Are Burn Multiples BS in an AI World? & Sam Altman Needs $1TRN of Energy

Jason Lemkin

There are only 2 ways of pricing a deal. You price a deal on hope, or you price a deal on the multiples. A $15 million-revenue company that's perfectly good and has reasonable growth is actually of zero value to a VC because we're in the upside-option game.

I hear too many folks saying, “Oh, you're triple-triple-double-double or better. You're golden. Don't worry, kids.” I think that's terrible advice in 2025. Terrible advice.

Jason Lemkin

Whatever the prize is for being the best company in AI, OpenAI is going to get that prize. Have a great day.

Jason, I want to start with one that you just suggested, which was fantastic. I want you to explain it a little bit. ICONIQ did a report, and there was a really interesting takeaway for you. Can you explain what that takeaway was and how we should think about it?

Jason Lemkin

Yeah, they did a 73-page State of Software report. We don't all have the patience for more than 80 charts. I wrote it up on SaaStr this week—my top learnings—but the one that kind of hung with me, along with the triple-triple-double-double discussions we've had here, which you had from Hemnet or from General Catalyst, right, is: What's the point of venture in 2025? Rory's other point is that 70% of the money is going into fewer than 20 deals.

But this one analysis that ICONIQ did was interesting: AI-native companies, the ones we're all so excited about, growing so quickly under $100 million ARR, have terrible free-cash-flow margins—–126%. Non-AI companies are –56%. They're still burning, right? Classic tech. But because they're growing so quickly, the burn multiple is actually much lower.

They're actually capital-efficient because they're growing so damn quickly that, even if they're burning a lot along the way—even if they've got afterburners on and they're getting to Mach 10, or whatever it was in Top Gun: Maverick, so quickly—if the capital efficiency is better, this is where VC should be putting all their money, right?

Can I interject and ask for an explanation for those who don't understand why the burn multiple is better if they're spending more? Can we explain it for those who don't understand?

Jason Lemkin

It's basically: How many dollars of ARR do you get out of each dollar that you're spending? What is the efficiency you're creating for each dollar of venture capital you're lighting on fire?

It's a weird metric because you could seemingly be efficient and run out of money, right? If you don't have enough money in the bank account, it doesn't mean you're profitable. David Sacks sort of coined it, and I think when everything was the same in SaaS and B2B in 2021, it made a lot of sense. All the companies were the same. They all kind of grew the same.

As you have companies with lower gross margins, at first they kind of broke it. Then VCs stopped wanting to fund everything in 2022; that broke it. And then AI breaks it because we've never seen growth like this, but the margins are lower. A lot of them have token costs, right? Some don't. We've talked about folks like Higgsfield, which is somehow almost cash-flow-positive at $50 million, or Lovable and Replit, which are burning a few million dollars of venture capital.

How much did Lovable raise in the last round, Harry?

$200 million.

Jason Lemkin

$200 million. Okay. But the point is: Oh my God, let's imagine you throw $200 million into Lovable and they're burning $6 million a month. That would shock most VCs, right? But if they're going to add $300 million of ARR, actually, the burn multiple is quite low. It's quite efficient from a company ARR-building perspective.

That's the thought. And then you should put even more of your money into these companies, right? Because you get the most leverage. The theory is you get the most leverage out of your venture dollar in these types of companies, right? Because you're getting the most ARR per dollar invested.

I think there's an old expression about models that's also probably true about ratios and rules of thumb: All models are wrong, but some models are useful, right? In the same way, all ratios are wrong, but some ratios are, at times, useful.

I think the burn multiple is a very useful ratio, but as Jason said really well, there's a whole bunch of implied assumptions that go into that, not all of which are true, right? So if you use it blindly, you're going to blow up. I thought it was a brilliant insight when David coined it, and it really is helpful to compare different companies at different stages and get a sense of capital efficiency, right?

But there are about 3 or 4 different assumptions in it that, if you forget them and just focus on the burn multiple, you're going to blow up. It's worth disaggregating that.

Jason Lemkin

First of all, let's say what it is. Burn multiple basically says it's a ratio between the amount of ARR you add and the amount of money you spend to add it. If you spend $2 in total burn for every $1 of ARR, your burn multiple is 2, right?

If you think about it—a big, crude comment here—if you're getting paid 10 times ARR, if you're being valued at 10 times ARR, you just spent $2 and added $1 of ARR that's worth $10. You're up, right? You put in $2 and you got $10 of market cap, so it is a very valid construct. At that level, it totally works.

Obviously, this is one of those weird multiples where lower is better, right? One is better than 2. And there's some idiosyncratic stuff: When it crosses into positive, it goes into a negative number, which makes your head hurt. We track it a lot internally, so we wrestle with all those things.

At a high level, it's a super-good way of comparing things. You can compare companies at very different levels, with different growth rates or even just different absolute sizes. It's a super-good insight.

Embedded in it, though, is a whole bunch of assumptions. There's a whole boring economist phrase here: ceteris paribus, everything else being equal. The implied assumptions are that ARR is real, but we know it's often not. The implied assumption is that it's net ARR, so you're taking out churn. But if you're growing fast, you can hide churn because you're churning customers from 12 months ago—which is maybe $2 million—and this year you're at $10 million, so churn is understated. You can hide that. So, is your ARR real? Is your churn real?

It does pick up on gross margin, but again, the same thing: If you're growing hyper-fast and your margins are moving, you're not picking up on it. And then, lastly, probably less so for these companies but definitely true for the AI model companies, it doesn't take capex into account, which isn't true for the $100 million ARR companies but is definitely true when it comes to capex. You can't ignore, even in a crazy world, $10 billion of capex in a company.

All those things mean that, when you're comparing, the model isn't like-for-like. That's one big-picture comment: There's a whole bunch of implied assumptions in there, which is why, even though we love those kinds of metrics—and the one metric you mentioned when we were chatting before, the Magic Number, which is about sales and marketing, we actually coined it back in 2004—they're all good assumptions.

But we've actually come back to saying there's a real advantage in seeing the GAAP revenue accounting as well, to make sure all the money is, for lack of a better word, just showing up for real. So that's one. There's a lot of noise in that multiple, and I think when they were all SaaS recurring-revenue businesses, all seat-based, all 80%–90% gross margin with no capex, all enterprise sales with low churn, it absolutely made sense. You could compare 2 companies.

That's why, by 2019 or 2020, it almost felt like, “Fill in the form, give me the valuation.” None of those conditions are true now, right? So, yeah, I agree. I think it's totally up for grabs.

Does it, as a framework, carry no weight then, given the volatility of all the different inputs, which means the output is less reliable? Respectfully, is it even a reliable framework to look back on?

Jason Lemkin

It's a decent framework. We still use it. It's absolutely a decent framework because there are things you can do to get the same idea. You can look at, for example, delta GAAP. In other words, you can see how the GAAP revenue changed.

If you were doing $2 million at the start of the year in ARR and $10 million at the end of the year, you know, the delta ARR is $8 million, right? But another way to get to the same thing, to check for, quote, honesty, is to look at whether you're recognizing a $2 million GAAP run rate in January and a $10 million GAAP run rate in December. It kind of checks on that.

There are things you can do to deal with that. But there's a lot more noise in the system, right? You have to worry about churn, and there are other issues—even beyond that—that we can come to. But even just at the churn level, are you picking up all the metrics? Is it forward-looking enough, especially on these trials?

Stepping back, there's an implied assumption. Remember, going back to my simple model, I was spending $2 million, getting $1 million of ARR, being valued at 10 times, and therefore creating $10 million of value, right? If that ARR evaporates a year later, then I didn't create value.

Rory O'Driscoll

So, there are implied assumptions around stickiness. All of those things are up for grabs. I think it is useful—I’m sorry, it was a long way to answer. I think they showed it is still useful, but we’re far beyond the stage that we were in 2019, where you could just plug the numbers into the number-cruncher and come up with a rough and pretty accurate estimate of the valuation of a company. We are not at that stage anymore. We are not in Kansas.

Jason Lemkin

There’s a meta-question that I think about mostly when times are good, when companies aren’t running out of money. Maybe it’s a question to Rory and to Harry, but venture in some ways is an ARR arbitrage, going to your point, right? When you’re north of 10× revenue, venture works. You put in this small amount of money, and the magic thing is—we can talk about free cash flow—

Yeah, and profits. But the reason it’s a tolerable business is we really get to trade on ARR, even through the IPO to some extent, right? We get to trade on this ARR, and as long as it lasts, it’s a great deal. As long as the multiple is high enough, that’s where the leverage is, right? I get paid off this ARR. And so, if the burn multiple is attractive, it just makes the whole thing run on afterburners or steroids, doesn’t it?

