AI Markets：与 a16z 的 David George 深度对谈

Let me start with what I think the big takeaways are from this piece, because this is the first time we've ever done this style of piece. We produce so much work and so much analysis—it's like exhaust inside of our team—and we thought, “We have so many different thoughts and points of view. Why don't we put them on paper and share them with the world?” So that was the genesis of this.

My big takeaways from doing this one are that AI demand is crazy. The actual uptake, growth, and quality of companies in AI are extremely encouraging from our standpoint. Companies are starting to run themselves better. I'm going to show you some statistics on that; there's been some buzz on X, including this morning, debating what's going on there.

But this crop of companies is more impressive than prior crops of companies, partially because the demand for their products is so high. That's the demand side. The supply side is healthy right now, but we are starting to see some signs of things that are stretched a little bit. I'll talk about what we see and what we're looking out for.

We've been fortunate to be a part of a lot of these great companies. The most exciting action happening in the private markets is AI, and it's happening in the private markets. We're going to show some slides about that.

Lastly, my big conclusion—and what has me so excited about where we are now—is just how early we are in this product cycle. Product cycles drive our business. These are 10- to 15-year cycles, and we're just at the very beginning of this one right now. So let's dive in.

We invest across all private stages. This is a chart that shows our activity. We're very busy, across all verticals. On the growth side, we've been most active in AI and infra apps, and then in American Dynamism, but we're also very active in our other verticals as well.

I'm going to zoom through some of these. I hate to do the a16z commercial, but I really like this slide. I think we have the chance to work with some of the best model, app, and infrastructure companies. Obviously—

I'm going to gong it on that side. Gong it.

I do like that slide a lot.

Soundboard effect here. I'm happy.

We debated how early to put that slide on the deck, and I said to put it further back, but I was overruled. Thankfully, anyway, here's some data.

We collect tons and tons of data as a growth team because we're seeing essentially every growth-stage company in the market, either as a portfolio company or as a prospect. We have a great data-analysis team, and we did some analysis. I think this stuff is super interesting. We geek out on it.

To me, the big conclusion from this is that 2025 was a year of accelerated revenue growth. Revenue obviously slowed in 2022, 2023, and 2024, following the rate hikes and the pullback in some of the tech markets. But 2025 reversed that trend.

It accelerated across different types of companies as we rank them by decile and quartile, but especially among the outlier companies, it really accelerated. You've probably seen us put this slide on a page before, but the fastest-growing AI companies are reaching $100 million in revenue significantly faster than the fastest-growing SaaS companies in their era.

There's a really important thing I want to call out about why that is the case: end-customer demand is so strong, and the products are so compelling. It's not because they spend more money on sales and marketing. It's actually the opposite. The best AI companies that are growing the fastest are not the ones spending the most on sales and marketing. They're spending less on sales and marketing than their SaaS counterparts, and yet they're growing much, much faster.

This slide shows the growth of the AI companies versus the non-AI companies. Roughly speaking, the AI companies are growing 2.5 times plus faster than the non-AI companies. That shouldn't be a huge surprise. The best AI companies are growing very, very fast.

We had to triple-check this data when we saw the top AI performers growing 693% year over year, but it matches our experience and the anecdotes we see from our portfolio companies. So that's growth.

This is the margin profile we're seeing in the data set. Again, these are internal data sets that we have of portfolio companies and companies that we look at as potential investments. Gross margins are a little bit worse for AI companies.

You've probably heard us talk about this before, but in a way, we feel like low gross margins for AI companies are sort of a badge of honor. If low gross margins are a result of high inference costs, that means, first, people are using AI features, and second, we believe those inference costs will come down over time.

In an odd way, if we see an AI pitch and the gross margins are super high, we're a little bit skeptical, because that may mean the AI features aren't actually what is being bought or used by the customers.

We're going to talk about ARR per FTE, but this is a new thing we've started focusing on. This is one of the things that got a lot of pickup and discussion on X in the last few days. ARR per FTE is a measure of the efficiency of how you run your company in general.

