所有人仍然低估了 AI 市场的规模 | Eric Vishria

I love asking you and all your partners this every time we hang out: You’ve got these singular investments. You don’t each do that many investments per year, and the ones that go on to work—so far, Sierra and Fireworks certainly are—you get to learn so much about the world through the lens of the company.

I’d love to do both of those, maybe starting with Fireworks. What do you know, or what have you learned, about the world and how it’s reordering itself by watching the world through the lens of Fireworks that would be surprising or interesting?

1. AI Inference Is Hard

One really interesting thing is that these models are big. These are 2 trillion-, 3 trillion-, and 4 trillion-parameter models. It turns out running those models is damn hard, and running them efficiently is super hard.

The way to see this—and everybody can see this—is that everybody from AWS to Azure to GCP to the neoclouds to the Fireworks-based sent-togethers of the world, like all of them, they all run these stock open source models that are available in their developer portals and everything else. The performance difference between Fireworks and a cloud provider is 5×, and that is just the speed performance.

Then you add on top of that throughput, which is not visible externally, only visible if you know the economics of these businesses, and you’re like, “Wait a minute. This is the same open-source model with the same NVIDIA hardware, and there’s a 5× performance difference and a multiple-times throughput difference?”

The way to simply understand that is that these companies are paying the margins of the cloud providers, running on top, and making money. How can that be? My big takeaway was, “Wow, this stuff is actually really hard to run. It’s just really hard to run, and there’s a lot of expertise involved in doing that.” It’s this very specific expertise that exists.

It is the kind of thing that, when you look at it as an investor from the outside—and we should talk about the early days of AWS—but when you look at it from the outside, it’s like, “This is a commodity. This is just a pass-through resale.”

Scale game.

Yeah, scale game, like whatever. And you’re like, “Oh, wait a minute. No, it turns out it isn’t. It isn’t at all.”

2. Cloud Maps The AI Market

Do you think that’s just a moment-in-time thing? I’d love to hear you riff on cloud. You watch cloud very carefully and closely. You know a lot about it, and I’d love to hear about the adoption curve there versus how people use these things and the nature of those 2 businesses in comparison.

It’s tempting to say there will be 1 or 2 scale winners—

Yeah.

—like there typically have been in a commodity market—

Totally.

—where cost to serve is everything and scale drives cost to serve down, and that’s the whole story.

I think the AWS example is so good. In 2006, S3 and EC2 launched, right? Those were the first 2 AWS offerings in 2006. You can start talking about it in late 2006 or whatever. In the 2007 annual letter, Bezos talks a lot about AWS—why it’s important, why it’s interesting, and everything else.

The investor reaction just isn’t good. I think if you put 30 of the smartest investors at that time in a room and asked them, “What’s the probability that this AWS business is a good business with durable long-term margins, super interesting, and not a commodity?” I think you would have gone 0 for 30 with really smart people that you and I know who were around at that time in 2007.

Fast-forward from 2007 to 2014. I joined the venture business in 2014, and a really common narrative in 2014 was, “Oh, my God, AWS is going to eat everything. There’s no enterprise opportunity left. It’s going to eat databases and infrastructure, but it’s going to eat the apps, too, and they’re going to offer everything at the cheapest and best price.”

We all had these amazing SaaS and software businesses that we were involved with or investors in. Part of the reason people loved them was that they were annuities, ran at 85% gross margins, and everything else. It was just like, “Oh, my God, AWS—Amazon—can offer things at 8% gross margin, and they’re just going to crush this whole thing.”

You’re margining my opportunity.

Yeah, you’re margining my opportunity. That was the whole narrative. Think about it: From 2014 to now, in enterprise, of course you have Snowflake, a direct competitor to Amazon Redshift, running on Amazon. You’re out-Amazoning Amazon on Amazon.

Right.

But it’s not just Snowflake. You had Confluent, Elastic, MongoDB, and Databricks—all of these amazing companies. That’s the infrastructure layer.

Then you have the whole app layer. In the app layer, think about offerings that they offered at the beginning. Datadog is a $100 billion company today. They had a competitive offering, and they did it. Of course, there was tons and tons of roadkill. There was tons of roadkill. They did run over a bunch of stuff.

But even then, in 2014, the thesis—the view that AWS was going to eat everything—was massively wrong, not because of all the examples I just mentioned. It was massively wrong because Azure and GCP were irrelevant then, and fast-forward to 2026 and they’re unbelievable businesses.

