in terms of the talent that’s leaving and the satisfaction with the quality of Scale’s output. Everyone’s wanting to use Surge AI—Edwin Chen, great guy. I liked him a lot, actually. A Mucker portfolio company. Go Mucker.

Rory, they’re just very upset with the output of Scale not being good enough. Are the wheels coming off the Scale and Meta train, or is this the media overhyping this and coming back at Zuck in a way that’s just unfair?

Guest

I just feel sorry for poor Zuck. Imagine poor Zuck opening the newspaper or his phone every morning, looking at the news. It’s got to be a challenging life being the CEO of Meta.

Dude, you get stuff written about you in AFR. Does it upset you when you read it?

Guest

It doesn’t, but it’s not at the same scale as them.

My sense is his skin has been thick since the movie.

Guest

I think if you remember back in the day, he was pretty upset when the movie came out. I think since then his thin skin is about 5 inches thick.

You do stop caring, actually. We’re at such a smaller scale than them, but you do learn not to care. Rory.

Guest

Yeah. Look, I don’t think it’s profitable for me to speculate on how he feels. I don’t know and, frankly, don’t much care. It’s not my problem, right? If he’s not feeling great, he can cry into his $200 billion and get over it on his own, right? Take it up with a therapist.

The actual most substantive question is: Does this new information make you feel better or worse, just objectively seeing how the deal is going, right? Is there an actual, real, fact-based takeaway right here? There’s information.

It’s not wildly surprising. When the deal was announced, you go, “That feels like an odd way to solve this problem. Maybe it’ll work, but it’ll be messy along the way.” This just feels to me like there are 2 different shoes dropping here, and they both feel like exactly the shoes you expected.

The first shoe to drop is people. It’s hard to give people $100 million, then give someone else $1 billion, and then give someone else $10 million, and have them all work together. There’s going to be some fallout. Even if they’re all amazingly talented people who want to work together, people have egos, people are human, and it’s just going to be messy.

So the fact that some people are leaving, whatever, I don’t know. I don’t know about the retention package, but it’s just not surprising. Some of it could be directed. Remember, we talked only last week, going back to your comment, that we were kind of up and down on this because they seemed to be organizing the thing in at least a structured fashion: you’re in, you’re out; you’re in, you’re out. So there’s this kind of human fallout that was predictable.

The second thing is the asset itself. The comments about how the Meta team aren’t as excited about the Scale data-labeling stuff make sense to me. From what you read, the requirements of data labeling have evolved a lot—from something very simplistic, like “This is a dog, this is a cat,” to answering much more complex questions. The training data it takes to pass advanced biology and advanced math is a different thing.

Rory O’Driscoll

I’m sure Scale aren’t dummies; they’re trying to do it. But there are other firms, and you mentioned 2. One of them is your portfolio company, because that’s what you do: promote the product. There’s a bunch of others, including Turing. There are a bunch of folks out there. I’m in none of them, just to be clear, so there’s no agenda.

There’s a lot of competition, and if you’re sitting there as Meta and your ass is on the line to deliver, you’re not going to say, “I’m going to buy from what is now our biggest investment just because that told me to do it.” You’re going to buy from the best. So there’s probably going to be some of that: “We didn’t get what we want.”

That raises a third issue. If you remember the structure of this weird deal, they put $14 billion into Scale as if it was worth $14 billion. To your point, Harry, it was 1× the last round. Then the VCs promptly took out that $14 billion, leaving Scale as an empty shell because all the money’s gone.

There’s this remaining asset that we agreed wouldn’t last a year, but we had to pretend was like a company. Now Scale and Meta, on its balance sheet, has a $14 billion investment in a company that probably isn’t worth $14 billion anymore. There’s no cash, and there isn’t a great business.

At some point, the auditors are going to say, “You’ve got a $14 billion venture investment there. Do you really think the empty husk of Scale, without all the team that’s moved over to Meta and with all the team that’s left, is worth $14 billion? We’d like you to take a write-down.” That’s going to be the entertainment factor at the back end of this year or early next year.

So, yeah, it was a quirky deal. It has a bunch of problems. It’s just been a step on the journey to fuck up a $14 billion acquisition.

Rory, do you have confidence Zuck’s master plan will pay off, or does this leave you less confident than you were before?

Rory O’Driscoll

His master plan—I simply observe the facts. He has a master plan. He’s won already. He’s worth $200 billion, and he’s got one of the 7 most influential companies on the planet. He’s won already.

All you can say is that he’s had some big bets that have worked amazingly, like WhatsApp. He’s had some small bets that worked brilliantly, like Instagram, the best acquisition of the prior decade. And he’s had some big bets that have flopped, like the whole metaverse thing. My gut is that this is more like the latter than the former. I could be wrong.

Culturally, I think the simple fact is that he’s assembled a pack of mercenaries. He’s gone out and hired all the best mercenaries out there, and something forced them to report to each other—a weird structure, power struggles, fiefdoms. But he’s put them all together in a matter of weeks, maybe months.

When I was a B2B founder trying to be driven but touchy-feely, I was sort of anti-mercenary: if you’re not on my journey, I don’t want you. Canva’s been doing this for, I don’t know, 20 years—something like that. But as time has gone by, with more nuance, sometimes you need mercenaries, and sometimes there are cultures where it’s okay. Sometimes there’s a tool for the job.

I just think we can pick at this, and I think the criticism is justified, but Zuck knows this is a bunch of mercenaries. Some of them are going to fall in battle. Some of them are going to quit, and he’s given $20 billion or $30 billion to a pack of mercenaries. I don’t know.

I mean, Cliff, Canva seems anti-mercenary from the outside, right? But maybe there are times when you’ve had to hire a pack of them to go into battle.

As we look at big, classical SaaS companies, we’re trying to figure out their role in the AI world. How much of your reacceleration would you attribute to the stuff you did in AI versus just getting through 2022, finding your sea legs again, and executing?