It encapsulates all of your costs—not just your sales and marketing costs, which is an efficiency measure we've always looked at in the past, but also your overhead and your R&D. The best AI companies are running at around $500,000 to $1 million per FTE.

The rule of thumb for previous software businesses in the SaaS era was around $400,000 in the last generation. Again, I'm going to talk about this a little bit more, but the reason this is the case is mostly because demand is very, very strong for their products. They need fewer resources to take those products to market.

David, maybe a quick clarification before we go to this slide: How do we define AI companies? Is that defined as post-ChatGPT, versus historical AI and ML companies founded by a certain time period?

Yeah, it's sort of post-ChatGPT, and some of them were founded right around that time. We'd give them a little bit of grace, but if their first product in market was an AI-native product, that's how we define it.

Got it. Maybe this is a good point—we can punt until later—but one of the questions a lot of folks are trying to understand is the magnitude of change in expected revenue and growth from companies from the SaaS era to the AI era. You've talked a little bit about the magnitude of revenue, but what happens to companies that aren't AI-native? Will they have a hard time competing against AI-native companies? Are they all shifting? Will we see more fallout? How should people think about their historical portfolios?

The way we're approaching this with our portfolio is that you need to adapt to the AI era or die. That's true on both the front end and the back end.

On the front end, you need to think about how you can incorporate AI into your product natively—not just attach a chatbot to your existing workflow, but reimagine what your product can mean with AI. You need to be aggressive about disrupting yourself and changing.

On the back end, I shared some of the statistics around the efficiency at which these companies are running. That's going to change, too. You need to be fully rolled out with the latest coding models for all of your developers and all of the latest tools across every function inside your organization.

The biggest uptake has been in coding so far, and that's where we've seen the biggest leaps. There have been major, major changes in the last 2 months—really, the last month and a half. Andrej Karpathy has written about this.

I was on a catch-up with one of our pre-AI companies. This is a founder who's very AI-deep, so he's adapting his company. We were talking this week, and he told me that he was frustrated with one of their products. He took 2 engineers who are very deep in AI and assigned them to build it from scratch with Claude Code, Codex, and Cursor. They had an unlimited budget for coding tools, and he said he thinks it's going somewhere between 10 and 20 times faster than the progress they had before.

The bills associated with that are actually high enough that it will cause him to rethink what his entire organization looks like. The conclusion was basically, “I need my entire product and engineering organization working this way,” and he thinks it's going to happen within the next 12 months.

But what does that mean for what the team design actually is? Where does product start, and where does engineering start? Even where does design start in that process? It feels like December was a turning point for coding.

Over the next 12 months, this is either going to hit and take hold in companies, or those companies are going to move much more slowly than their peers. As it relates to the pre-AI companies, the message is: adapt.

We have another example of a pre-AI software company whose CEO has gotten totally AI-pilled. He's saying, “We're going to become an AI product.”

We’re going to ship: your employees are now your AI agents. How many agents do you have? Those are the things that he’s talking about.

We have another company that was very extreme about it. The CEO said, “I now ask the question for every task that we need to complete: Can I do it with electricity, or do I need to do it with blood?” This is the extreme mindset shift that’s happening with our companies.

I’m happy to see that our pre-AI companies are moving very fast and trying to adapt, but they very much need to adapt to this new era, both in the front end, product-wise, and in the back end, in how they run their companies.

Totally. Tactically, with almost every portfolio company, you have to go line by line to understand where the founder is on that journey and how much they’re implementing from the ground up.

What you said in terms of blowing up existing operations is also happening in post-AI companies. Increasingly, people are looking every 6 months and saying, “The things we built 6 months ago could be vastly improved based on what’s available today.”

If that keeps happening, pre-AI companies need to increasingly 10x their efforts to catch up to that point.

Yeah, the good news for the pre-AI companies is that the business-model evolution is still in its early days. The most disruptive thing that can happen to you is a technology and product shift and a business-model shift at the same time.

