Bessemer on the $3.6 Trillion AI Startup Race
Newcomer PodNovember 17, 202600:23:5121.84 MB

Bessemer on the $3.6 Trillion AI Startup Race

Inside the $3.6 trillion AI startup race, Bessemer reveals the companies leading the next wave of artificial intelligence.

Byron Deeter and Talia Goldberg unveil the 2026 AI 100, a ranking of leading private AI companies created by Bessemer Venture Partners and Newcomer. They break down startup valuations, revenue growth, and where founders expect the most value to emerge across AI models, applications, and infrastructure.

The presentation explores the shift toward a mix of open and closed models, persistent compute shortages, and opportunities in robotics, healthcare, and cybersecurity. Goldberg also lays out the revenue needed to support the enormous investment in AI infrastructure.

Then, Deeter joins Eric Newcomer to discuss why Anthropic ranks ahead of OpenAI, how Databricks and Stripe fit into the AI economy, and why 2027 could bring a wave of AI IPOs.

[00:00:00] We released our first AI 100 on stage at the Cerebral Valley AI Summit. Newcomer partnered with Bessemer Venture Partners to rank top private AI startups in the world. You may know Bessemer from their Cloud 100. They had the top, most respected ranking in the era of software, and they are making, with the help of Newcomer and the Cerebral Valley AI Summit, the switch to AI. So some of these names you expect. Anthropic is number one, OpenAI. Number two, Databricks. Number three.

[00:00:30] But I think there's a lot of interesting finds in here if you're not following the space closely. It's a super important exercise, and it allows for a ton of analysis from Byron Dieter and Talia Goldberg at Bessemer. They presented on stage at the Cerebral Valley AI Summit, and then Byron and I sat down afterwards and had a wonderful conversation. It is wonderful to be with you all here today. Great to see you all. Thank you for having us.

[00:00:56] And in a few minutes, we're going to unveil for you the AI 100. These are the 100 most important companies in the AI world that are driving this next phase of innovation. Now, for 10 years before, we published the authoritative list of the cloud 100 companies that are leading the cloud wave and the predecessor cycle. As we know, these cycles compound.

[00:01:23] As we saw in the internet wave, they democratized access to information and commerce. The cloud wave did that for compute. AI is doing that for intelligence, which makes it larger than the prior waves combined. AI is doing that for us. And so, without further ado, here's the list.

[00:01:47] Starting from 100 up, you'll see that to make this list, the companies had to be valued well north of a billion dollars. In fact, the list is so strong that there were dozens of companies north of a billion dollars who didn't make the list. Importantly within this group, some stats. The average company raised a shocking $1.1 billion. Another element of this, San Francisco's back, baby.

[00:02:18] 58% of the companies on this list are San Francisco Bay Area based. The next closest, New York with 12. And a hat tip to one company who's still fully virtual. All right. Let's give a round of applause to these great companies. And feel free to check out the list at the QR code there in a little bit more detail.

[00:02:46] The list of companies that you saw there is worth $3.6 trillion. And the top 10 represent about 70% of that value. But it's not just the giants. The median company on this list is worth over $9 billion. Physical AI was the second largest category by value. And data and agent infra has the most companies by count. That said, as is common in our industry, the power law is in full effect.

[00:03:15] It's not just the top 10 representing 70% of value. But Frontier Labs, which are just 9% of the companies on this list, account for well over half of the value as well. And we expect this power law dynamic to get even more extreme. All right. Switching gears to how companies are valued today. At the peak of the Zerp era in 2021, the average Cloud 100 company traded at 34 times revenue. The AI 100 is at a record 57 times average revenue multiple.

[00:03:45] And that's with a very generous denominator because it even excludes all the pre-revenue companies, which for the first time there are a number on this list. Those are companies like SSI, Instinct, Mind Robotics. Their revenue multiples are effectively infinite. But we've seen how quickly these companies can scale and how much strategic value they have and how fast that revenue can inflect when it starts. But nonetheless, it shows the value of companies has grown pretty dramatically.