Rory O'Driscoll

High is bad, low is good, negative is better. I think the real point is: let’s not obsess over the ratio. The real comment, I think, is that when all you’re graded on is growth, it’s not easy, because growth is really hard, but at least it’s a one-dimensional scorecard, especially on ARR, right? When you’re graded on growth plus profitability, which happens to us all at some point in time, it gets a lot harder.

You’re right: there’s often, especially in enterprise software companies, an ability at the margin to push really hard on the sales and marketing pedal or the free-user pedal, and you get some revenue, but just not commensurate with the marginal spend. As long as you’re just being rewarded for growth, you can do that. But once you have to deliver profitable growth, it all gets harder, right? So, that’s not the shoe that’s dropped yet, but it will.

I just want to make one other point on that. It’s so important. People talk about burn multiple and they’re like, “Oh, my burn multiple is good,” but they sometimes just forget that there’s also absolute burn and then not having money. They do forget—

Because the implied assumption, again—I remember thinking when David Sacks published it, and it was a very clever comment—is that if you have a good burn multiple, you should, in theory, be fundable, right? If you’re adding a lot of ARR and spending a lot of money to add that ARR, then in theory, you are fundable because you are venture-value-accretive.

But that’s a theoretical construct, and cash in the bank is an actual, material construct. Sometimes I see people tell me their burn multiple and not tell me their cash balance, and I’m like, “So, yippee. You could have a great burn multiple and still be out of cash on Friday. I need to know more.”

So, to your point, Jason, you can’t lose sight of just having money versus not having money. You see that behavior sometimes where you’re like, “I don’t care about your burn multiple. I care about the fact that you have less than 6 months of cash. What are we going to do about that?”

Guys, I have many companies with good burn multiples, and they’re going out to fundraise now, and they’re not getting love. They’re not getting attention, and they’re going, “Harry, I don’t get it. I’ve been brought up to understand burn multiples, to understand growth. What is going on?”

I’m seeing a very stark, binary world of haves and have-nots. Are you seeing the same? If you are, what would you advise this generation of founders who have good companies and good numbers and are feeling very confused by a rejected VC community?

Jason Lemkin

Wow. It almost sounds like therapy, doesn’t it? “I’m confused. I’m rejected.” But on a serious note, it’s a super-interesting subject because there’s an embedded set of assumptions in there. At some high level, does no one give a shit about anything that was founded before 2022? That’s really what you’re saying, right? Is all this old stuff—how uninteresting is it?

But I think the high-level comment is that it’s not that simple. It’s not going to be just, “No one cares.” Look at the recent IPOs. Many of them were non-AI-native, by definition. Most of the average companies that just went public were plus or minus 10 years old. By definition, they’re pre-AI. They’ve built perfectly good businesses capable of going public, maybe getting some lift from AI, but they’re a thing, right? So, I don’t think it’s all going to just, quote-unquote, go away.

But I think what you are wrestling with is that we’re in, for lack of a better word, an AI-first world in terms of mental models. When VCs look at any deal, there are only 2 ways of pricing a deal. You price a deal on hope, or you price a deal on the multiples.

When you price a deal on hope and growth, you can lean in on anything, right? You can get prices that, quote-unquote, make no sense because the growth ultimately comes and it all pays off. Once you start valuing things on, quote-unquote, the fundamentals today, then you can value a public company because at $400 million, it’s not nothing.

But to what we were talking about before we got on, a $15 million revenue company that’s perfectly good and has reasonable growth is actually of zero value to a VC because we’re in the upside-option game, right? It’s a perfectly good company. Someone should lend them some money. They should get profitable, right? But at super-subscale, the mental model of the VC is saying, a lot of the time, you can’t get from here to a big IPO. That’s the business I’m in. Therefore, I don’t have any embedded option value, so I can only value it on fundamentals.

If you’re doing $400 million, I could multiply 400 by 4 and tell you you’re worth $1.6 billion. You might like it, but I’ll give you the money. But if you’re worth $4 million, you have no value because $4 million is never going to be an IPO. Therefore, I’m just not going to do it.

There are a lot of companies that are going to have to build a much more capital-efficient model. Maybe it can make great outcomes, but it’s the zeitgeist—the groupthink—that’s not in your favor. I don’t know, Jason. Does that kind of resonate?

Jason Lemkin

I think that’s right. I see something that’s worse, to Harry’s point. I think Harry’s point on X was, “Listen, I’ve got a couple of companies that are growing better than triple-triple, double-double, right? They have an AI element, they’re interesting, and they’re struggling to get funded because they’re not ultra-breakout.” That’s a slightly different point.

Rory O'Driscoll

It is a different point. And not only is that true, I’m seeing something more problematic, that’s at the edge of toxic, which is that boards and investor syndicates that I’m a part of aren’t aligned on this. They’re not seeing it.

I am seeing many VCs that have been around for a while, especially ones that are doing just fine, right? Maybe they aren’t going to every AI hangout in San Francisco or everything, who, when they hear numbers like this, have no concern. There’s no concern.

I had a portfolio company kind of like this. My advice to these guys is: just take it. If it’s decent, just take it, because some of these VCs are still living in the past. They’re still living in the past, and I think they give founders terrible, borderline-inadvertently-toxic advice. They’re still giving 2022 and 2021 advice from the corner office, and I think it’s dangerous for founders.

Can you clarify that? What are you saying? I’m genuinely confused, which, by the way, speaks to the complete lack of certainty on this issue. Are you saying the bad advice is to need money, to raise money, or not to? I mean, are you saying it’s a dumb thing?

Jason Lemkin

Here’s the bad advice, Harry. I mean, Rory, Harry, in my company, we’re at $15 million ARR. We’re growing 100%.

Okay. Our burn multiple is good.

Rory O'Driscoll

We’re an AI-enhanced mug-making company. It’s good, but the TAM isn’t enormous. The numbers are there, right? But it’s not people.

Jason Lemkin

I see VCs saying, “Don’t worry, you’ll get the round done. Take your time, and let’s optimize around price. Let’s see how it goes. There’s no rush.”

Then I hear, “Hey, Scale wants to put in money at 250 on that deal,” and my advice today is: Rory’s a pretty good guy, but even if he isn’t, take that deal now.

I hear too many folks say, “Oh, you’re triple-triple, double-double, or better. You’re golden. Don’t worry, kids.” And I think that’s terrible, terrible advice in 2025. Terrible advice.

Rory O'Driscoll

I agree. I think those are perfectly good—in fact, they are great—numbers, right? And with an upside story, you could fund them. But I agree: a totally non-AI story, if you’re doubling at $15 million and you’re still so far below a $300–$400 million exit value that you’re many years away from it, if you can get a deal done, you should take it. You shouldn’t be optimizing.

What you’re saying, which is good advice, is that if you’re one of those companies, you should be getting your funding done and being damn glad to get it done, right? It may well be that 4 or 5 years from now, you’ll have the last laugh and you’ll be sitting there going, “I told you, you idiots.”

This is great. And you can email all the guys who turned you down and laugh, right? But right now, there was a lot less money for that deal, and it makes sense. Again, it's the comment that if you had $200 million, I can tell a story because, again, I'm going to repeat myself: of the 15 IPOs year to date, 10 of them have almost no AI story, right?

So it's not like you can't make money not doing AI. That's absolute bullshit. But to start today at $10 million and believe that 7–8 years of compounding can get you to an IPO 8 years from now, that's a much harder undertaking in a world where everyone feels that AI is the story.

So for those companies, Jason, you're right. In fact, we have one in our portfolio that I'm thinking of specifically. You should just get the deal done. Raise at a reasonable price. Continue to grow, but be capital-efficient. Don't get lost in just your burn multiple. Focus on your cash.

If you're right about your business, you'll be right in the end. I think a key part of being an entrepreneur is being willing to prove everyone right, even when they all think you're wrong. But you should operate for the next couple of years as if cash is pretty damn tight and scarce.

But I think—and I want to hear Harry's anecdotes—just one thing I think that's changing: I don't think there are any non-AI deals anymore. I agree, but what I mean is, listen, there's cybersecurity, there's fintech, and then there's B2B and B2C. I think that's all that there is in our world.

I don't think that even if you're not an AI company, you're outside of this trend. The other thing that's said in the ICONIQ report—the September report that just came out—is that 94% of public software companies call themselves AI companies, and the majority mention their AI agents. Adobe has $5 billion of AI-influenced revenue.