I think of business models as a spectrum. I’m talking about enterprise, or B2B, just to keep it simple. The spectrum is basically licenses—and these were the pre-SaaS license-and-maintenance business models—then you had SaaS and subscriptions, which were typically seat-based. That was a big innovation, and it was very disruptive.

The architecture and cloud delivery were disruptive, but the business-model change was also very disruptive. Just look at what happened to Adobe as it went through that transition.

Then you have the transition to consumption-based, or usage-based, pricing. This is how the cloud providers charge, and many of the volume-based, task-based businesses have already adapted and shifted from seat-based to consumption-based pricing.

The next iteration will be outcome-based. When you do a task—and ideally, when you successfully complete a task—you get paid based on the successful completion of that task.

The only area where that’s really possible to pull off today is probably customer support and customer success, because you can objectively measure the resolution of something. But we’ll see what happens with the capabilities of the models. To the extent that functions besides customer support can measure those kinds of outcomes, that would be a huge disruptive force for incumbents.

Honestly, moving from seats to consumption might be a big disruption if the composition of companies changes as well. But that next one is the really big one.

For sure. Speaking of blood versus electricity, we should go to AR over FTE—this next slide here.

Yeah, yeah, yeah. The big debate on the next slide was, “Oh my gosh, look at the AI efficiency gains that are happening in the market.”

There’s a little bit of that in companies running themselves differently. Take the example I gave about the 2 engineers who are rebuilding the product. Sure, I would say my observation from our companies, even the AI-native ones, is that they run leaner, partially because they’ve grown so quickly and the demand is so strong.

I wouldn’t say yet that we’re at the point where companies have fully reimagined the way they run themselves. I think this is partly the result of our data set being the best of the best companies, with extremely high demand signals. They have fewer resources to serve that demand and, frankly, there have been general efficiency gains in the technology market coming out of the kind of 2021, most-bloated era.

We’re starting to see some early signs of that efficiency, but the wholesale shift to running your company totally differently is still early in that journey.

The coolest example I’ve seen in the public markets that anyone can read about is probably Shopify. Tobi’s awesome. He’s a CEO who’s close to the company; he’s in a bunch of our groups and stuff. He does a great job, and he fully embraced this a couple of years ago.

One of our staff writers actually wrote this whole big deep dive on how Shopify AI’d itself in terms of employee direction, process, et cetera. That’s probably just scratching the surface of what’s going to happen over the next 5 years.

That’s a good segue to the next section: What are these companies actually doing? Our favorite topic is lawyers. Lawyers have only increased in this new world of AI meeting lawyers, not the opposite.

I love the tweet—I don’t know if you saw it earlier this week—in which a corporate lawyer was quoted saying, “LLMs have actually increased my workload because every client thinks they’re a lawyer now.”

That’s a good segue to Harvey, which is the next slide.

That’s very good. Harvey’s so great. This is a real test for me because I love talking about our portfolio companies, and I’m supposed to go through this section quickly because I think people know these companies, hopefully.

One of the big things that we look for—and one of the questions that came in was, “How do you know that revenue is going to be sustainable?” These companies all grew really, really fast, but is it fleeting?

The big thing that we push ourselves to do is go super deep on revenue retention, renewals, and product engagement: actual time spent, how often people are logging into the platform, and what their activity looks like when they’re in the platform.

What you see on this page is that, with the onset of the much better products they’ve built over the last couple of years, plus the improvement of reasoning models, it turns out lawyering and reasoning go hand in hand.

Users are spending about double the amount of time in the product compared with before. It turns out that AI is really good at lawyering. Again, there aren’t fewer lawyers, but I think AI is making lawyers a lot more efficient.

The most important thing as it relates to Harvey is that users are spending a lot of time in the product and getting a lot of value out of it, which is great.

Let’s go to Abridge. Unless you want to keep talking about lawyers—I was just going to make a comment. In all the 7 years that I’ve known you, I wouldn’t have ever discerned that you’re from Kentucky, other than in this moment. By the way, you say “lawyer.”

That was a tell. I have a couple of those words in my vocabulary. My wife always jokes, “You go home, have one bourbon, and then you talk like you probably did when you were 18.” The Kentucky came out when it came to lawyers.