I mean, Canva, for us, we’re growing faster than ever. So you’ve got to decouple valuation and company value.

I agree. Decoupling valuation—COVID was a mass discovery event. Everyone was sitting on their ass at their computer all day. It was great for Canva. But what was the question?

If you look—and I didn’t pose it clearly—but post-2022 and 2023, you decelerated, right? You’re right: leave valuation out of it entirely and just talk revenue. Now you’re obviously reaccelerating at huge scale, and you’re doing what every SaaS company before the AI companies wants to do. That’s the only way they’re going to be relevant and exciting again, and obviously every one of us owns lots of them and we’re trying to figure this out.

For you, do you think AI was the igniter of growth in 2024 and 2025? Do you think it was just execution? How much of it do you attribute to the AI initiatives you took in the last year and a half?

I would probably say 20%. One thing you need to buck the trend of as you become a larger company is insular thinking and treating your user base like a wet tea towel that you need to wring out. Ninety percent of our user acquisition is organic, so we just needed to reaccelerate all our core flywheels.

3. Lovable Raising at $4BN and Vercel at $9BN: Justified or Madness?

AI enhanced that. Going really heavy on international enhanced that. We spoke about paying up for the team.

There are companies that are being paid up for, Rory. I’m not shilling, so before you get me for shilling, I’m not shilling, but one of them is Lovable, and it’s in the FT: “New $4 billion round.” By the way, Cliff, notice what I’m about to do here. I’m about to neutralize my argument.

Vercel, another company in a similar space, has a $9 billion round apparently in the works. The question is: do these markups, very rapidly, literally within a month or 2, really make sense, or is it excess capital supply that is exuberant and desperately trying to find a home in an AI company?

Rory O’Driscoll

I’d say it’s a bit of both, but definitely the latter. There’s the FOMO of missing out, and that’s real. People are throwing cash and realizing that we’re on a curve here with this AI boom. Most people are thinking we’re not at the top of the curve. It’s not going to fall off. We’re a lot closer to the top than we probably were maybe a year or 18 months ago, but it still feels like there’s money to be made.

Harry, I heard you the last time saying you’re still investing. This gravy train probably isn’t going to end immediately. It will start cooling off at some point, and I think investors are just realizing they need a good chunk of their portfolio in this category.

I do worry, as I mentioned before, about some of these companies crossing the chasm to the mainstream and turning $100 million in revenue into billions in revenue. But companies like Lovable are definitely well positioned to do that if they keep executing at the rate they are. They’re also creating a new category.

The multiples can’t make any sense, because we knew this when we did the round. Anthropic’s revenue has tripled in 4 months. No question, there was some risk it wouldn’t make the plan. Of course there’s some risk, but it wasn’t that high. We didn’t derisk. We’re all investing on forward multiples.

Lovable was worth $1.8 billion 60 days ago and $4 billion today. I know the ARR growth has been tremendous, but it’s probably exactly as predicted. If either of these had accelerated their plans, it would be one thing, but at tripling at this rate, what’s that?

This is where I’m going to get in super trouble, but fuck it. It’s late at night, I’m in London, and fuck it. They’re out-accelerating the plan. They’re at $125 million to $130 million now, give or take. They plan to end the year at $175 million. They’ll be above that, I think—at $185 million to $200 million.

If you’re at $185 million to $200 million and you end next year—say they do a 2× or 2.5×—say they’re at $450 million to $500 million. Is it that nuts?

I’m not saying it’s nuts, Rory. What I’m saying is, your data is more valid than mine. What I am saying is that most VCs should have had that in the model 47 days ago.

Rory O’Driscoll

I'm not saying that they didn't achieve the progress. It's just, in the public markets, you miss by 1% and you get your head cut off, right? You're down 30–40%, even though the range of variation is quite, quite tiny.

We're always valuing future growth, and I just—it's not that I'm saying if Lovable really, in 30 days, blew out the highest plan any VC had, then I'm with you. But I don't believe Anthropic did. I believe Anthropic set a crazy number, as did OpenAI. These numbers blew our minds when they put them out there, and they hit them or exceeded them. The VCs couldn't put that in their spreadsheet.

I'm going to commend it because I think the Monday.com comment put a pin on that, which I think actually proves the opposite point. Stepping back, because I think it's a really interesting subject: the second round 2 months after the first. I've been thinking about it a lot, right? I'm going to do big buckets and then go down each.

There are only 3 reasons, logically, that this happens. One is that it was priced right 2 months ago, and new information has occurred, or something has changed such that the new price is worth more. That's kind of what we're talking about now, right? Is that happening? If that's happening, that's option 1.

If that's not happening, then the second thing is somebody underpriced the first round and now someone else is figuring that out. There's been some kind of mispricing. Either the second round is too high or the first round's too low, right?

4. Will AI wipe out SaaS margins with 10% GPU taxes? Or is Notion the exception?

And the third, and the most zany one—but I think it's a nontrivial thing—is this validation concept: “Oh my God, Sequoia were willing to do $2 billion. I would never have offered $2 billion or $2.2 billion before, but now I want to get in,” and so they're going to offer $4 billion. You get this kind of last-round-provides-the-validation-for-the-next-round phenomenon. Those are 3 different things, and I think they're all going on to some extent.

There's another point as well. If this company is going to be a $20, $40, $50 billion company, who gives a fuck whether it's $2 billion or $4 billion, right? If you can write the thesis that this company is going to compound some level of growth over the next 5 years and is going to be one of the major players in a new category, then who gives a fuck?

You're right, Cliff. But logically, I'm just going to be that painful person. If that's correct, then the people who did the first round underpriced it. If the company says—just be logical—if the company did exactly what it said it would do and raised money at $2 billion 2 months ago and can raise money at $6 billion now, it should have raised money at $5.5 billion 2 months ago, right? They underpriced the first round.