My point is, I think we're leaving the day where there are 2 types of companies. Now, we may debate what an AI-native company is, but I just don't think most VCs are going to pick up the email or the phone anymore. They're just going to assume everything has an agent. It has to.

One thing that has really shocked me is the mimetic behavior—and this sounds obvious, given the sheep-like analogies applied to venture—but it's how concerned investors are about going against a kingmaker, whether it's Harvey, Abridge, or any of the kingmaker companies.

We have a couple of companies that are the second or the third, and going against the kingmaker in the Valley is the most unpopular thing in the world. You cannot get funding. That's obviously very binary, and of course you can—I'm being deliberately binary—but wow, investors are not willing to fund anything if it touches a kingmaker or is in close proximity.

In other words, founders listening: when you raise, raising to deter others from raising is really a working strategy right now.

It's definitely a strategy, right, and we'll see if—and it does have an impact. The closer your customer base is to also being venture-backed companies, the more relevant it might be as a deterrence, because your customers might also feel that you're the king.

But I wouldn't overextrapolate. I do see the effect. Maybe there are 2 separate things. Do I believe that thinking exists in venture? Yes, I do. I don't fully share it, but I acknowledge that you have to factor it into your decision-making and your risk analysis.

The question is, is it a binary no, or is it something you factor in and then look at the facts? We're the latter. We have done deals where the leader has been funded by one of the top firms, and we've also done deals where you look at it and say, “Oh, and by the way, they're doing really well. Maybe you're not going to get there,” right?

So it definitely is a factor. Then the second question is: provided that second company can access capital, do the customers give a shit? The answer is that they do in some markets where it's very Valley-centric. If your first customers are also VC customers or VC-backed companies, then you get this dual loop.

But look, the reality is, you've raised money. Someone has raised money from Sequoia, they have a big portfolio, and they're known to be aggressive. You're like, “Hmm, do I want to do that?” If you're competing against that and selling to oil and gas companies, they can barely tell the Sequoias from their KPs. You know what I mean, right?

So it's TBD. It's a thing. I totally agree: it's a thing. It's important, but not dispositive, I would say.

So what's different now, going to Harry's point, right? I mean, it's always been hard. VCs have always been less excited about investing in number 2 and number 3, outside of the 2021 bubble.

I remember when Sequoia called it the—sorry, who did Uber buy?—Postmates. I remember when they called it the Postmates effect in 2021. There was so much money to be made that Sequoia decided they were okay investing in number 2 and number 3, because if you could make billions off Postmates, you didn't have to be in number 1. They called it the Postmates effect.

But people also understand that there are different number 1s in different segments. If you're really verticalized, there are different number 1s. Revolut and Chime are not the same company, right? We can come up with a million examples.

What is a little different in AI is that, in many cases, there's not an established brand. There's so much change, so much new budget, and so much confusion that so many buyers are under pressure and have a desire to make a purchase.

They want to buy a Harvey in legal, or they want to do something like Clay, which is powerful, but they may not even know what it does. They know they're under the gun in GTM. Being number 1 is so powerful when people know they want to do something, they've got to do an LLM for legal, or they've got to do this AI research for their clients, and they're asking, “Tell me who the hell to buy, Harry.”

That will calm down in a couple of years because the leaders will settle down, right? That's why Lovable and Replit are in a death match, and it's very powerful. They're both at 9 figures in revenue. They won't kill each other, right?

But them and other folks like them want to be that brand that nervous buyers don't know who to buy. I was just talking with someone at Bolt that closed a massive deal against Lovable the other day, right? They'd heard of both of them, but no one at Lovable called them back. No.

So you've got to buy. Bolt became trusted, so they bought Bolt because they were kind of equal on discovery. But who do I trust? Do I trust the one where the humans are in the deal and helping me, or do I trust the one where it's 90 days to get an appointment? Right?

We're not going to wait a year. When I was at Adobe, we waited 5 years to implement Salesforce. It's just not happening with AI.

Two comments on that. One is, if we're actually going to be responsive to Harry's question, you have to separate being number 1.

Yeah.

Which I agree with you. In the end, when the money is made, number 1 makes 67% of the market in a business market. Number 2 makes 20–30%. Number 3 makes 10%, and anything after that doesn't even matter, right? In a consumer market, it's even more skewed.

So I agree: in the end, when the total is written, you want to be number 1 in a segment. You'd be better off in a subsegment and being number 1 than being number 4 in a bigger segment. I totally agree with that. If nothing else, if you're not number 1, don't spend like you're number 1.

That's true. Even if you're growing pretty quickly, like Harry said, if you're the clear number 2 or number 3 and 80% of VCs are going to drive by, if you burn $100,000 a month, you actually may have the best exit—for founders, dilution- and time-adjusted. But don't be burning $2 million a month.

That actually triggered something else that I should have said because I glossed over this, and I think Harry hinted at it, but just to call it out: if 2 companies have $20 million, one from a great VC and one from a less well-known VC, it helps at the margin.

But what I didn't say—and Harry mentioned it, so I want to go back to it—is the thing we're seeing now: because that first company got $20 million from a great VC, 3 months later they get another $60 million from a bunch of people who want to follow that great VC. Now it's not a fair fight anymore because they got $80 million.

We're definitely seeing some of that, where it's not so much the money itself that's creating the momentum; it's the fact that the money sucks in more money, right? In the end, SoftBank proved to everyone's complete satisfaction that money alone cannot make winners, which was very kind of them—to run that economic experiment and prove the negative.

So it's not, in the end, dispositive, but there's no doubt in my mind—and I think that's what you're referring to, Harry—you are seeing some cases where you go, “Wow, not only did they get a great firm, they got Sequoia, they got Kleiner, they got whatever, but, oh my God, there's another $80 million on top of that from other people.”

Now you've got to say, not just, “Do I put this $20 million in this other company?” but, “Do I think this other company is nuanced enough and clever enough and has a differentiated enough strategy to beat the wall of money?” If they haven't, then it does cause a pause.

This is one of those examples, Harry, where—and I do this a bit during the course of the conversation—I end up going, “Hmm, I get that point.”

I should nuance what I’m saying a little bit. It ain’t as easy. The wall of money makes my trying to be pure and saying it’s just about the company perhaps a little unrealistic in today’s market. Do you think we are near peak madness, guys, or do you think we’ll look back and laugh at ourselves for having this conversation, given the might and the size of the markets that we’re entering?

You mean that it’ll be so much better? You mean it’s like—what’s his name? Alan Greenspan talked about irrational exuberance, and everyone remembers that, but it’s worth pointing out that the markets kept going for 3 more years and never went back to 1996 levels. So is that what you’re saying? Is it going to keep going, or do you think we’ll be laughing because it’s gone backwards?

I’m saying, are we just so [__] peak that we’ve got $25 million ARR companies being valued at $5 billion or $10 billion, xAI being valued at $30 billion, Mistral being valued at $10 billion with nothing? And that said, are we actually going to look back and go, “What morons? We’re about to see the biggest transition in spend from software to human labor budgets,” and realize how small-thinking we are?

Well, one of those 2 things has to be true, because that’s actually the interesting insight. If that massive transfer of labor doesn’t happen, then all these valuations are wrong by an order of magnitude. One of 2 things is going to happen in the next 5 or 7 years, right? Either you are going to see pretty profound productivity changes—you’re going to see companies like OpenAI hitting $200 billion to $300 billion in revenue super quickly. That’s option A. Option B, AI is still going to be wonderful, but you’re going to have a readjustment period that’s going to make your head hurt, and you’re going to go, “What were we thinking?” Without speculating yet—we can in a minute on which of them it’ll be—one of them is going to happen soon.

Jason Lemkin

We’re just getting started on what we’re going to do in AI in B2B. It’s just getting started. It is so early. Much like so many of these other waves, the direction’s correct; we’re early. I feel it. I live it through our vibe coding. I live it—we’ve replaced 11 people on our team with AI agents. I can see the future pretty clearly. We’re just starting.

For venture and for investment, though, not only am I worried about—we’ve never had 20 billion-dollar pre-revenue seed rounds before, right? This has never happened, right? Beyond that, what I’m even more worried about is that now you can’t—we’re back to the era where, if you have a billion-dollar round, you can’t even get in TechCrunch. You certainly can’t get on 20VC. There’s no way Harry’s going to swap you just because you’re the 14th company this week to raise at a billion dollars for your AI vertical SaaS company. You’re not getting on 20VC.