[00:04:13] So you might be thinking, wow, this seems really high. But the average ARR growth is 480%, which is more than 5x the average historical growth rate. So sometimes you think investors like we should pay a premium for that growth. But on a growth-adjusted basis, the AI 100 is the cheapest cohort in recent history, with an average 12 times growth-adjusted revenue multiple. So now Byron's going to switch gears and share a little bit of some trends that we've been seeing.

[00:04:42] So in addition to asking this group who submitted to include data and company information, we asked them to predict the future a little bit. These are the companies that are driving the AI wave, and so we asked them to complete some survey questions around what they're seeing and their outlook ahead. And because this is the group that's driving the future, we put a lot of weight in the results and their input into how they see the future. First question was around model usage.

[00:05:08] Perhaps not surprisingly, the meaningful majority of them are primarily using closed frontier models today. However, by the end of next year, the interesting insight is that number is expected to fall to 36%. Now, they're not fully shifting to open source, but what we see is this mix. Closed, open-weight, custom models, fine-tuned models, etc. There's a cocktail that's emerging.

[00:05:32] And what we're increasingly seeing people try to compose is the right combination, and the answer is both. Perhaps the least surprising stat in the survey was 85%. Not only is compute scarce today, but they expect it to stay scarce. Now, we're not sure who the other 15% are, and if you have any excess compute, come find me afterwards. But what we're starting to see is business models emerging, not only around neoclouds and inference clouds,

[00:06:00] but even GP rentals and all sorts of things around the scarcity of compute that remains. But if this is the least surprising stat in the survey, perhaps the next one is the most surprising, which is an 87% number of where people think value will accrue. Not the model providers, but the application layer ahead. Now, perhaps this is a little self-serving, because as we look across the room, the majority of you are targeting these markets.

[00:06:27] But it's an interesting insight that speaks to the depth of the list. And it sets the stage, we think, for what's coming in the years ahead. When we look at the app layer in the current list, you see, and admittedly, it's a broad definition in the AI era, but there's about $629 billion. And an average valuation for these participants of almost $16 billion, which is about double the non-application layer members of the list.

[00:06:53] And you see that this app layer spans all sorts of categories, HR security, legal, accounting. And we expect from the survey responses and from the emerging companies below this list, much more activity in the application layer specifically ahead. Now, as we think about leading into what some of those areas will be, Tali will preview the future. So where do we see a lot of growth ahead? One of the reasons that I am so excited to be in this industry and in this job today

[00:07:22] is because there are so many insane opportunities ahead in a few really important and very underserved areas. The next frontiers that we're really excited about are in cyber, in bio, and in physical AI. And interestingly on this list, cyber and bio are particularly underrepresented. I think there were just three companies in those categories. So consider this a request for startup. Bessemer is excited to fund more companies in those domains. And importantly, these domains are pretty capital intense.

[00:07:49] And they're requiring historic levels of capital, particularly in physical AI and in bio. But speaking of capital, there's no scarcity of capital right now flowing into the space. Six companies on the list have raised over $10 billion. And of course, OpenAI and Anthropic have raised an order of magnitude more. But even if you exclude those two outliers, the average capital raised is still almost $4 billion. And the median is over $1 billion. That's an insane amount of capital flowing into the market,

[00:08:18] into these companies, but also into the CapEx that's supporting them. And there's been a lot of fair questions in the ecosystem about what's the ROI on those investments and what's the ROI on that CapEx. Usually we end with a prediction. But today we're going to end on math. Because the math is our answer to this and why we think we are so, so early. Global GDP is about $115 trillion. And about $65 trillion of that is labor.

[00:08:45] AI is the very first technology that does some of that work. About $8 trillion of CapEx is going into the ground over the next five years between now and 2031. This is the largest build in history. But for that to math, you need close to $6 trillion of revenue to be generated annually on the other side. That sounds like a crazy big number, especially when you look at the software market, which today is just $1.4 trillion.

[00:09:13] So we need four times that about to generate that amount of revenue. What gets us comfortable is that you look at the labor. Less than 10% of that labor going into AI would account for that. So AI doesn't have to replace the workforce. It just needs to capture a small sliver of it to make that math work. So the other part of people equations are the founders. We all know that the people are driving this revolution. And an exciting part is that the next mafias are already forming.