It's a little like that whole IPO weirdness discussion. “Oh my God, you priced your IPO at $38 and the stock opened at $76. You left money on the table.” It's actually the private version of the same thing.

Rory O’Driscoll

I think it's the Harry effect. I haven't listened to a Harry podcast where he hasn't drummed it up. I think you've added $2 billion of market cap to this company single-handedly.

You should get more shares.

But let's go back to the first one, because I think the first one's interesting. It's actually a fact-based comment: Is there net new information? Are they worth more? You'd like to think the whole world lives in the first area, because if it doesn't, then you're into weirder shit.

The second thing is mispricing, and the third thing is just kind of a psychological, dog-hierarchy, high-school-hierarchy phenomenon.

Jason, your point: You said the Monday.com thing about you miss by 2% and the stock goes down by 30%, right? My comment is that's actually proof why you can, in fact, see these step-ups. If you're underwriting 30% in the next 2 months and you get 35%, by the same logic that if you miss by 5% you go down by 30%, if you achieve or outachieve by 5%, you can justify a higher valuation.

[Speaker?]

I mean, the beta's off the charts, right? Some of the Anthropic thing could go down, as that first example is the performance this year. I'm willing to bet that, no matter how hard you tried, no one had modeled that after they went from $100 million to $1 billion last year, they were going to reaccelerate in Q1 or Q2 of this year, right? And they did, obviously, with Cursor, Claude Code, et cetera.

So there probably is new data. I would argue that's an example of where you have a 3× step-up from the early round, I think, with Lightspeed earlier this year to today. At least some of that is justified based on new information, which is that they have reaccelerated at a scale that probably no one imagined they could do it at, right?

Sometimes that follow-on round 2 or 3 months later might be based on new information. I don't think it's the majority of them.

Rory O’Driscoll

I'll tell you 1, just on this point—one that, to me, shows the inefficiencies in this, or the shoot-from-the-hip-ness. Lovable—Harry's favorite company—closes at $1.8 billion on July 17, 2025.

13 days later, the exact same company, called Replit—that's the one I use—closes at $3 billion. Basically the same ARR, basically the same company. I can tell you my views on security and rogue AI agents, but come on. Most people can't tell the difference.

The revenue is basically the same, $100 million-ish, right? One's worth $3 billion because it's marked up by Andreessen funds; the other's worth $2.2 billion because Accel wants the deal. I don't think either of those deals was perfectly efficient.

Is this fundamentally bad for companies, Rory? If these companies are getting hundreds of millions of dollars shoved down their throats a month or 45 days after they've just taken a couple hundred million more, do you believe that is fundamentally bad for the company?

Rory O’Driscoll

Did it change your views, Cliff, having an extra billion on the balance sheet, or did it not really change the company?

We've had a billion sitting on our balance sheet for ages.

That's a flex, people. “I've had a billion.” What have you guys been doing this week? Okay, flex away, big guy.

You want to sound all wise, owlish, and say, “Don't take too much capital.” But the truth is, it's a rocky journey. There's probably some bumps ahead. I think most of these founders will be happier with a bigger balance sheet, right?

I think the really great ones are the guys who can take the capital and then have the discipline not to use it foolishly. Sometime in the next 2 years, in many of these markets, there will be a shakeout. If you've pissed it all away in performance marketing, shame on you. But if you have that capital ready to move decisively, you might find a good opportunity for it.

It all comes down to confidence in your ability to execute and capture TAM and market share. With Canva, we took a very different approach. We were so bullish on where we were going, we wanted to minimize dilution.

So we raised as little as possible at every stage. This is a high-risk maneuver, and I don't recommend this to founders anymore. I say be a bit overcapitalized, but we would run it kind of to the bone in order to take as little money as possible.

We'd go for the highest valuation possible just to back ourselves to hit that next level and get as minimal dilution as possible, which worked out well for us but was a riskier-than-probably-recommended maneuver.

5. Quarterly Results for Snowflake, Mongo, Okta, Zoom Skyrocket: Is B2B SaaS back from the dead?

Guys, I want to cross the chasm, so to speak, and move from the world of privates to the world of publics. This will be a fun one, because we had quite a big week in publics—crushed for B2B. Jason, baby, B2B public markets is back.

Snowflake, Box, Elastic, Okta, Zoom. I mean, Zoom beat. That's like Madonna coming back from the dead. I mean, that is like—

Sorry, Rory. Don't worry, I said it, not you. Jason, is this just the return of the good old days for SaaS, baby? How did you analyze uniformly great results from everyone?

[Speaker?]

I don't know if it was quite uniform, but it is interesting that some folks—I mean, let's spitball it as half of the public B2B leaders are finally getting an AI tailwind, right? Or they finally are getting one. Not everybody. We love Salesforce; we had Marc on. They haven't seen it yet. They have the demand, but it hasn't hit yet.

Box, Zoom, and MongoDB should be crushing it, because every time I spin up a new vibe app, I need 2 or 3 databases, right? That's just one corner of the world, but MongoDB should be crushing it. It's exciting to see.

The Atlassians from down under, the Dropboxes, and the Asanas aren't seeing it yet, but it should. Like Cliff said, you've just got to be smart. We're not building our own LLMs. If you have a billion-dollar installed base, you have a distribution channel, to Cliff's point.

It's kind of sinful if you haven't reaccelerated by the end of 2025. You kind of failed as a founder because, yeah, you may miss some of the cool kids, right? They may not be using you, but you have a billion-plus of distribution. You have no excuse. You had 18 months.

So thank God we're seeing it, right? It would be almost catastrophic if none of the leaders were getting an AI tailwind boost. It's good, but it's not a dead-cat bounce. Outside of Box and Snowflake, we're seeing modest, modest reacceleration.