But when I met Harry, of course, you’d get on the next week, right? I mean, if you raised $200 million, you’d get on one of the first 50 episodes. So my point, what I wonder is: are our loss ratios correct? As venture capital firms, especially with larger funds—but actually, I’m just as worried about C funds because they’re paying such high prices for such low valuations—as long as we get it, as long as we’re cool if 80% of our unicorns implode or blow up or make no money—more importantly, just don’t make money for venture—as long as we’ve got it all right, even if those are at 1x or 2x ARR and we modeled it properly.

But I’m pretty sure we didn’t model it right in 2020–2021. Those unicorns—the B2B unicorns—we did not model them correctly in terms of our loss ratio and expectations. As long as we have it right, I’m cool with the party. It feels almost risk-free again, like it did in 2021. It feels risk-free.

Rory O'Driscoll

I also think, by the way, that Alexandr Wang’s acquisition created this very dangerous precedent in venture investors’ minds, where they’re like, “Well, if Alex is worth $14 billion, Bret Taylor’s worth $30 billion, and then Ilya’s worth $30 billion, and then Mira’s worth $10 billion or $15 billion or $20 billion or whatever it is.” It creates this kind of tidal wave of justifications for why we should pay entry prices where they are, rather than accepting that Alexandr Wang is just an anomalous event that may be seen in isolation.

Jason Lemkin

Irving Fisher, the big American economist at the time, famously said in the summer of 1929 that stocks had entered “a permanently higher plateau.” Well, it turns out he was wrong by 90%. So whenever I hear the word “permanent,” I’m like, “Permanent?” Have valuations permanently changed? No.

Well, talking about permanency of value creation or sustainability, Figma is down 63% since IPO day, now at $53 a share. There are VCs who are still locked up.

Rory O'Driscoll

Maybe times haven’t changed that much.

Jason Lemkin

No surprises here. It turns out stocks that open 250% above their IPO pricing probably will go down from there.

Rory O'Driscoll

Look, you can’t be critical about anything with Figma, right? It’s as close to as good as it gets, but it’s still trading at—

Jason Lemkin

26 times revenue.

Rory O'Driscoll

Yeah, agreed.

Jason Lemkin

Right. Actually, the bigger issue is I think you could argue that the markets are too generous today. The top group of public B2B companies trade at 20 times ARR, only averaging a 30% growth rate. The bar is actually—we’re lucky. On the one hand, we have this AI boom. On the other hand, perhaps because of profitability or whatever, the markets are fairly generous today in ascribing relatively generous multiples to growth that seems fairly modest based on our historical standards.

The bar isn’t that high on ARR growth to have a decent public multiple. If Figma was trading at 8x revenues, I’d be like, “Let’s all quit. Let’s just go golf and live off the fees, like 20, like 2,000, okay? Let’s not invest at all, because the world’s been destroyed.” But Figma is still at 26 times revenues.

Rory O'Driscoll

Embedded in what Jason said is a meta-worry here, because whenever you’re pricing a deal, you’re always going, “This is the way I think of it: How much extra do you pay for how much extra growth?” They say, if it’s doubling, can you pay 20 times? If it’s tripling, can you pay whatever? I’m just throwing out the numbers here.

You’re doing that, and you mentally have a benchmark. I always think the 10-year Treasury is the benchmark for the financial interest-rate world. In my mental model of SaaS for the last 20 years, the public-company median was the equivalent of the 10-year Treasury. The public market, as Jason said, up until about 2019, was 30% growth, roughly 6 or 7 times revenues—NTM revenues, right?

Then you said, “Okay, if that’s worth 30%, then maybe 60% growth is worth twice that in multiple,” and you could build your little valuation ladder in your head from there. You’re right, Jason: today, that 30%-growth stock in the public markets is trading at 20 times or 15 times, twice the long-term average. Therefore, one of 2 things is true: either it is different because it’s more cash-flow-positive now or there’s some good reason, or even the core bedrock price is wrong and, when that goes down, everything drifts down with it.

That’s like when the 10-year Treasury trades—everything that’s linked to the 10-year Treasury goes up and down with it. If the bedrock, if top-tier SaaS at 30% growth, reverts to 7 or 8 times, which was where it was happily for a decade and a half, and everything else goes down accordingly, it’s going to be a tough day. So, that gnaws at me every once in a while.

Jason Lemkin

We've seen Klarna get hit as well, dipping below its IPO price. StubHub was hit massively—absolutely bombed.

Do you think there's going to be an impact here in terms of people being less willing to go public, with the more recent IPOs facing a harsher pricing environment?

Jason Lemkin

Somewhere, Bill Gurley is lighting a candle and saying, “I told you so.” In many respects, this is good. I should qualify that from the banker's perspective—actually, not even from the banker's perspective. Their argument is always, “You've got to pay them on the upside, because sometimes they go wrong,” and this is what going wrong looks like.

If you bought some of these at the IPO, you're hurt and unhappy, right? If you bought StubHub, I think you've been down pretty consistently, and a lot, right? Klarna traded up and then is down. So what it says is: what's the rational buyer of IPO stocks going to do? He's going to say, “I actually want a little more give in the valuation to make sure I can get this done. I'm probably going to want a wider spread, because they all price off the public comps and say, ‘Look, I can already buy any one of 10 names at 10 times NTM revenue that are already public. If you're going to make me buy some new shit that I haven't seen before, that hasn't traded before, I'm going to pay you 8 times so I can make a pop,’” right?

And that's the pop. Sometimes it goes wrong and they get a huge pop. But that's the thinking, right? You've got to give me a discount to buy the new thing. Now they might say, “Hmm, I've been thinking there's more risk here than I thought. 7,” right? And you're going to see that.

So I agree: you're going to see some reflection in the capital availability. I mean, it will impact pricing at the margin. I don't think it's the end of the world. I think the trend is okay, given the markets, but there's definitely going to be a little more weariness and a little more focus on price from the buyers. And then the question, to your point, is what happens on the seller side? Do people say, “I'm not going to go out because I'm not going to get my price?”

Speaking of what happens on the seller side, EA is going private in a $55 billion deal—the largest take-private of this kind—with Jared Kushner behind it. I'm just like, wow, that's impressive, Jared. Really well done for being the broker behind this. How did you analyze that? It's a momentous deal.

Jason Lemkin

It is. It's the largest LBO in history. It's $18 billion of leverage, which feels like a lot for a venture-backed deal. But if this was an industrial manufacturing company, those dudes leverage those things at kind of 6 times EBITDA, right? So it's not a lot of leverage if you're a recurring-revenue, boring business, but it's a lot of leverage for a hit-driven games business.

So, interesting at that level, right? Biggest deal ever, but I've got to say, Silver Lake are part of this, and they've been astonishingly smart at the decisions they've made, from Airbnb to—I think the real killer is the Dell EMC acquisition, where the Dell take-private was a work of genius and has made Michael Dell one of the richest men in the world. So I look at it and go, pretty smart money at the helm. But yeah, quite a big-ass deal.

I'm not a gamer. Jason, are you a gamer?

Jason Lemkin

You know, all my gaming energy has gone into vibe coding.

Jason Lemkin

Ah, it's the same dopamine hit, really. Interesting. The attention economy has gone elsewhere for Jason. Yeah, okay. It's gone elsewhere. Howie, are you a gamer?

Howie

I honestly think people who have time to game need to get on with other, more important things.

Oh, David.

David Kellogg

No, I'm really sorry, Rory. Come on. When I was younger, it's like: go to the gym, get outside, spend time with friends. Come on. You're going to level up a character that dies when you could level up yourself? Come on, really? Spend hours on Grand Theft Auto? It's like, go play Grand Theft Auto and hijack cars. No.

I used to like the one where you go around and shoot. I remember first-person shooters. They were great in the early 90s. It was a big invention then, man. The first Doom—it was like, wow. But yeah, not a gamer now. It got all [?] in your day, wasn't it, Rory?

Jason Lemkin

The thing I don't understand—listen, I'm not an expert in these massive buyouts, but what I like from a venture perspective, from an investor perspective, is that EA is obviously an iconic company with iconic products, all this stuff, but it's shrinking. It's not even growing, right? 1%.

And so, yes, getting acquired for 5 or 6 times revenue, or whatever the enterprise value is—the exact number—it's pretty healthy considering that it's not growing. The stated strategy is more AI, which, Harry, startups—if they're number 2—can't even get funded with that. Now, we could say EA may be number 1 standalone, right? But that doesn't even make logical sense to me as a reason to spend.

There may well be some arbitrage. One possibility is to restart growth and get the multiple higher. There's got to be some version of it, but something's got to change. I don't know if the pixie dust of AI is going to revive these franchises for a variety of reasons. My point is, I love it because we all have a couple of EAs. If I could sell a few things growing at minus 1.31% in my portfolio for 5 or 6 times revenue, just send me the e-signature. I don't need to read the disclosure statements. If it's 6 times revenue for something growing at minus 1%, I'm all in, right? I don't have a lot of assets, but I've got a few.