It’s 10:25 a.m. I have not had any bourbons today.

Important distinction. It’s important distinctions, yes, exactly.

Abridge is another super exciting one. Doctors rave about having access to Abridge and how much time it saves them and how much better it makes their lives. One of the customers we talked to described it as a trusted deputy.

The chart on the right shows something we look for. The blue line shows the growth in users, and the green line shows the engagement of those users. As they’ve massively grown the number of users, you’d be a little worried if engagement among the incremental users they were adding was going down.

Instead, they have extremely high usage among the people who use the product, and that has held steady and grown a little bit even as they’ve added tons and tons of users.

These are examples of the kind of data we look for to make sure we feel confident that the revenue these companies are generating is sustainable. These companies are growing faster than any of their predecessor companies, but the growth is very sustainable. It’s high engagement and high retention, and that’s critically important for us.

Same thing with ElevenLabs. Voice is the centerpiece of so many of the new AI tools. I talked about customer support on the B2B side, but so many other personal and business tools start with voice.

The usage growth is the thing that I love to look at on this chart. It’s staggering. This company is growing very fast and is a great example of one of these companies that runs extremely efficiently. ElevenLabs is really a great one.

Navan is the next one. This is a different example and a good example of what I was describing earlier. They were early to this AI shift, and they spent a lot of effort making sure they could take full advantage of the AI capabilities and make their business better.

The biggest way you can see it in their business today is in the handling of resolutions.

Part of what they have is agents that have to handle travel bookings or travel changes. AI is now handling 50% of those user interactions. This is hard stuff—travel bookings and changes to travel. This is not something simple like, “Tell me the balance of my bank account.” This is a complex workflow that AI is now able to handle.

The way you see that in the business is a 20-percentage-point expansion of gross margins over the last 3 years. That’s exceptional impact. You need to adapt or die. Their competitors are not adapting; they’re very old-school. While they’ve been sitting still and doing things the old way, Navan now has 20-percentage-point-higher gross margins than those incumbents.

Flock is doing absolutely incredible work. I’ve talked about them so much. It’s the most compelling customer value proposition that we see in our portfolio, because their ROI is solving crime. We’ve covered the 10% statistic before: each year, Flock is solving 700,000 crimes. The data point on the right also shows that, per officer, where there’s Flock, they’re clearing almost 10% more crimes. That’s a huge impact on the community. Obviously, they have a great business and financial model that goes along with it, but the impact from their product is exceptional.

Okay. By the way, I don’t know if you see the chat lighting up with people saying that they’re 3 bourbons deep.

I didn’t see it.

For what it’s worth, there is one question about how you think about the benchmark. If you were to think about traditional industries, like finance, for example, and use JPMorgan as a benchmark, how would you calibrate the Fortune 500 in terms of AI adoption? Maybe I’ll overlay that question with the one that Xavier mentioned as well. There was that study about enterprise adoption from MIT at the outset of last year, and they were measuring all sorts of wonky things. Maybe say a little bit more about how and what you’re hearing from Fortune 500 CEOs.

What we’re hearing from Fortune 500 CEOs, I would say, is—and maybe this is the key link between those 2 points—“We have to adapt. We’re dying to understand what AI tools we need. We’re ready to change. Our businesses are going to fully roll things out. We’re ready. We’re going to become AI companies.” That’s quite different from what is actually happening.

I think the biggest disconnect between that mindset and actual change in the businesses is that change management is hard. It’s hard enough to get people to just use an AI assistant to help them do their jobs better. Coding is probably the easiest one to get people’s minds wrapped around. Customer support is such a better, faster, cheaper, obvious thing. But in terms of general management of businesses, changing business processes and change management, it’s extremely hard to do.

I’m not surprised that there are anecdotes out there that suggest things are moving slower than expected. But for the best companies that are fully embracing it and actually know what to do, it already has tremendous business impact. I think there’s going to be a reckoning over the next 5 years over who can actually embrace change, push through change management, and adopt all the best products—and those that don’t. I think there will be major differences in productivity. We have some charts later in the slides that I can talk to, but the expectations around productivity enhancements, growth, and all that stuff are high. I think a bunch of companies will achieve those expectations, and the ones that don’t are going to be at a huge disadvantage.