It's great to see them have it. Otherwise, we have to give up on all the public guys and bet on the Canva, Databricks, and Anthropic, and give up on the last generation.

And again, yeah, they're taking shit, because that was the one that jumped 40–45% in the stock price.

And it's back to the point Jason made earlier: the year-on-year GAAP growth rate is back up to 24%. I think it's higher quarter-on-quarter, so I think they guided a little more aggressively going forward, but they were at that rate 2 years ago, right? So it's not like they're 10x-ing or something like that.

I think what happened here is everyone got into the, “Oh my God, SaaS is dead. Everyone's sad. None of these guys are going to make it.” And this is back to the Monday.com comment: these markets are trying to tell you it's not just about fundamentals; it's about how you perform relative to expectations. If your expectations are low and you just do moderately well, you can have a 45% jump in the stock price in a week, right?

And if you look at the absolute stock price, if you look at the revenue multiple, it's just back to where it was 2 years ago. It's a great core company. It's like the Mark Twain reports of my death were greatly exaggerated. Well, it turns out reports of the death of SaaS and software were greatly exaggerated, right?

If you're a good CEO—and Dev Ittycheria is an extraordinarily good CEO—just like the Cliff story there, I noticed, Cliff, he didn't say, “Oh, it was all just AI.” You got your shit together. You did a whole bunch of things. You raised your expectations. You said, “Hey, we're the leader in a big market. Let's make stuff happen.” And if at that point you're valued at 7x and then you beat plan even by a little bit, you get that kind of bounce. That's what happened here.

Cliff, when you see this and Monday.com getting hit for being a couple of percentage points off again, I would never ask about timing or anything quite that ludicrous, but do you go, “Yeah, that's an arena I want to be in”? Or do you sit and watch Cheeky Pint with the Collisons and go, “That's the arena I want to be in”—sitting drinking a non-alcoholic beer with the founder of Cognition, doing a handstand with Vlad, enjoying the wonderful splendor of the private market?

I still love my beers alcoholic. I haven't followed that trend, but I love it.

There's a lot less scrutiny as a private company, but as a late-stage private company, with the likes of all the big cats that are playing in public markets already invested in us and continuing to do so, our reporting obligations and our expectations to beat and raise are pretty much the same. So it does get me thinking, like, what is the real difference?

And then I think to your point that you've made on previous podcasts, the public markets are valuing companies a lot higher. So when the public markets were valuing companies lower than the private markets, you were kind of like, “Well, whatever, whenever.” But now, at a lot higher marks, it is appealing. It is becoming more appealing to go public.

In the show, we actually mentioned you—you probably heard it, sorry—where we were like, “Figma goes out, sees the pop, and if I were you, I'd be going back to the team saying, ‘Let's Forrest Gump this one. We should go out now. Let's run for it.’”

We're gearing up to be ready to IPO. We want to be an IPO-ready company. We recently—you mentioned Zoom—brought in Kelly Steckelberg, who led their IPO, and she's been a fantastic addition to the team. She was their CFO.

So our goal is to be ready. When we actually go out is another question, but we're gearing up to be an IPO-ready company.

Can I ask a question we've talked about on this show a bit? You have $1 billion in cash. You're profitable, or cash-flow positive—I don't care which one; probably both.

Nice.

You're able to do tender offers for your employees and provide liquidity. And for whoever of your early-stage investors wants out, you can probably flip their shares. Why?

I'd love them to sell more shares because it has helped solve my problem right now of not having enough sell-side.

Yeah. So why, at a meta level, IPO? You have access to capital, and even M&A probably isn't a reason on its own, unless you want to buy something for $10 billion. Why would you IPO?

That's the question we've always asked ourselves, and I think there are 3 key points. There's availability to capital, which we have access to. I think it's probably liquidity, and there are restrictions, particularly around employee liquidity and what you can do in the US and whatnot around that piece.

We do believe that, at 13 years old as a company, our employees should have liquidity. They've created all this value. How can we make it easy for them to access the wealth that they've built up? While annual secondaries are a mechanism for that, it's pretty janky, and particularly in some jurisdictions, it's downright impossible. So that's probably the biggest one.

You also get a bit more publicity. Personally, we don't want to be more in the public eye. We're happy just being in Australia, working away, building great products.

Guest 2

I'm not going to be too nice to you, Cliff, because after last week, Harry gave me grief for being too nice to our guest, Mr. Benny.

Give it to me. I love it.

Guest 2

Yeah, but I'm actually going to be nice this time because I totally agree. And you mentioned the other one in passing, and I just want to put it back on the table because you said it: “Oh, and by the way, the public markets now are giving me cheaper capital than the private markets.” If all the numbers are as reported, you're getting roughly 10x revenue, and the fine folks at Figma are getting between 17x and 30x, depending on how you value the current price, right?

I think that's a byproduct of these large crossover funds. They probably have 80% of their capacity allocated to public markets and 10% to 20% to private. So you're chasing a smaller pool of capital, even though we're in a good position. Ultimately, the volume of capital dictates that multiple, and there's such an immense amount of capital being deployed in public markets that it just is driving up those valuations.

Guest 2

In this conversation, which we've been having all week, I'm a huge believer that companies at scale like yours should be public, right? If for no other reason, it is bizarre that we've evolved the system whereby, to allow ordinary people to invest in you, instead of paying 50 bps to Fidelity, we have to pay 2 and 20 now and enrich the middleman like us.

God bless it, but it doesn't seem like a mission-driven company would make that their mission. Call me cynical on that, right? The whole structure is absurd—these high fees—and it's exactly what you said: having to get permission from your employer to get liquidity as a secondary after 13 years. It's better than no liquidity, but it's a little bit serf-like, and when you're public, you can make your own choices.