Speaking of size that blows you away, one of the things that blew me away this week is the sheer size of data center requirements and energy capacity that Sam Altman needs with OpenAI. Very candidly, he would need more energy supply than India's capacity today in 8 years. OpenAI is planning for 125x energy capacity in 8 years. Is this sustainable? Is this within his “I need hundreds of billions of dollars more to make what I plan to do” framework? Analyze this unwaveringly infinite, insatiable demand for energy.

You're right: the energy demand is an outgrowth of the compute demand, and the compute demand is an outgrowth of their ambition. As long as the revenue keeps growing like it is, and the capital keeps coming like it has, they're going to be allowed to make that bet, right? And as we said last week, it's pretty clear that he and the other 5 or 6 people making that bet—and it is only 5 or 6 people—are all going to make it until some unequivocal feedback comes back to say, “Dude, this isn't working and you have to stop.” That hasn't happened today.

Will it stop on the road between what's roughly a $12 billion run rate now and a $200 billion run rate in 2030? My gut is that somewhere along the line it will, but the way life works is, if someone's rolled the dice and won 6 or 7 times in a row, he gets to keep rolling. That's all you can say, really.

I've really internalized this whole debate. You can step back and stare at it and wonder, but what do you actually do with this discussion? In the end, it's fun to talk about, and each week, as we talk about it, you do what a good little forecaster does: you update your priors based on new information. That's what I'm doing literally every week, because we have some version of this discussion every week.

The weird thing is, every week there are some hints that say, “Oh my God, this thing can work forever.” Typically—and this is my construct—it's in the technology, in new things like that. I'm sure Jason thought, “Claude can now code busily for 30 hours with no human intervention.” It's amazing shit on the technical side.

I did 3 hours myself. Yeah, keep going. It's a big change.

So it's huge. The technology keeps moving forward, right? If you look at this, my bigger point is that if you look through a science lens, this is all doable and possible. I think if you look through a finance and economics lens, at some point in the next 2 or 3 years, the scale of the ambitions becomes too hard to fund. You can find evidence of that, right? Then you pays your money and you takes your choice, and in the end, you manifest that bet by deciding whether you want to buy your puts or calls in NVIDIA.

You ask about power for data centers. Here's just the weird thing I'm thinking about—and I'm not an expert, right? I believe that Sam's going to solve this problem. But, like a lot of things, I believe it's hard for us to appreciate the scale of the problem. The $100 billion that he's doing with NVIDIA—the one they just announced—needs more power than all of New York City.

So it will only be a couple of years before the cities of the future don't even have humans in them. The next New York will all be GPUs, with maybe 20 people managing an entire city—a New York City of data and GPUs. They will be producing AI. An entire New York will have 20 people in it.

We'll be this weird future where, out of however many major cities we have in the US—20 or something like that—half of them will now be just AI cities, full of GPUs, because this 10 gigawatts is more than New York. That's the weird thing. I'm like, what are they even going to look like? Our country will be dotted with these Stargates that are larger than New York City, with only hundreds of people working in them and the equivalent of billions of digital minds.

That's why I don't think we can even fully appreciate the slope of the curve. I think Sam will figure out fusion and the trivial things on the way to get there. But Jesus Christ, half of our cities in the United States may be these massive Stargates with no humans. I'm going to zoom back out to the wider concept and the thing I said earlier about how the technology side is going to be exponential. I think the economics will grab hold of it. I think what Jason said was a microcosm of what will happen.

I don't think the level of growth—I think the growth rate will slow more quickly. I think all the practical realities required to make that much power, roll out that much data, sell that much software, and get enterprise to adopt that quickly are going to matter. I personally think that the rate of adoption forecast over the next 4 or 5 years that's implicit in all these data center assumptions will, in retrospect, prove to be too optimistic. So there, I've crept out on the limb and said it. I think we'll be revising those forecasts down over the next 5 years.

I feel very stupid, honestly, because I look at $1 trillion required to fund data centers for OpenAI alone. I'm like, I don't know where that money comes from. I know we said, "Oh, there's 5x the demand for Anthropic's round." That's cute. That's like $50 billion.

This is $1 trillion for data centers alone, for OpenAI alone. Where is the money? Sovereigns don't have that. I mean, well, actually, funny enough, they do. I mean, they'd have to put it all in.

Yeah, agreed. That was just being precise, right? Maybe the better point, Harry, is that you're dealing with numbers at that scale. A trillion dollars is a lot of money. The joke used to be Everett Dirksen's quote: "A billion here, a billion there, and pretty soon you're talking real money." Now that feels laughable because the real truth is, in OpenAI land, a billion here, a billion there, and pretty soon you're not talking about that much. We're talking about a trillion, but a trillion? You're talking about real money.

The trillion. Here's what I think, again. I've said it many times: I have more and more respect for the way Sam communicates. Simple things said a little bit ahead of time, said more clearly than we realize, including the beginning of Stargate, which I didn't understand, and why the hell Larry Ellison was there with Donald Trump. I didn't get any of it.

I think he is willing the trillion into existence. I don't think the answer is clear. I don't think you can get every sovereign wealth dollar, but whatever it is, I think it's sufficient. Here's my view as someone who just lives in AI today: 12 agents vibe-coding 2 hours a day. If I had to stop, if I had to only use GPT-5 and Claude 4.5, and I couldn't get any more GPUs, and I couldn't run any longer context window, and I couldn't do anything else, and I had to live in that, it would be okay. It would be a bummer, but it wouldn't have to stop.

Let's say it had to slow down because there wasn't a trillion. I think Sam is just willing as much of this into existence as possible because of the future. If we come up short—if it's $400 billion or $600 billion—we don't have to buy all the GPUs. The world would be okay if they had to last 6 years instead of 3 years, or whatever the depreciation schedule is. So I think he's willing it into existence without it being a certainty, because we can stop at $700 billion.

I think that's actually right and quite insightful, and it allows us to make a really important distinction. You see this a lot with the very best entrepreneurs. The sheer act of willing something into existence like this has just been amazing, and the broad direction they took in 2016 has been entirely vindicated. It's entirely probable that the same broad direction is correct for the next 8 years. What that means is, as long as OpenAI stays ahead of that train and on top of that train, they're going to be the winner. Whatever the prize is for being the best company in AI, OpenAI is going to get that prize. That's his job, and he's doing it better than any other CEO of this decade.

That's true. It is also equally true, as any CFO who's had an ambitious CEO knows: just because a CEO says, "We're going to treble next year," doesn't mean we should buy real estate and hire people as if we're going to treble. Maybe we should plan for a double and be ready to hire more if it starts to happen. The problem is not big visions from a big-V, the most visionary CEO of our decade. The interesting thing is, if you start valuing everything as if all that's going to happen and more—if you start valuing Oracle as if all that debt's going to be paid off and more—we might look back and say, "We saw this visionary person leading us to the promised land with big metaphors like 'trillion,' and we foolishly thought it wasn't a metaphor. We thought it was a PO, and we literally borrowed money against the PO." Five years from now, that's where you could be.

We could say, "Oh, I get it. OpenAI is still the best company on the planet for AI. Its growth rate has slowed to a shockingly small 50% or 60%. It's freaking amazing. It's doing $30 billion, growing at 50%. It's astonishing. But maybe they don't need a trillion dollars of capex this week." Then the ripple effects of that will be where the fun starts. To me, that's at least as likely a scenario as achieving the full kind of thing.

As I think about the metaphor of the CEO and CFO, we've all been on those boards where you have the wildly aggressive CEO, and you don't want to trample them. You don't want to say, "Don't be aggressive," because their aggression is what made you all this money. But you do want to say, "Please get an experienced CFO who quietly will make sure that we don't run out of cash, that we don't actually spend that until we see the revenue coming in." That's obviously what the economy as a whole perhaps should be doing here. Maybe we shouldn't be borrowing every dime on the assumption it's all going to happen.

Every time I do this, I say to myself, "How much should I have in the S&P this year?" But it's been up.

No cash.

No cash. I'm thinking less and less right now. I'm looking at it thinking, I can't be this good. This has to be a peak. This has to be a peak, baby.

And then you also said about revisiting your priors. I always thought that Zuck earned the right to do the next thing. He earned the right to do the next thing. And I have to say, my faith in Meta's AI strategy has just dwindled and dwindled and dwindled. I hold that in stark contrast to them being my largest public position, to be very open.