Chime said they reduced their support costs by 60%. Rocket Mortgage said that they saved 1.1 million hours in underwriting, up 6x year-over-year, and that was $40 million of run-rate annual savings. We’re seeing pockets of it in non-AI businesses, and I think this is going to be a really interesting year to watch over the next 12 months. You’re going to see a ton more anecdotes, but there will be companies that can figure it out, and there are going to be companies that don’t.

Totally. A lot of these corporations have had to orient their businesses to be ready for AI as well. There’s one version of just using a chatbot, right, and how much productivity gain that actually gets you. Probably not a lot. But if you have to completely upend your systems, information, and backend to be ready for AI, a lot of that is probably latent and being built up now before you actually see the outcomes associated with it.

AI winners are driving the public markets. They account for almost 80% of the S&P 500’s return. This is the major thing driving the economy and the stock market. Public markets are doing very well, but the fundamentals are sound. Prices are going up; there have been some blips like the last couple of days, but they’re generally doing well.

The fundamentals are very sound, and I would say the evidence of froth is minimal. Recent performance is driven by EPS growth. Multiples have contracted slightly—maybe more than slightly if you’re a SaaS company—over the last few days or couple of weeks. But I would say the market is priced, in general, on earnings and earnings growth. Earnings multiples are higher than average, but nowhere near the dot-com bubble, adjusted for margins. You can look at the charts and see where we are, and that gives me some comfort.

The earnings of the companies that are the biggest drivers of the market, in general, are pretty sound. The companies are good. The health of these companies is pretty good, and the valuations are higher than average compared with the past, but they don’t feel super alarming. I often say that the leading tech companies I was just talking about are the best businesses in the history of the world. If you just look over a long period of time, they have shown margin improvement that suggests that’s probably true. That’s on the left side of the page.

Investors are paying for profits, not loss-making growth, and that’s a big contrast from the 2021–2022 era—sort of the 2021 era—and obviously a big contrast from the dot-com bubble, adjusted for margins. Multiples are not that high. To summarize 5 slides’ worth of material, the market is higher than it has been in the past, but I think there are high expectations for a reason. We’re optimistic about the impact of AI flowing through to earnings overall in the public markets in the coming years.

I’d focus your attention on the right side, which is a 4-box: low growth, high growth, low margin, and high margin, pairing up those types of companies. This chart shows how they trade. There’s a premium for the best companies. What you see in the 2 columns on the right is high-growth, high-margin companies and then high-growth, low-margin companies. Your bad box is obviously low growth, low margin, and those companies shouldn’t be rewarded. They should trade low, and they do.

But the companies that are high growth and high margin—and high growth and low margin, as long as they have good unit economics and they’re scaling into their margins—should be rewarded. I think this is good. If you’re not high growth, even if you’re high margin, it’s tough out there. That’s not surprising. I’ve talked about this in the past in many different forms, but ultimately, growth is the biggest thing that drives returns over 5 to 10 years. It’s nice for me to see high growth rewarded more than low growth. But if you have high growth and high margin, you’re one of those great businesses, and you’re being very rewarded.

This is just like what we’re going to talk about on the supply side of the capex buildout. The buildout is massive, and the size and concentration of the investment are inherently risky, given how big it is. While it has some bubbly features, the underlying fundamentals bear little resemblance to previous bubbles. The investment is financed primarily by historically profitable companies—very profitable companies that I talked about.

Debt has started to enter the picture. Cycle times have accelerated, which is good, but we’re closely monitoring the cost of training and the economics of that whole equation. Right now, the paybacks for the big model companies that spend money on training models are pretty good, but we’re monitoring that closely.