[00:09:40] This is the talent flywheel that's happening in AI. 14 of the companies already on this year's list were founded by companies, people that came out of other companies on the list. Open AI themselves have eight, and you see many more. Now what this feeds into is this compounding effect. When we first did our Cloud 100 inaugural list in 2016, a decade ago,

[00:10:08] the sum total of companies on that list were just under $100 billion in value. Roll that forward this decade, 36x increase. The average company is now $36 billion, as Talia shared. And the total market cap is $3.6 trillion in this ecosystem. It reflects the staggering growth rate across the list.

[00:10:33] And as we shared briefly, 20 companies, 20% of the list didn't exist two years ago. So 24 months and under is driving this more and more. And as we look across this room and the attendees of this event, and we look at the emerging companies, we look forward to more of you emerging on this list in the years ahead. So please join me in a round of applause for this year's 2026 AI 100.

[00:11:03] And we look forward to joining you next year and revealing the new names, hopefully, from many of you in this room that joined this crew. Thank you. Thank you. Thank you. We're here at the Cerebral Valley AI Summit. You've just gotten off the stage presenting the AI 100. Thrilled to partner with you all on the list. Likewise, Eric. Thank you. It's a mix of objective revenue factors, valuation, and a little bit of your read and our read on how various companies are doing.

[00:11:32] Let's just reflect, I guess, on the top 10, right? Anthropic, obviously, is number one. And that would have been a surprise. I don't know how long ago, like two years ago, definitely a surprise. Now, I think, you know, the juggernaut. Yeah. You go back a couple of years and people were thinking of Stripe and Databricks and those companies, Canva, et cetera. OpenAI had leapt into the lead. They were the presumptive winner. And I think the surprise over the last really 12 months, but going back 18 months, it started to reveal itself,

[00:12:02] was that leapfrog effect building. That Anthropic and really their enterprise business, the API economy that they were building for tokens, has overtaken OpenAI. And they're the leader now. It is a private company list. Right now, there are zero public companies. Barely holding on to the private status. But they are, I mean, literally, we're a week away. So they slid it under the deadline if the IPO goes forward as scheduled. And so the inaugural list appropriately captures that energy and ecosystem. And you're an investor, Anthropic. Obviously, part of our role is to keep you honest.

[00:12:31] I think no one is really going to contest that Anthropic should be number one. How far out relative to OpenAI? Yeah, it's OpenAI number two and I think Databricks number three and then Stripe number four. So how close does that feel? That separate? At this point, the market clearly views that distinction. And so it's unambiguous. I think the market will also back it up over the coming week as they're both out financing. And we'll see the Anthropic IPO for the world to see here in a matter of days.

[00:12:58] And then it's really how we think ahead, how this list will reveal itself. The exciting thing is also the emerging list beneath that. So there were dozens of companies valued at over a billion dollars that couldn't make the list because this economy is so strong. The company, you know, bottom one just to make the list was well north of a billion. Simile, a company we love is leading. Mr. Irrelevant at 100. I know. You use that term, you're a sports guy. I literally stop my team from saying Mr. Irrelevant. Come on, it worked for Brock Purdy. It'll work for Simile.

[00:13:27] They're a great company. And that just shows the quality that, you know, that there are 100 on this list. And that's the depth of this ecosystem. What do you make of, you know, there were some, I mean, even Stripe, right? I mean, whether to include companies or not, you know, it's sort of, there is the AI filter. I mean, Stripe obviously acquired OpenDNS, which I think helped solidify that they take AI very seriously. But how did we think about who to include and who not to? Yeah, and that's part of the criteria is published online.

[00:13:55] And there is this definition of kind of AI revenue quality. And so, you know, SpaceX is even an example. As they first went public, you wouldn't have thought of them as an AI company. But with their cursor acquisition and obviously the data center revenue, if they were still a private company, we probably would have included them post-acquisition of cursor. Stripe in a similar way. You could also say categorically, where do you put them? Are they fintech? Are they app? Are they platform? Are they infra? They sort of blur a few lines.