So I'm totally with that answer. In terms of companies at scale, when they're ready, they should go public, and it feels like the better way to run a business at scale.

Yeah. People deserve liquidity, and with our customer base—we've got 240 million monthly active users—a lot of them want to invest in Canva. You saw Figma had huge retail demand. We want people that have helped create our success to share in that success, and we really want to deliver for them.

So it very much works into our mental model of the world. We're not anti.

I love it. Are you not the perfect contender for a direct listing? I've looked into this, and while I've looked into it in depth, it just—you've got a great consumer brand. You've got 240 million consumers that would love to buy in. Direct listing all the way, baby.

Yeah, it is an option. I'm not sure it's going to be the option for us. We'll look at all options when the time comes. You can still get all those dynamics. You're not really getting much out of it.

And if you look at all the historic direct listings and how they've gone over time, I believe the data proves that none of them have been greatly successful. A lot of them have been really successful companies over a long period of time, but in the period post-direct listing, none of them have really nailed it in the short term.

You are right in one sense, but it's always worth pointing out that the definition of success is weird, because you're right: they didn't nail it in the short term. In other words, the stock didn't go up a lot after the direct listing, but a little part of me wants to say that's the freaking point, right? If you get the direct listing totally successful—

The people buying don't make any money, right? And people like their pop.

Guest 2

You've mentioned this in a prior podcast. It is about getting the right long-term investment. You want people that are going to hold your stock if you deliver. If you deliver—being a key point—you need to deliver for 5 or 10 years and compound that position.

So, yes, from a sort of logic perspective, it makes sense: supply and demand match that. You definitely don't want a huge pop, and I think you can sort of manage that through how you stagger the lock-up periods, et cetera. There are ways to manage that so it isn't everyone locked up for 6 months and then it drops. And there are better ways to manage that.

But, yeah, I think you're ultimately optimizing for the large, long-term shareholders that are probably going to hold 50% of your stock for an enduring period of time.

Rory O’Driscoll

And the relationship you build with those investors is incredibly important, and they’re, from what I understand, pretty anti-direct listing. And that answer—that very cogent answer—is, Harry, why everyone in theory will argue this. Then, when you’re the guy on point with your life’s work on the line, just like Halligan, you’re like, “Do I want to be an experimental baby on the biggest day of my life, or do I just want to land this freaking plane?” There you go, baby.

Well, I think your point on Figma was totally right there, Rory, which is that everyone was like, “Oh, well, here’s the test case for why we need to have a direct listing.” And you were like, “Well, had it been a direct listing, it would not have listed anywhere near the price that it went to. It would have been 3 to 4 bucks higher, maybe at $36 to $40, but it would be ridiculous to assume it would have been a $75 starting price.” I think that was really well articulated.

6. Is Jensen Huang right there will be $4TRN in AI gains?

I was listening to Jensen on an earnings call. This is what’s so great, Cliff, about doing these shows: I actually have to do some work and really listen to earnings calls again. He said, “Over the next 5 years, we’re going to scale into, with Blackwell and with Reuben, effectively a $3 trillion to $4 trillion AI infrastructure opportunity.” Can that level of capex be supported by enough AI-driven revenues, guys?

Rory O’Driscoll

Yeah, I mean, look at it: $4 trillion. You want a 20% return on equity, you’ve got to be generating $800 billion of profit. That’s a lot of profit when Meta and all these guys make a couple hundred billion a year, right? So you’ve got to believe you’re going to create another 4 Microsofts, another 4 Facebooks, to justify that kind of spend.

It feels deeply lofty to me and not grounded in the macro. On the other hand, it’s hard to argue against a guy who built the most valuable company on the planet, so you can give him credit for the specifics. I don’t see where the macro works. But whatever. If you want to make that bet, Harry, there are Nvidia puts that I keep my eye on that you’re more than welcome to plow into anytime you want.

Tell me when you do, Rory.

Rory O’Driscoll

The hell I will.

Jason—

[Speaker?]

All I do know is, I know it’s a small percent of the economy, but when you listen to what Cliff’s saying now, when you listen to what Marc Benioff said last week, basically Marc said, “We’re like 0.1% AI-penetrated in the Salesforce base,” right? Salesforce is coming up on $50 billion, right? They alone are going to have $200 billion of AI attached to their model.

I’m not saying Marc’s going to get all of it, but the attach is going to happen. It’s just so early that it’s hard not to see everything easily being 100x bigger than it is today. We just started. We just started. It feels like 100x now. Does 100x get us to that number? I don’t know. But I do think that Jensen and Sam Altman have a pretty good sense of it, so I’m not betting against it. We can ask Cliff how deeply AI is penetrated there.

Yeah, we have billions of AI usages in our product per month, and that’s accelerating. It is just beginning, and the amount of calls and inference we’re going to rely on is just going to grow exponentially as these products evolve.

It’s easy to see 100x growth, right?

But they’re all getting distilled, and they’re all going to get run on-device a lot more. So there are optimizations that are coming as well.

Thank you. Because you’d mentioned the Notion comment about spending 10% of the revenue on AI—in GPUs, inference and model training, right? I mean, turning back, they’ve gone from a 90% gross margin to an 80% gross margin, which is effectively a way of saying that, to deliver their AI magic, they have to part with roughly 10% of their revenue to the big AI companies, just as they probably did roughly the same with AWS.

So, turning back to Cliff, do you envisage spending 10% of $4 billion—$400 million—on Nvidia chips, third-party models and GPU acceleration? Does that feel way too much?

100% yes. 100%.

Yeah.

Wow.

But especially in the short term, it will probably be less than that over time. You’ve got to separate training your own models from serving AI. Currently, yes—I mean, if you look at Lovable, what is their pass-through in regard to what they’re paying Anthropic, or whoever the model providers are? It will be a lot. It’ll be way more than 10%.