Most importantly, we think that AI is going to be the biggest model buster that I’ve seen in my career, certainly. I’ve written about model busters, so I won’t spend too much time on them, but they’re companies that grow faster and longer than anyone would have modeled in any scenario. The iPhone is the classic case of this. If you take consensus models from before the iPhone to 5 years later—4 years later—consensus models were off on Apple’s performance by a factor of 3x over 4 years. This was the most covered company in the world at the time.

I think the same thing is going to happen in many pockets of AI, where the performance massively exceeds what any expectations in a spreadsheet would show you.

So, tech in general is itself a model buster, but since 2010, tech has delivered high-margin revenue at unprecedented speed and scale. So it often looks expensive early, but repeatedly surprises to the upside and creates value far in excess of the capital that's required to grow. I have no reason to think it'll be different this time around.

So, relative to the dot-com era, capex is actually supported by cash flows, and capex as a percentage of revenue is considerably lower. So that's the simple headline. We can zoom to the next slide, but I feel much better about this capex dynamic than I do about the dot-com-era dynamic.

Obviously, hyperscalers are the ones who are bearing the biggest brunt of the capex, and this is a very good thing for our portfolio companies. This is great. I am all for it: get as much capacity in the ground and get as much supply as you possibly can on the ground for training and inference. This is a very good thing.

Again, the companies that are bearing most of the brunt of this are the best businesses of all time that I had talked about before. So, one thing that we're starting to monitor is the introduction of debt into the equation. You can't finance all of the forecast capex that's to come with cash flow, and we're starting to see some debt. So we're following this closely.

We're generally not invested heavily in companies with exposure to debt. Do I feel comfortable with a bunch of the companies on the page financing with cash flow, continuing to produce cash flow, and using debt, even with Meta, Microsoft, AWS, and NVIDIA as counterparties? Of course. I feel great about that. I mentioned the ones I feel great about. I don't feel great about all of them, so not all counterparties are the same.

We're starting to see private credit get a little bit more involved in the data center buildout. Again, the company that's very well covered that is making a bet-the-company move into becoming a cloud is Oracle. They've been profitable forever and reducing their shares forever, but the amount of capital that they are committing is very large. It's a big bet.

They're going to go cash-flow-negative for many years to come. If you follow some of the buzz around it, the cost of their credit default swaps has gone up to around 2% over the last 3 months. So we're watching things like this. Again, this is all generally good stuff for our portfolio companies, but we want to make sure that the market overall is healthy as well.

This is just a slide that shows the magnitude and the pace of change of AI, comparing the AI buildout and AI revenue to what happened with Azure. AI revenue is coming along relative to the cloud. It took Azure 7 years to reach one year of AI revenue. This is just Microsoft-reported data, which I think is a cool way to frame how quickly this has happened.

The build took a very long time. Again, this AI buildout is happening much faster, but it took 10 years for Azure revenue to surpass its capex. I think that sort of ratio or equation is going to happen much faster with AI.

We don't need to geek out too much on depreciation, but this is one of the topics that gets a lot of buzz in finance circles: What are your assumptions around depreciation of chips in particular? I would say the pricing for older GPUs is very solid. Early users stick with models a bit longer, but later users quickly switch to the new thing. So that's the right side; that's the model side.

On the chip side, 7- to 8-year-old TPUs—Google actually disclosed this—actually have 100% utilization. We very closely monitor the price of chips in the secondary market, and the price to rent A100s and H100s has actually held up very well. So older generations of chips are still getting fully utilized. This is not something I worry about yet, but it gets a lot of buzz from alarmists who like to talk about risk in the system.

All right, some positive stuff. The big thing that we talk about all the time is this paradox: as tokens get cheaper, consumption goes up. All the hyperscalers report that demand is well in excess of supply. I believe them when they say that.

I interviewed Gavin Baker, a friend of mine, at our AI Summit, and he was comparing the buildout of the internet and laying all the fiber to the buildout of data centers here. His big line was, “There are no dark GPUs.” There was dark fiber: You had to lay fiber, and then it lay there dark and wasn't used. If you put a GPU in the system in a data center, it gets fully utilized immediately. So that's a very good sign in terms of demand meeting supply immediately.