I sympathize. You just—

Never bet against Zuck. But I'm looking at it. I'm looking at Alexandr Wang. I'm looking at the treatment of Yann LeCun. I'm looking at how they structure teams, and I'm going, this is not well-run. OpenAI and Anthropic are coming for you. Microsoft and Satya are great. Sundar's got Google. Am I wrong to have such unwavering allegiance?

Well, it doesn't sound like you do. It doesn't sound like you do. You're voicing disloyalty here. One thing's clear: Mark Zuckerberg is not the communicator that Sam Altman is, getting up there with the thick glasses and saying things that—if I don't understand where the hell he's going, and Harry doesn't understand—not saying he's not getting there, but good God, we don't understand. He's not one of the great communicators at the moment, Mark Zuckerberg. He might be a great connector with technology because this Facebook engine is unkillable, but man, he's a crappy communicator, right? None of us understand. I'm not saying his AI strategy isn't S-tier, but I don't think any of us understand it, not for the life of us, where the hell it's going.

I think you can say it isn't S-tier, bluntly. It's a desperate attempt to throw money at a problem, bring in dream talent, and try to throw it together in a way that hasn't worked very clearly, very quickly.

Yeah. But he was clear just this week that he'd rather burn the $20 billion in operating income and fail than become irrelevant. That was Altman-level clarity, but maybe not in the way I wanted to hear it. It makes total sense: he'll burn every billion—$20 billion of operating income—to be in play rather than not be at the game. I just don't know what the game is, unfortunately. I don't get it.

Agreed. I mean, I think there's a lot to unpack in that. You're right that statement is the most important statement: the man who owns, with unparalleled power, the $200 billion-revenue, whatever it is, $70 billion free-cash-flow business, is totally willing to spend the money to make the bet.

So the bet's going to happen. You're exactly right, and the bet's going to happen because that's his evaluation of the risk-return. And then, to the point you echoed back for me: when you're successful, you earn the right to roll again, right? I said that, and I stand by that statement, but there's 2 different statements. You earn the right to roll again, which doesn't mean you were right. If there were a Kalshi bet—and maybe I should check it—that some version of this AI strategy will not produce meaningful revenue despite a $20 billion burn and will look more like Meta VR and less like Instagram and WhatsApp...

Two of the most brilliant acquisitions of the last 2 decades. I’d take that bet. I’d also be glad to have backed the leader who got 2 right. Remember, you get 2 right, you get 2 wrong. The 2 right more than swamp the 2 wrong. If this guy was just building a venture portfolio, he’d have a 50% hit rate and he’d have a wild DPI. I don’t get it for this deal, but there you go.

You do, but it doesn’t mean that you rate the quality of their decisions in the way that you used to.

Well, you could also say it’s just a harder bet. You’re right. It’s a harder bet. People are obviously vitally important, especially CEOs around founder power. But, to some extent, the wider comment is that the bet for the last 15 years was writing this brilliant invention that you had in 2003 and just optimizing it. And that’s hard, but it’s a lot easier than now, when you’ve got to invent a whole new thing a second time.

With the exception of Mr. Jobs, very few people have ever built a whole new thing differently. Right? Because Jensen's right. The zoom-out comment here on risk in AI is not, “Oh, like that [bleeped] statement that AI is going to help us target ads better.” That’s in the noise. The big-picture comment is that if you spend 2 hours a day on ChatGPT, that’s 2 hours a day that you are not spending on Facebook, and we live and die on our attention. So we’re just going to make [bleeped] until somehow we get people to come back and play with us. We’re going to have characters whisper sweet nothings in their ear. Whatever it takes, right?

How significant do you think it is that ChatGPT now enables a buy-in-ChatGPT feature, totally opening up commerce so users can absolutely buy following recommendations and suggestions?

It’s clearly a trend they’re all exploring. There are two different kinds of protocols on this, one from Google and one from GPT, for how to do this. All companies that are involved in e-commerce are looking at this. There’s clearly going to have to be some monetization of all these free users because, as we just discussed, this stuff is $1 trillion worth of expenses, and $1 trillion isn’t going to cover itself.

There are only 2 or 3 things you can do with free users: you can sell them [bleeped], or you can sell advertising to them. They’re going to press this button. This is obviously going to happen, and I’m sure the other shoe to drop at some point is advertising. It all makes sense because it’s the only way—it’s 1 of only 2 ways—to monetize the free users. So it’s going to happen.

I just think it’s an experiment.

Yeah, exactly.

I think we’ll see a lot of these, and it’ll be confusing to us because they’ll all get a lot of PR. They’ll drag out the Collison brothers or Tobi Lütke, and they’ll all do joint PR. But whether, for OpenAI, this is a top-5 initiative or whether this is just another integration, at the end of the day, I’m just not sure. I’m just not sure.

I mean, is this the future of e-commerce on ChatGPT? We’ll see. There are a lot of arguments that it isn’t. There are a lot of arguments that that’s not how people buy the hottest shoes or the hottest watch today. There’s a lot of data from Instagram and Pinterest talking about how people purchase. So you’ve got to do something like $2 billion of revenue to move the needle at OpenAI for next year, right?

You’ve got to think in billions to have a new product, right? Especially—you’ve got a CEO of apps, right? I think that’s a tough job.

Because this is coming in like—when it took a little while for Google Cloud and Google Apps to figure out their footing because the numbers were so big, right? It used to be that Google internally almost made fun of Google Cloud for years because, in the Diane Greene days, it wasn’t even a rounding error, right? It was a distraction. Now, obviously, it’s a force of nature, right?

But it’s a tough job because if you’re the head of—if I’m the CEO of apps, I’ve got to come up with a couple of multibillion-dollar revenue streams that get there in 2 to 3 years. That’s nontrivial.

And just to be clear, you’re exactly right. Just for context, there are 1,000 billions in a trillion. If you’re going to spend a trillion in capex, you’re implying that you have $1,000 billion. It brings it into scale. You’re exactly right, which, in my view, exposes the absurdity of $1 trillion of capex. It’s going to be really hard to cover that nut.

I do want to discuss 1 very important one, which is Fiverr in talks to buy DBT. Both were super-hot companies. DBT was really super-hot. Fiverr reported last time $400 million of ARR. Contra said before it was $100 million. Taking them together, given growth rates, they’d be over $500 million if they were to combine. How did you guys analyze Fiverr buying DBT in this kind of combination coming together?

Smart. I’ll tell you why. The products seem to be adjacent, rather than overlapping, so you probably have a better-together story. There are always puts and takes one level down—how well the customers overlap and that kind of stuff—but at a zoom-out level, this is the kind of thing that simply has to happen over and over again in everyone’s venture portfolio because we have 6,700 unicorns.

We’ve processed 5; 15 have gone out the IPO gate year to date. So, 20 for the year. That implies we’ve got 30 years of this stuff to get through. Every time 2 companies combine, we have the unicorn list, right? It takes 2 midsize companies—I mean, $400 million was nearly there—and makes a bigger company. This is part of the job venture is going to have to do to whip their portfolios into shape to be IPO-able, right?

It’ll be noisy. It’ll be a hassle. I’m sure there’ll be all the drama of private-to-private transactions, right? But do the DBT investors do okay in this transaction? I know it all depends, but how do you expect this to play out?

I thought so, right? I tried to do 1 of these myself recently, but I’m not Andreessen Horowitz. The fact that Andreessen is the lead, or close to it, in both deals makes it much simpler on many levels. Not only does it make it easier to get people together in the conference room, right? Not only does it mean you already know each other, but just on paper, if Andreessen owns 20% of Fiverr and 20% of DBT and you combine them, it does kind of suck.

When you own 20% of a portfolio company and combine it with another leader, it totally makes sense on the spreadsheet, right? Great outcome. And now I own 8% after the deal. I go from 20% to 8% because I combined them and there’s dilution and all this. It may make sense in the real world, but if I own 20% and 20% and I get 20% together, there are a million reasons you should mash your own portfolio together, right? It just makes it easier.

It does make it easier. But I understand what you’re saying. Just to be clear, what you’re saying is: if I had ownership in 1 company but not the other, I have 20% ownership. I have 20% of the upside. And now you merge, it’s a 50/50 deal. There’s some dilution, and now you’re down to 8%. That is fundamentally the reason why these deals are hard. The preference stack makes it even harder.

But even on an ownership basis, there’s a little part of you that thinks, “I have this little at-bat, and if this company takes off, I’ll get 20% of the upside.” Now what you’ve got to say to yourself—and when you do this deal, you’re saying—“If this combined company takes off, I’ll only get 8% of the upside.” So your leverage of your bet has diminished markedly.