[00:14:24] And so with some of the business evolution, as they're driving this agentic API economy, they straddle a bunch of those worlds and they've reaccelerated. So a huge hat tip to the Collison brothers and to that company for helping to lead not only the cloud wave and that API economy, but now the agentic AI wave and to kind of get their mojo back. Cloud, I was waiting for that word. You, for a while, were synonymous with the cloud wave. Obviously, this list is an evolution of the cloud 100.

[00:14:54] Yeah. What made you decide to say, okay, have the companies changed or the brand changed? And how did you think about making that shift? Yeah, I've been part of the software industry for decades now. Cloud was the macro cycle of the last decade. And in 2016, we published the first cloud 100. Last year was the 10th anniversary of that list. And the inaugural value of that list was $100 billion, $99 billion.

[00:15:24] As you think about the list and the evolution of the cycles, we are now in the cycle that dwarfs the entire cloud wave. It's an order of magnitude larger. Literally, add a zero to what can be possible. Already, the value of these companies are 36x. The sum total of this list is $3.6 trillion. The mean value, therefore, is $36 billion of a company on this list. And that just shows that when you went from democratizing compute in the cloud wave to now democratizing intelligence in the AI wave,

[00:15:54] the markets are entirely different. And this is that next super cycle that we're going to experience in tech, in software, and in business. There was something that felt perhaps incorrectly solid about a cloud list where it's like recurring revenue. People knew how to value them. You look at the AI 100. They're much more valuable collectively. But, you know, companies we're interested in, like Merkur, where it's like helping AI training runs get smarter and smarter. It's like, will that be a strong business forever?

[00:16:23] They're just, and then obviously there are labs where some of them are almost pre-revenue. I mean, Prometheus. Some of them are. Zero revenue. Almost. It's just, I'm afraid to even say it. But you're like, Eric, they are pre-revenue. Literally infinite multiples of revenue. But I think you actually, you highlight the extremes very well, which is the neoclouds, the, you know, the training data sets, which have extremely high revenue numbers, but perhaps the market is questioning durability. And then you have these really bold physical AI, foundation models, et cetera, you know, bio labs.

[00:16:53] That are doing things that are valued in the billions. That have fantastic founding teams, but have zero revenue. And in some cases won't have revenue for a while, but deserve to be on the list. You know, an isomorphic as an example. And so you get in these, which does have revenue, but, you know, it's a high multiple in revenue. Or, you know, mind robotics or pie, physical intelligence, those sorts of companies. And so that's where market leadership, that's where some of those variables weigh in to how the market values them and how the list ranks them.

[00:17:21] Newcomer produced this AI sentiment report interviewing anonymous VCs. And one of our takeaways was that there was a huge amount of enthusiasm for robotics. You see that on the list here. That is obviously a category that's so far from the cloud era. What, I don't know, what's your personal level of confidence there? What has to happen for these companies to succeed? Yeah, so we feel that momentum building. You are reporting on it, I think, very correctly.

[00:17:46] And yet one of the negative surprises on the list was that we saw fewer physical AI companies, robotics companies, I'd say life sciences and healthcare, and cyber companies than we expected. Interesting. I think that's the future opportunity, not a statement on how it's going to play out. We think those three categories are among the most interesting in the years ahead, and I think those numbers are going to build. So we very much think that we will see more of the robotics revolution ahead.

[00:18:12] And personally, I think life sciences is the single area of AI I'm most excited about just because of its impact for humanity. I think that lives will be saved, outcomes will be changed, longevity will be realized because of what's going to happen over the next several years. What do you think of the persistent privates? You have Stripe, Canva, Databricks. I mean, great companies, but also we're talking about this as Anthropic is about to go public. Do you think that trend is going to change?

[00:18:40] Do you think it's the same thing where there are companies transforming from one generation to the next? I think 27 is going to be a really good IPO year. We've had a conspiracy of factors that have geopolitical, with market, and then with the mega IPOs that have just created this logjam unlike anything we've ever seen in human history. The fact that we can come up with a list of 100 private companies valued at $3.6 trillion is shocking. And that speaks to the quality of this shadow pipeline.