I mentioned this earlier: Earnings growth should come for these companies. This is our expectation. If it doesn't, then they will probably be disrupted if they can't change. Change management, again, is the biggest reason why we see things that haven't dramatically shifted yet.

Honestly, to me, it's not the readiness of the technology itself. It's probably product buildout that needs to get built around the technologies, and then change management and putting it in production. So revenue growth has scaled at a staggering clip relative to other categories.

This shows how quickly generative AI in-app revenue has grown from 2023, when it was basically—you can barely even see it on the page—to now. This is a slide that we've shown before, but basically this compares the clouds, public software companies, and how much net new revenue gets added in 2025.

The far right is what I like to look at: Public software companies added $46 billion of revenue in 2025. If you just add up OpenAI and Anthropic on a run-rate basis, they added almost half of that. And I think if you were to do that same comparison for 2026, the entire public software industry—I mean, SAP, this is not just SaaS, including SAP and older software companies—I think the AI companies, the model companies, will be something like 75% to 80% as much.

So it's just staggering how quickly that has happened. These are pretty detailed slides, these next couple. These are slides showing what is implicitly expected in AI performance based on where stock prices are today and in analyst models.

Goldman Sachs estimates $9 trillion of revenue flowing from the buildout of AI. So if you assume 20% margins and a 22x P/E, that translates into $35 trillion of new market cap. There's been about $24 trillion of new market cap that's been pulled forward. Now, we could debate if that's all attributable to AI or otherwise, including large-tech performance, but there's still a lot of market cap to go get where you could have upside if those assumptions are right.

This is another cut, or a few cuts, trying to address the AI payback question. Current estimates put cumulative hyperscaler capex at a little less than $5 trillion by 2030. So, if you do napkin math on that, to achieve a 10% hurdle rate on that $4.8 trillion, or almost $5 trillion, of investment, annual AI revenue would have to hit about $1 trillion by 2030.

To put that into context, $1 trillion would be about 1% of global GDP to generate a 10% return. It's possible that happens. It's also possible we could fall a little short of that. But I think it's limiting just to look to 2030. I think the payback of this probably happens over a longer period of time, between 2030 and 2040 as well.

But framing it up, that's about 1% of GDP to get to the payback number of a 10% hurdle rate. All right, “Heard it on the street.” What we've started to do is build software to track what all of the AI—or what all of the tech public technology companies—discuss in their earnings calls, mentions of AI, and how relevant it is to our business at the early stage and the growth stage.

Then we package it all up and share it with our CEOs, so they can have a simple, digestible format of what they need to know about AI as it relates to public technology companies, how it impacts their business, and so on. We shared a bunch of the stuff that we track here.

Awesome. There was one question before we moved to the private section, which a lot of folks on this call care about. Before we get to that, where are we calibrating to your trillion dollars in AI revenue, thereabouts in 2030? Where are we today relative to your guesstimate of AI-enabled revenue, and how far off are we from that trillion-dollar number?

We're probably in the—I would probably guess—in the $50 billion range.

Yep.

Just add it all up. And there's no perfect way to do it. I know some of the big inputs.

Yeah.

The harder stuff to track is honestly the big tech companies. How much real AI revenue do they have? The clouds can, from time to time, give a percentage uplift from AI, but depending on how they want to paint the picture, they can play games with that a little bit.

So I think that's a rough swag, but we're probably at $50 billion. It's growing way, way, way faster than 100% year over year.

Yep. And then arguably, that revenue—I mean, ChatGPT launched 3 years ago, but substantially most of this traction happened in the last year and a half or so, if we're being really generous, too.

Is that a fair characterization?

Yeah, that's right. Yeah.

And look, it's not just ChatGPT now on the consumer side. Google has a business, and xAI has a business.

And then, on the B2B side, not only do the big model companies all have large API businesses, but the clouds have them too. A lot of the sales that are model sales are also flowing through the clouds.

Yep. Okay, cool. We have some questions on the private-company side, but I'll let you get through the section and then I'll tee you up for it.