I get it, and I remember thinking that when we looked at some of these deals. But what I’ve got to internalize—and I think these guys have done a really good job internalizing it—is that 20% of something that’s not going public is not nearly as interesting as 8% of something that is going public, right?

If you believe that it’s not a continuum of value, a sliding scale of value, but rather it’s like electron states, there’s just a gap, and then you’ve got to go to the next state. If you’re above critical mass and you can go public, you get the keys. If you’re below critical mass, then your only option is, you know, to [unclear?] and whatever pain that involves.

So I think that’s why—wouldn’t you look? We’ve had these discussions in some of our companies, maybe. Would I prefer to have 20% of my bet? Yeah, but I’d prefer to have a bet that’s worth something, right? And I don’t mean worth it in the sense that a $100 million company is not worth it. But if you want to get to the IPO, and the IPO window is for $300–$400 million companies, you’ve got to do what you’ve got to do.

Of course. Of course. If you believe the upside is not bounded, per se, and you're optimistic—

It's so much better to combine 2 portfolio companies and own 20% together. I understand you can't argue the intellectual argument, but it's tough. VC firms are a collection of GPs, and it's a collection of interests. If my one winner goes from 20% to 8%, that's tough enough as it is, right? But owning 20% of something that's accretive—that's hard to argue against intellectually or emotionally, isn't it?

Emotionally, yes, but intellectually, no. You're right. I get it. I do the anchoring, and someone mentioned something you guys asked about last week or 2 weeks ago about individual portfolios. Again, I often do this: I sleep on my answer and revise it. Individual portfolios should be less diversified than group portfolios because there's some value to the firm.

And this is another one of the values the firm has: We have to be able to talk as a partnership and say, "Hey, even though you're going to go from 20% to 8%, this is something we need to do as we think about liquidity." So it's not ideal, but there's no point hanging on to a dream that's not going to happen when you can get a reality that is.

And look, the one thing—the fatal mistake that always scares me—is not the dilution, right? The thing that scares me is you go from a decent deal that's well run, where you know everything about it, to merging with something else, and then the combined entity screws it up. That, to me, is the really shitty outcome, where you took your 20% bet and turned it into 8% of a disaster, right? Which is why picking the partner and having it make industrial sense is key.

And that's why I think this deal felt to me, from a distance—I'm not the infrastructure guy at scale—but it felt to me like that's a damn smart, obvious combo. That will get critical mass. You won't be looking at the S-1 going, "Why are these companies together?" You'll be going, "Oh, yeah, I get it."

It was funny. There was a deal superficially similar to this, in terms of the numbers, that I tried to work on. There was the A investor, the seed investor, and the pre-seed investor, all along the chain. The A investor wanted to jam 2 of his companies together.

You were just telling him to—

The pre-seed investor had another company that I thought was mid from his portfolio that he wanted to jam together. But I get it. It was almost as big, but mid. And then I had this idea. I'm like, "Listen, I have a third company, a fourth company to combine. I have no shares in this other company. I'm going to go through 50% dilution, but I know the CEO, and he's the best in the industry," and the A investor and the pre-seed investor were both fine. I mean—

Yeah.

But both would maintain ownership in their own thing and their own properties, and nothing's happened. Nothing's happened.

And that's why you need to be an active investor. That's why you need to be a board member, because it is tempting to try and take care of yourself at the same time, but you can't. You've got to do it, because the whole point of this—remember, the bad thing about doing that is you actually will create the situation I just talked about. If the CEO isn't saying these companies obviously belong together, then it's probably a dumb idea.

Right? So backing—

When the CEO doesn't drive it, it's weird too, right?

Oh, poorly. No.

Yeah, exactly. This is what happened with Clari, though.

And this one's easy to be critical of, right? It's easy to be a critic. It's just combining a series of properties that Vista has underinvested in with another one that has scale but isn't growing. This is the bad—I mean, it's easy to say this is the bad version of Rory's story. Rory's saying, "We've got to combine 800 B2B unicorns," or whatever it is, right? But mashing together a bunch that are growing single digits is the suboptimal strategy, right?

If that's the case here, I mean, Drift is probably shrinking based on just looking at how the deal happened, right? It's probably shrinking and leaking all the Salesforce data for Clari, leaking everybody's data too because it's being ignored, right? To take it on the chin and push back, but I agree: In a way, the combination of a Salesloft-type company and a Clari-type company makes a ton of sense. You have a sales engagement platform, and then you have a forecast. It intuitively makes sense.

I wonder if tech PE firms now question their business model a little bit more when they see the multiples that you can get on the money that's being moved by your KKR and your big firms, combined with the increased loss ratio that will happen from an increasingly volatile new AI world. Are you suddenly going, "Shit, I'm not getting paid for the risk that I'm taking on—the multiple on the upside given the displacement on the downside"?

To be explicit, because you didn't make it up, what is it you're saying? Are the tech PE people looking at their business model, which looked so secure for so long, of buying SaaS companies, just running them, paying down the debt, and optimizing, and are they saying, "This might have more risk than we thought and less upside than the other game"?

100%. Buying your Pipedrives or your Coppers—or, you name it—gosh, it's harder than ever because, opportunity-cost-wise, you can move more money with better multiples elsewhere. And then, secondarily, displacement-wise, there are more and more ways in which they're getting attacked through better and better startups.

They might be saying that, just like sometimes we say, "Oh, my God, PE looks so easy. They just have these big sums of money and they do it." I think it'd be a mistake. Generally, the record of people trying to transition to a totally different sector is pretty mediocre.

I remember in 1999–2000, a bunch of the PE firms, like Hicks Muse, piled into telco just at the wrong time. I do think they're probably—so, to my view, going from what they do to making non-control, late-stage investments just because, you know, pick a name, Thrive, do that well, would be, in my view, stupid, because Thrive are really good at that and they're not.

I do think they should be saying, "Any deal we underwrite today, you better have a clear understanding of the AI downside risk." To Jason's point, if you have a lot of downside risk, maybe you shouldn't be doing these, right? I do agree with you there. It'd be fun to speculate on how—I don't know, actually, Jason—how embedded would an AI-native app have to be for you to say, "This thing is good for 5 more years of revenues"?

The problem—the problem is, and I think PE thinks about this maybe from a slightly different perspective, but I think this is the concern for anyone that's been doing B2B for a while. These products didn't change from about 2008 until 2023. They're the same products. Brian Halligan will agree, and I said this with Henry Schuck, and he's like, "Yeah, looking back on it, none of our products changed for a decade," right?

And so it wasn't just that we had high NRR, which was the spreadsheet glue for the PE model. It was the fact that the products couldn't change. I led the seed round in Pipedrive, and that product didn't change. It took them 4 years to launch a mobile app.

Yeah. You used to have 4 years to launch a mobile app. That was my first venture investment. It was a $1 billion cash exit, $1.5 billion—my first investment. But 4 years—you could have 4 years to launch a mobile app. Could you imagine today waiting 4 years to launch your AI copilot? You're dead in the water, right?

And so that was the part that was underappreciated and under-discussed. Yes, 140% NRR meant we could buy Marketo and fire everybody, but the products can't be static, and AI is the accelerant there, right? That's why I worry there aren't enough buyers for any of this stuff, right? Because it's the rate of change. It's unprecedented in business software. Unprecedented—the software.

Jason Lemkin

Yeah. I just want to double down on that—whatever the cliché is. That is such a big insight, right? Basically, there are 2 consequences of that. You're right: There were 15 years where we made the same product, and you didn't have to think that much about product direction at the macro level. It was roughly the same form factor. You look at Salesforce in 2002 and 2022; it's the same thing, right? And that's now changed. You're exactly right, and that's huge.

Yeah. The weird thing is—

Jason Lemkin

A new tech risk.

A new tech risk. And the weird thing is, even on our side of the table, for these new post-LLM startups, I'm finding—and I think we talked about this before—that product-market fit, when you locked into it in SaaS land, just didn't unlock for 10 years. Whereas here, you can lock in and out of product-market fit as the models change and approaches evolve, and you look back at the product a year ago and think, "Oh, my God, it feels totally obsolete." So there's more risk on our side of the table too, I think.

Jason Lemkin

Yeah, that’s why it’s good that growth is higher, because the risk is less stable. It’s less stable, right?

Yeah, that’s actually very clear.

Jason Lemkin

I think, to your point, that’s what makes our job harder than ever. I’m not asking for sympathy—I know it’s not easy making money—but that’s what makes it hard. The predictability of markets in the old days was easier.