[00:19:09] No one wanted to challenge SpaceX or Anthropic. I think OpenAI will have that effect in the early part of next year if they do choose to go public. My hope and sort of my expectation is that by the middle of next year, the floodgates may open. And you could see a wave of really high quality companies that say, we're going to do this. And the market is so thirsty for them. They're struggling to buy chip stocks and data centers as their only way to play the AI macro trend.

[00:19:36] As soon as software companies are out there and there's a chance to do pure play tradable names, the public market's going to love them. And so I think that these companies will give them what the market wants and we'll start to see that. But you're going to have to get through these. A lot of what people can trade right now are the, you know, infra companies, you know, the core weaves. What, IREN, you know, what do you make of obviously the public backlash to data centers in the United States? Does that change the list?

[00:20:04] I mean, this list was mostly composed before this blue wave. Do you think the signs that the public, American public is so negative on data centers is going to have actual impact on what companies succeed? I mean, the first thing I'd say is I think a lot of the data center backlash is misplaced. And that will be litigated over time in public opinion, but part of it's marketing, part of it's politics.

[00:20:28] But for regions to turn away economic growth and jobs and vitality. Now, I think it's short-sighted. Of course, it needs to be regulated. Of course, things like power prices for the community need to be managed, et cetera. But I think there will be a settling. I don't think putting data centers in space because you can't do them domestically is the logical conclusion. There'll be some price, some tax, some agreement that's struck. Yes, and you're exactly right. It'll be a cost input.

[00:20:56] It will raise the cost of doing it, which is okay. But I think we will work through that cycle. But I absolutely think that the demand for intelligence is almost insatiable, that it will continue. And therefore, the need to provision compute is going to continue for a long time. Any observations about the great rivalries on the list? I mean, obviously, there's Anthropica and AI. But then in the more nerdy domain, there are the inference companies, Fireworks, Base 10, Modal, together to some degree. Absolutely.

[00:21:25] Databricks and ClickHouse. And you get these app, Harvey and Legora. I mean, it's fun. The prizes are so big. And I think what people are realizing is it's not zero sum. That in the prior software eras, you would have grill economics and oligopoly dynamics. But oftentimes, it consolidated to virtually one. We're not seeing that. These folks are beating each other up. But the markets are growing. And if we're thrilled to be investors in Anthropik, we're fortunate enough to get number one.

[00:21:55] But hey, OpenAI is a great business. And what Gemini's building within Google is a fantastic business. And what Grok and SpaceX have been able to do is a great product. And so there will be multiple winners. There will be trillions of dollars of value creation still ahead in these markets. And if you look at the application area and the layers on top of this enabling infrastructure, that's the wave that's going to build next. The craft of the list. How much do you want people to read this list?

[00:22:24] And you're like, these were the... It's almost obvious. Like, this is who it's supposed to be versus, oh, we wanted some surprises in there. Like, what do you try to balance? The goal is to make it factual, objective, and correct. And we hope for both as a byproduct of that, which is the discussions are always awesome. And companies, you know, lobbying and protesting. And why wasn't I on it? And then they look at the list like, oh, yeah, actually, turns out I get it. Like, I forgot how many great companies there were. And when you get this juggling of, you know, instead of being 35, I should have been 25 or whatever.

[00:22:54] And then they look at number 26 and 27 or whatever. And they're like, actually, you got me. I know. Well, even our contributions to the list, easier to say, let's move this up than to say, spike that. Then who are you possibly going to move? Yeah, sure. I get it. It feels like they should be top 10. It feels like every one of them should. But give me a name in that top 10 who you should bump to do it. I can't come up with it. And so it's a great mark in time, which is, of course, there will be movement in the list for next year. And that's one of the things I love is looking year over year at the changes, the new emergence, those sorts of things.

[00:23:23] But we will look at this list. And I feel very good in saying that the biggest acquisitions, the biggest IPOs are going to come from this list. And objectively, it captures today in history who the leaders were. Byron Dieter, Bessemer Venture Partners. Thanks so much for partnering with us on this. Thank you, Eric and the newcomer team. It's been a fabulous relationship. Thank you. Thank you. Thank you.