But over time, they’re betting on distilling these models down, understanding user queries and where I need the foremost, frontier, best model versus where I can deploy the model that’s on-device or the model that we’re self-hosting and running. Companies will get a lot better at picking the right model for the right job and only using the expensive models connected through an API to OpenAI or Anthropic, or whoever, for the most premium queries where you need that answer.

Ninety percent of it will be run on-device or be self-hosted. Over time, we’ll use the best models. So take images, for example: if there’s the latest and greatest image model that has additional capabilities, it may cost us 4 cents an image, but we know we can get that cost down to 0.02 cents an image, and we’re banking on that over a 6-month period. We view some of those upfront costs that are eating a chunk into our margin as more of a marketing cost than a long-term, enduring cost of goods.

Got it.

Rory O’Driscoll

And that’s a huge difference. The assumption of getting 10% from every software vendor is crucial to the idea that you can spend $3 trillion. If Cliff and all the other Cliffs optimize and that 10% becomes 5%—

I mean, which is still a hefty tax for you to pay.

Rory O’Driscoll

Yeah, but I don’t want to speak for Cliff. I imagine creating a static image as it is today, you could bring it down an order of magnitude, but—

When Canva adds everything that Gamma does, Gamma’s consuming a lot of tokens to build dynamic presentations for every single person on my little team on the fly. This is not an image, right? They’re coding every presentation from scratch, right? That’s a lot, and it’s only pretty good. Imagine when it’s great, and they redo every presentation 3 times and run it through multiple models, and then Canva does it. Canva has a higher bar because you have 240 million users.

Yeah, but we don’t need to code it, right? That’s why we’re building our own foundational model to generate a presentation that’s phenomenal. It doesn’t need to code every line of a presentation. There’s a heavy compute cost to create a presentation. We don’t need to essentially go to Anthropic and write a whole website every time we want to create a presentation. There are a lot easier ways to create presentations at a lot lower cost, so we’ll be thinking about that just like we’ve thought about it.

But if you were doing the Gamma approach, which you’re not, you might, to Rory’s point—let’s compare.

We’ve got Canva Code, which you can use to code a presentation; you can code a website. That is a high compute cost. We’ve got the equivalent of Lovable. It’s more for creating widgets and for education purposes, et cetera. It’s got 20 million active users already. It’s going really, really well.

But you could roll that up. It is cool, and I use it today, but you could do much more. You have it pretty locked down. What you can do with it is create assets and overviews. It’s great, right? But you could spend a month and this could be Lovable Prime if you wanted it to be, right? You’d use 100 times the tokens.

Totally. Totally. Yeah. That is an expensive product to serve our customers. That is the most expensive product to serve our customers.

Rory O’Driscoll

The macro—this sounds really arcane, but it’s actually going to drive a huge amount of downstream implications for the whole discussion we’re having on AI, right? Big picture, the cloud business pre-AI—AWS, Microsoft Azure and Google—is plus or minus $150 billion to $200 billion of total revenue, right?

If every software company spends as much on AI inference and AI training and the whole enchilada as they did on cloud compute, that’s a $200 billion-a-year business, just as I said—rough analysis. There’s probably some double counting there, right? That’s pretty damn impressive. It’s still going to be hard. That’s the minimum they need: $200 billion of top-line revenue and $100 billion of profits.

Going back to that, you wouldn’t want to spend $4 trillion to make $100 billion of profits, is that so? I think it gets back to this: for $3 trillion of capex to have a return, people like Cliff running businesses like that are going to have to yield a lot of cost over to the hyperscalers and the model providers. I don’t think a large number of software executives are going to do that, at least easily, so it will be interesting to see how that math actually shapes out and whether you can, in fact, command a return on that level.

My gut is that something over and above what we’ve seen now is required. Jason, you’re pushing, and you always do. Maybe that is there. Maybe it is: instead of optimizing compute, you throw compute at everything and end users are willing to pay for it. But it’s something more than what we’ve got now.

I just think we’re underestimating—I mean, Cliff made the point—we’re underestimating the processes we run today. We’ll figure out how to use fewer tokens, our own models or other things. If the world doesn’t change, it will come down an order of magnitude a year, or possibly faster. But our ability to use orders of magnitude more tokens—Canva could turn this on tomorrow, and we could consume massive amounts of tokens.

They already have the product. It’s already cool. It’s just here. We genuinely have to think about this deeply because we’ve got 240 million users.

We’re about to launch in October a whole slew of new AI products, deeply integrating it into every part of the workflow. And we need to seriously run the math: if 20%, 50%, or 80% of our users use this 10 times a month, what are the costs going to be? They can look pretty big and eat into your margins very significantly, so we need to be sort of double-checking.

I don’t feel comfortable shifting pricing accordingly.

No, so we are doing that, actually. We’re moving to a unified credit model around AI. Your free subscription gets a certain amount, a premium subscription gets you a certain amount, and then, if you’re a super-active user, that means it can’t eat into our margins too much. You need to maintain that margin, and we’ll have that scaled, usage-based pricing beyond a certain point.

When you look at usage within the company itself, Jason said something I think was very apt a couple of shows ago. Jason, you can remind me specifically what you said, but you said about basically equipping developers with—I can’t remember the number—was it $10,000 a month in terms of assistance through coding tools?

[Speaker?]

That’s where Farhan at Shopify was pushing it, up to $10,000 a month. If you can prove the ROI, that’s the budget.

When you think of equipping your engineers at Canva today, would you feel comfortable in a future world equipping them with $10,000 a month of coding agents?

We haven’t done that down to an individual level because I don’t believe that, with over 2,000 engineers, doing that at scale would be the right approach. But from an engineering leadership perspective, we encourage all our engineers to use the best security-certified coding tools that can increase their efficiency, and we’re not price-sensitive around that at all.