Well, I'm happy to go into questions if you want on it. A lot of the stuff that we've talked about—the big themes for me on the private-market side—are that companies are obviously staying private longer, but this is such a real asset class now. Over the last 20 years, the number of public companies has been cut in half. The vast majority of companies with $100 million-plus in revenue are private, something like 86%. So that's a major shift.

You can skip a couple of slides forward. I'll talk a little bit about power laws because that's interesting, and maybe some new stuff that we haven't talked about as much. Value very much concentrates in the outlier companies. The collective valuation of North American and European unicorns is about $5.5 trillion. The 10 largest ones, if you just take those, comprise almost 40% of the entire value. And that's actually doubled since 2020. So value is being concentrated in the biggest and best winners. I'm trying to count in real time: 4, 5, 6, 7 of the 10 are portfolio companies. We've got a reasonable amount of coverage on that.

Power laws are happening in the public markets too. Large-cap has tripled since 2019. So what constitutes a large-cap company has actually tripled since 2019. And I think the chart on the right side is super interesting. This was new data analysis that we had done.

If you look at the lifespan of an average company on the S&P 500, that's what the chart shows. That's what the numbers represent: once a company is on the S&P 500, how long is it on there? On average, over the last 50 years, that amount of time has declined by 40%. So disruption to companies happens faster and faster and faster, which I think is a very interesting dynamic and sort of matches what we're seeing in terms of the speed of change in the markets driven by technology.

So we always like to talk about power laws in our business too. I didn't choose the title of this slide. [Laughter.] I recognize all of the questions and concerns about it. The volatility-laundering thing is a big debate in our circles too, mostly around founders who are trying to debate the merits of the private markets and the public markets.

The Collisons did an interview—I think maybe it was John—where he talked about managing your stock price and avoiding volatility. You can, in an orderly fashion, bring your stock price up over time, and that makes it easier to retain employees, hire employees, manage morale, et cetera, et cetera.

I get the merits of that. I also think there are really strong merits of being a public company as well. I think we're going to have a really interesting 18 months where we're going to have some of the big private-for-a-very-long-time companies go public. And that's a good thing, in my opinion, too.

Some of the stuff that we show in this chart is just volatility and the observation that, over time, volatility has gotten a little bit more extreme in the markets. To me, this is a little bit cycle-driven too. I know short duration is sort of what we're measuring, but there are merits to both. Companies can get much larger on the private side. We have embraced that new reality. I think it's been a big benefit to our business in terms of continuing to invest in these companies over time. But obviously, there's a path of being a public company and getting liquidity, which we care a lot about too.

Awesome. On that note, there were 2 questions I'll queue up for you here. One on Databricks: can you talk about their transition from being a pre-AI company to a fully embedded AI company and what that's been like?

Yeah. First of all, I think you need to look at leadership. I mentioned Tobi: the reason Shopify has embraced it is because Tobi has led from the top, and he runs the business with AI at the center. He performance-manages everyone to make sure that they do that.

Ali is the same. Ali is this unique blend of commercial terminator—I mean, we call him the technical terminator. You need to have a commercial instinct and understand the importance of the value-creation opportunity in AI, and then you need to actually be deep enough in the technology to know what to build.

It just so happens that their cloud data warehouse, or what they call the data lake, is actually a great place to have your data and run AI workloads on top of it. That was sort of a good place to be for them, and then they've aggressively iterated on new AI products. They have this new product called Agent Bricks, which we're super, super excited about. We think it's going to be really big and transformative for them.

They have the big AI-native companies all as customers. They have the technology, they have the low-cost technology, and a big thing that we look for when we're making investments in companies is who their customers are. I would far prefer the customers of our portfolio companies to be the modern-thinking ones—the DoorDashes of the world, the Instacarts of the world, the Ubers of the world—than the very, very old-school, stodgy companies, because that means that their technology is evaluated by smart technologists and they pick it.

The cutting-edge AI companies are all building on top of Databricks, and they have the chance to grow with them as they scale. It's also a really good validator that they have the right technology.

We'll close out here. Thank you, David, for taking us through that.