A lot of things Jeff Lawson said I’m still processing. It was a good show. Even if people didn’t watch it, they should all go watch that one. His point was that if he were running Twilio today, it would probably be thriving because he sold at the API level and could benefit from the AI boom, right?

I’ve seen that with RevenueCat and others in my portfolio, and the seat model is under risk, right? I got burned out by LinkedIn people saying the seat is dead because it’s obviously not true at some level, right? Seats are growing, but, good God, now that we’re running 12 AI agents, we only need 2 seats of Salesforce.

Yeah, because you don’t have the people.

Jason Lemkin

We just don’t. And if Salesforce—if Agentforce—can do all 12 of those agents, then we’ll actually end up paying more to Salesforce. It hasn’t happened yet, but this change of not needing as many seats—we thought it was a layoff thing in 2022 and 2023. It was, “Oh my God, we’re laying off people. Smaller headcount is an issue,” but ultimately, if the economy grows, you get past it, right? You get past the layoffs and the companies reaccelerate. It’s a transitory thing, like a global pandemic.

But the agents taking over for humans and creating fewer seats in software—my God, it makes the PE model worse and our jobs harder. We could literally get by because now we have 6 agents that plug into Salesforce. They’re like 6 human equivalents, but Salesforce may change its API pricing, and they sure don’t need a seat.

It just makes it even tougher for PE to buy these seat models. Not only do the products not last a decade, but the AIs don’t need as many seats.

Totally. Two things on that. One is, you’re exactly right: seats don’t have to go to zero to be a lot more variable than they used to be. But then the second is, I was thinking of you, Jason, because you are always, if I may say it so pleasantly, so brutal about the impact of AI in terms of employment and the consequences.

I always recoil because it feels a little mean, but then I decided I could finish. I prefer the way you talk about it to this version of corporate speak. Let me give you corporate speak. I saw it, and I’m not dumping out of it.

The CEO of Accenture, when they said they were laying off a bunch of people, said, “We are exiting on a compressed timeline. People reskilling based on our experience is not a viable path for the skills we need.”

That is just a brutal corporate-speak epitaph. You think, “We think you’re no good in the AI world. You’re out.” I just thought—I mean, it’s entirely correct. There’s nothing objectionable about it; it was just such a wonderful mix of corporate speak and finality. I had to laugh and print that out today and look at it.

Wow. There you go: “We are exiting on an expedited timeline—on a compressed timeline. All you people who are no good, have a great day.”

I’m going to finish on my wild card. I’m not naming names, and it is not political, but we always stick to advice for founders. As we move into a more and more political world, do founders have a primary fiduciary role to team members, investors, and shareholders to do what’s best for the company over freedom of political expression?

It would trouble me to say yes, in the sense that it would trouble me to think that just because you’re the CEO of a company, you’re not entitled to your personal opinion separately from that, right? That’s a troubling reflection of the times—that you’d even have to say that—because you should be able to dissociate the two most of the time, right?

I’ve been on boards where we’ve wrestled with that. You have a particularly outspoken CEO, and I’ve come to the conclusion that if they’re expressing their personal political beliefs on a personal basis, I’m inclined to let them do it, right? I’m not even inclined; that sounds too controlling. I would be actively resistant to stopping them.

Jason Lemkin

I think the recent trend at the company level of saying less has been smart. A lot of companies took a lot more positions 3 or 4 years ago, and it’s been almost fun to watch. Just like universities, they’ve realized that, speaking as a corporation, you probably should stick to the mission of the corporation.

In university language, the University of Chicago principles have been proven to be so much cleverer than anything else that all those other colleges are scrambling for that safety. As a company, you probably want some version of the same thing. You want to stay out of the culture and social wars, especially when they’re so vehement.

As a company, I think companies should stay out, and I think they should think long and hard before getting into anything else. What’s hard, and what I’m struggling with a little, is: when you’re the CEO, do you really give up all personal rights to have a political opinion?

It’s worth pointing out that there are a lot of roles in society where the job does involve exactly what you said, Harry: not having a political opinion. It used to be, for example, that people in the military were scrupulous about not declaring their political opinion. When Eisenhower was solicited as a candidate for president of the United States in 1952, they didn’t know if he was a Democrat or a Republican, and both sides asked him to do the job.

I wish that wasn’t the case, but the wisdom from some of those old learnings about staying out of it is making me tweak my opinion a little. Do you understand what I’m saying? I want everyone to be able to have a political opinion, because I think everyone in this country should be able to speak. That’s why this country is so freaking amazing.

People should be able to express their opinions, and we have to get a lot better at not trashing other people for them. But I do recognize that, in some cases, institutions don’t express political opinions. It would seem a shame that that needs to extend, but I understand the point.

Jason Lemkin

Well, I’ll give you a tactical answer. Especially for folks who are active on social media, once in a while you’ll say something that either you shouldn’t, or maybe you should but you went too far, or you said it the wrong way.

Harry and I have known each other for a long time. I can’t think of very many times, but I think a couple of times one of us has DM’d the other and said, “Hey, here’s a tweet. Maybe you didn’t really mean it,” and we’ve deleted it or modified it. It happens—not all the time, but it’s happened multiple times.

There’s a level of trust. Harry and I have done this, and I’ve done this maybe with 10 CEOs that I know a little bit. I would do it with Jeff Lawson, who I barely know. I would do it with Brian Halligan. I’m not saying—but I would do it with folks we’ve had on the show, right?

I’ve done that multiple times. I’ve said, “Listen, you be you, but just to let you know, this tweet may not have landed the way you thought, or it might bother some folks on your team, or it might bother some folks.”

I can only think of 1 public-company executive who responded positively to that. Not that no one was negative—no, because I don’t do this all the time. I’m not preachy. It’s always a quiet thing. I try to be, but I only do it when I know the impact was more than they thought.

I can think of 1 that all 3 of us know well. He was like, “Holy crap. I didn’t—I did. You’re right. Thank you.” Sort of thank you. It took a beat to get to thank you. It wasn’t an instant thank you at first. It was like, “You’re wrong,” but it was a thank you.

Every other time—and I’ve been told this by all the other 9—it’s been, “You might be right, Jason, but I don’t care.” They say, “I feel so strongly about this. If I alienate 40% of my customer base, if I upset some of the less-represented folks on my team, if I do whatever, I feel so strongly about this. I don’t care. If I piss people off, I just don’t care.”

So I’ve become much more reluctant to do that. It’s rare, right? I only do it at the time when I think I can really be helpful. The moment I see what I think is a mistake, I do it maybe once every 4 or 5 months.

I can only think of 1 time where it was well received. What I learned from that is, like a lot of things in venture—and these aren’t companies I invested in, but they are public-company executives I know—it doesn’t really matter what I think.

I will provide some feedback at times, and I’ll provide it multiple times, but at some point, it’s your company and your keys.

How important is it to post that?

There, we agree on that. But what does it matter what you and I think? My learning is that some folks who have been on your show—some folks who have very strong opinions, who I’ve talked to, just a handful—they just don’t care. They’re cognizant of the risks.

They're cognizant of the downside. It's not a mistake. Every once in a while, someone makes a mistake, right? There's a subtle distinction between bringing politics to work and not having that, which I think has been validated as the correct strategy. And I think that's almost—I want to say fully a given now, but I think that would appear to be the consensus, because everyone tried the other theory, tested it to destruction, and failed.

As a board member, I would struggle a lot to attribute some business blame to someone having a personal opinion that's clearly their personal opinion, because we have free speech in this country. It shouldn't be, right? So I struggle with that. And then you could say to yourself, well, let's just, to play out the other extreme, what happens if that personal opinion alienated 50% of the country such that they literally canceled all your business, right? Then you could argue at some point maybe you aren't the right person to run that company, or your team left—10 great engineers left.

But my learning is 9 out of 10 of the executives are fine with that. Let them go. I think Brian Armstrong was fine with it, wasn't he? I think Brian Armstrong said—and I didn't agree with Brian—but I think he said, “Go. There's the door.”

The one thing I do think is the attention economy is also more fickle than ever. And just like Deel and Rippling was such a big deal—

I think that's a smart point. It's—

It's not—actually, everyone forgets it. No one cares. This story, whatever story—do you remember that Elon and Donald Trump broke up in the most blazing of rows? No one. That was yesterday's news.

I think that's actually very insightful, Harry. Just keep moving forward, and people move on. The rearview mirror changes so quickly; it vanishes so quickly.

Guys, thank you so much, as always. This has been wonderful. You have changed or distracted me for 2 hours, which has been awesome. So thank you so much. You're awesome. Cool. Rock and roll.