We know that ultimately the playing field levels out and there will be competition. We’re very open to whatever tools they want to use, but there are always 2 or 3 great ones, and it’s already consolidating.

But to push on that, because you probably have got the $20-a-month level as a given, the $200-a-month level—do you see, because Jason’s visualizing a world where you can go 1 order of magnitude beyond that, to not the $200-a-month but the $2,000-a-month level and even beyond that?

Yeah, we need to rethink our seat-based pricing model because some of the tools, particularly around marketing tools we’re creating, enable a single marketer to deploy tens of thousands of pieces of content. One person can create so much content, feed that into all the social platforms and wherever their marketing visual content ends up, get feedback on how that’s performing in the wild, and feed it back into the creation loop.

One person can do inordinate amounts of work, and that’s using a huge amount of compute. You can’t charge $20 a seat for that level of breadth. You give that functionality for a per-seat price, but beyond that it needs to be based on consumption. So it’s a hybrid seat- and consumption-based model.

We saw Monday get hit because a lot of their growth relies on SEO. You’re seeing SEO really reduce as a customer acquisition channel for a lot of companies. I know you’ve only got 10%, which is paid, Cliff, but given 10% being paid, are you moving forward with the assumption that SEO is going to be a much smaller part of your customer acquisition funnel moving forward?

I think SEO is about 15% of the 90% organic to date. It used to be our number 1 channel, but now our user flywheel, word of mouth, and people sharing designs are our biggest.

We’re seeing a lot of it. SEO is growing for us. But also, we’re the number 1 productivity app on ChatGPT, and we’re the 5th-highest domain that ChatGPT refers to. Out of all the websites, it’s like Google, Meta, blah. We’re number 5 in regard to essentially SEO for LLMs.

Because we’ve invested a lot over the years, they’re obviously taking a lot of the same signals that Google’s taking, and anything we’re losing on the SEO front is translating to LLM SEO, which is a huge tailwind for us. To give you an example, a year and a half ago, 0.02%, I think it was, of the images uploaded to Canva were from ChatGPT. That’s now over 5%.

The volume of content being generated in these LLMs is being propagated into Canva for editing, for how they’re using it in designs, and for that collaboration, storage, and deployment—that whole visual communication workflow that we excel at. I know what Harry’s fishing for: he’s trying to figure out how his investments are doing.

Have you proactively tried to win in terms of how you show up on ChatGPT in the same way you did on SEO, or has it just happened organically, just by virtue of being who you are?

I would say 100% we have won. As soon as these LLMs started taking off, we had the conversation: is our SEO team working on LLM optimization? There’s definitely a team at Canva working on that.

Yeah, this whole thing about SEO being dead is stupid. It’s dead for folks that don’t have a brand, don’t add value, and don’t have reach. I just popped it into Claude: “Best overall Canva. What’s the best design product to make a YouTube thumbnail?” Best overall: Canva. It’s just—

In fairness, I think you have to distinguish between people like Canva, where they have a product to sell and they’re totally happy to sell it via ChatGPT, and media companies, where the only product they have is their content. If Google or ChatGPT serves up the answer and no one clicks on the website, then they’re toast.

So, I agree. I think for Canva, this is not existential. You guys are fine. But if you’re a mid-tier review site and you just get scraped and summarized, well, thanks for playing.

Cliff, if I were to ask you, you’ve got OpenAI at $500 billion, Anthropic at $183 billion, and xAI at $100 billion. Where do you put your money?

All of them.

You can choose 1.

I’m a big fan of all those companies. I’m not going to choose.

We’re still a constitutional country. You have the right to remain silent.

Yeah. Yeah. Yeah. Well, yeah. They’re all doing great work.

Right.

Rory O’Driscoll

Neutral.

Guest 3

He doesn’t fall for your traps, Harry. Not like me, who foolishly feels the need to answer these questions and get into trouble.

No, but I genuinely believe they’re all companies that are going to be the foundations of our AI future, that are going to feed pretty much every single product. So it’s like betting on Amazon and betting on power. The power question is an interesting one, because every big revolution when it comes to a technology shift has largely been power-based.

I think one thing that’s interesting is the amount of energy that Jensen’s $4 trillion investment is going to take is just insane. It’s akin to what Tesla has done with electric cars. While I see AI pessimists and environmental pessimists saying, “Oh, so much more energy—it’s going to be bad for the environment,” I actually think it’s going to rapidly accelerate our shift to green energy, particularly nuclear, which we’re just going to have to solve.

Once we’ve solved it, and it’s way more economically viable than burning fossil fuels, it’s going to kickstart the entire shift to renewables or zero-emission energy sources, which I think is ultimately going to be huge for the environment long-term—like, midterm.

Before we wrap, there’s 1 interesting topic that we talked about before, which is, you said, “Oh, you should invest in Riverside.” And I was like, “Oh, no. I saw it at seed and I missed it, and then I’ve seen it every round since, and I didn’t want to do it.” You said it was an interesting thing about the VC regret pathway and not engaging later on.

I had it again today with Revolut, when people asked me why I wasn’t investing in Revolut, and I was like, “Well, it’s a bit embarrassing as an early-stage investor to buy Revolut off Goldman Sachs—my bank.” That’s when you really fuck up as an early-stage investor.

Rory, Jason, I’m just intrigued to hear your thoughts on the ones that you’ve missed and the regret pathway on investing later.

Rory O’Driscoll

I think you should do it. That’s the short answer.

And for the viewers, this happened before the conversation went live. Cliff was talking about folks who looked at Canva early on, passed, and then really struggled later on to pony up and pay obviously much higher prices.

Guest 3

I’m the exact opposite, right? Many of my most successful deals I’ve passed on prior, and I’ve just learned: if you pass on something, you come to a conclusion, and then you get another data point a year or 2 later and they’ve done what they said they’d do, you literally don’t need any more information.

It’s like you said, because it’s so much more telling. With a new deal, you’re starting off and all you’re seeing is 1 data point. The difference in information content between 2 data points over time, both of which are positive, and 1 data point where you have no calibration is almost infinite.

I can think of 2 or 3 deals way back in the day. I passed—I didn’t get Amazon in 2003, and a year later I saw it at twice the price and I bought all I could.

Rory O’Driscoll

Same thing on Box. I had passed at the start of 2010 and didn’t even get it done. Nine months later, I literally woke up and said, “What’s the dumbest thing I did all year? I didn’t do that deal.” And I went down and did it.

I think I’m trying to discipline myself to do it even more because, when you go back—let me just repeat it again—when you see the company, when they say they’ll do A, B, and C, and you pass because you don’t believe they’ll do A, B, and C, and then they do A and B, and even if they do C-prime, a little less than C, you have what you need to know. Then Cliff’s comment applies: now you’re seeing a category leader. You can lean into their execution.

You really should say to yourself, unless you think there’s a time problem—unless there’s a time problem—you should say to yourself, “I was wrong. How do I change my weighting and lean in here?”

Rory, does that leaning in apply in an AI world where sustainability of revenue is a question? Because a lot of people will say, “Oh, I’m going to do $10 million in a year.”

Rory O’Driscoll

I think they’re 2 separate issues, because you are right about one thing: sustainability is a lot harder in AI. We’re seeing a lot of people drift in and out of product-market fit. But that’s going to be true in the new deals as well, right? The new deal that you see where you have no context from 2 years ago and it looks golden today—that can drift out of product-market fit too. So it’s a separate factor.

I do think, even in an AI world—taking it one step beyond—I think the really positive sign in an AI world would be that you find the founder that, for whatever reason, you passed on 2 years ago. The best of all signs is this: the product has evolved 3 times, because that’s what’s happening in AI land, and the founder’s been able to evolve it.

Then you’re like, “Oh my God, this guy has a survivor gene. Run, don’t walk.” Because I think that’s one of the identifying characteristics of the people we see figuring it out: the damn thing keeps changing, but they just keep changing faster than the other guy.

Again, I think there’s always signal, because the hardest thing—the thing you can’t change in this business—is time. You can’t compress time. You can’t fast-forward. You can’t rewind. So when you have 2 data points over time, that’s just so freaking powerful, right?

I totally get it because I wrestle with this. You get hung up: “Oh my God, I could have done it for $10 million or $100 million or whatever. Now I’ve got to pay $500 million.” You just have to look yourself in the mirror and say, “That is the tax you pay for being stupid. Pay the tax and just get off the stupid train.”

Does the same apply for follow-on rounds as well? It’s amazing to me when I’ve seen investors have the inside lane on all the company data and the company’s performing like crazy. They got a big chunk very early—call it Seed, A, or B.

The investors that have done best out of Canva were early-stage funds, but they realized, “Holy shit, we’re onto something here.” So they raised SPVs or additional vehicles to move further up the value chain, going later and later stage, and compounded their position, or at least didn’t get diluted over time. They’ve done the best.

There are a lot of early-stage investors who call themselves disciplined: “Oh, we only stick here.” But if you’re on a winner, keep betting on that winner, is my approach. It’s amazing to see how differently investors treat follow-on investments as well.

Rory O’Driscoll

It’s a great point, and I think you have 2 separate issues. You have the individual investment: are you making the right investment? In other words, did you really think that the third follow-on round was overpriced because you thought the market was smaller? Were you wrong on the investment? That’s one factor, and we can talk about that.

Separate from that, you have the institutional thing. Are you set up to do those big rounds? Do you have to raise an SPV? Are you able to raise an SPV?

I would say, Cliff, one of the things I’ve internalized is that, especially being scarred by probably 3 decades, it doesn’t help sometimes. You’re right: in a company like yours, the correct response is to pile in at every level and find some way to do it.

There are 2 separate things. One is, do you still think it’s a good deal? To your point, one of the things we’ve observed is that the round after the round we do, if it gets a quick outside-led up-round and you have positive data, it always feels expensive. You go back to, “Oh my God, it’s 2X more.”

That’s the round you should do every time, because it’s roughly in the same strike zone as your sweet spot. It’s not like we typically invest at, plus or minus, $100 million pre. The $20 billion round is hard to get your head around. But if you do the round at $100 million, and then 12 months later they’re at $300 million or $400 million and everything’s working, that’s a signal that we have constantly underestimated, constantly corrected, and been validated.

Or actually, there’s been a price inflection point, but there hasn’t been a company inflection, and so you’re actually paying up for little company growth. I’d rather pay up for the $800 million to $1 billion where there’s real company inflection and the price inflection matches that.

Rory O’Driscoll

If you believe your inside information points to the negative and you know something, then yes. But I give Peter Fenton the credit for saying that the outside-led up-round—the follow-on in the outside-led up-round—was the strongest positive signal, and they consistently underestimated the value of that. I do believe that, to your point, Cliff, is the case.

I do admit sometimes that, if your business is relatively early-stage, it is hard to think: how do you go at $20 billion, what do you do, and how much do you put into it? But there’s no doubt—we’ve talked about this before—that being willing to massively concentrate on a small number of deals gets you the last absolute dollar of outperformance.

You do have to, as you say, though, be able to distinguish Canva from the 10 other companies you’ve had that got to billion-dollar pre-valuations in 2021 and aren’t worth a billion.

True. In the end, you still have to be vaguely good at picking, unfortunately.

Guys, listen. I can’t thank you enough for this. You’ve been fantastic. Cliff, I so appreciate you joining so early in the morning. It’s so great of you to join, and I really appreciate it, man.

Thank you so much for having us. I appreciate it. It was a great chat.

Rock and roll. Thank you, Cliff.