Is AWS the biggest? I think it’s about a 40–30–20 split. You ended up with an oligopoly of those businesses, and even outside of those big 3, you have Cloudflare, which is a cloud provider of a different sort and another $100 billion company. So you have these smaller players that emerge as $100 billion companies outside of it.

What’s the takeaway? The takeaway to me is that there’s a bunch of zero-sum thinking and not realizing how big—

What if it all works?

What if it all works? It all works. Of course, getting the relative winner right matters. There was roadkill, and so all those things still matter. I’m not saying spray and pray. I’m not saying that at all.

But I’m just saying the market was so big that 1 vendor could not scale and consume it all. They just couldn’t consume the whole industry. It’s different now, like all these things, but that whole notion right now, with what you and I are seeing in AI, sure feels like it rhymes.

Anthropic’s going to do everything.

Right. Really? Is that really right? To me, that view that it’s just, “Oh, this 1 company’s going to eat it all,” doesn’t hold.

I would tell you that scaling—while cloud scaled very quickly, cloud did not scale anywhere close to as quickly as what’s happening right now—for that company to actually scale and deliver it, requires a ton of infrastructure buildup in this case: energy, power, shell, chips, memory, obviously the algorithms, and everything else on top.

It feels to me like we’re going to end up with an oligopoly of winners. I really believe they will be these $100 billion, crazy, smaller winners. It just feels like the same thing is happening.

Can you talk about it also from the demand side and compare it to how you watched cloud get adopted in enterprise versus how you’re seeing AI get adopted now?

If you go back to 2010 or 2011, you’re now 3 or 4 years into AWS being an offering.

Enterprises were super skeptical—traditional enterprise, blue-chip enterprise. You had digital natives out here. You had new companies forming that were using the cloud. I think, famously, Snapchat was built on GCP.

Mm.

And I think at one point in this era—maybe 2012, 2013, 2014—Snapchat was like 40% of GCP. Those kinds of things were happening.

That's like Cursor being 30% of all these things.

100%. 100%.

Same thing.

Same exact thing happened.

Yeah.

Same exact thing happened. Enterprises were very skeptical, I think, of cloud until by 2014, 2015, 2016, it was like, "Oh, yeah, yeah." Then I think the big banks and financial services and insurance companies and more conservative blue-chip enterprises were like, "Oh, wait a minute. This is actually different, and we're going to have to pay attention."

It became an issue in recruiting for them because they couldn't get the best developers. The best developers wanted to work on the easiest platforms, and all these kinds of things happened. So the difference now feels profound to me because, while it is 100% the case that blue-chip enterprise AI is not well absorbed and well adopted—and it's not the same thing as going to Cursor and walking the halls of Cursor versus walking the halls of a big New York financial services firm in terms of their AI use—they want it. They want to figure it out.

They're running experiments. They are spending against it. They're trying to figure it out and talking about it. They're not being dismissive about it. I think they view it probably as more of an opportunity than they did the cloud in terms of the potential impact on their business. I think they probably view it more as a threat than they did the cloud, and maybe they also learned lessons from the cloud in terms of what's possible here.

I feel that they will continue to try to figure out and absorb it. Having said all that, I do think one of the more interesting things is, if you go from Silicon Valley out into the world and talk to these companies—these enterprise companies—and you realize what the pace of adoption is and what the barriers for adoption are and everything else, there's a ton of opportunity to help the enterprises get there.

One of the things that I tell the companies I work with is, "Hey, let's be their AI Sherpa." If we're in that position to be their AI Sherpa, where we're crossing both worlds, that's very valuable, and I think that'll continue to work.

3. Sierra Builds For The Jagged Edge

What is Sierra teaching you about the adoption of this stuff? It's really interesting to contrast it with Fireworks, where Fireworks is the infrastructure provider. Sierra is feeling its way through what are really cool things that we can do for end consumers—enable companies to do for end consumers—starting with customer service, but I know with Horizon now going beyond that.

Again, same question as for Fireworks: What do you know that the world doesn't fully appreciate because of how you've seen that?

My partner Peter and Brett—I think this is their third company working together. It's a 20-year relationship, which is an amazing and great place to start. One of the things that I think makes Bret so special, and the team, is that they're technologists, but they have actually lived in the enterprise world for a long time and really understand it and everything else.

I think he's personified this whole idea of, "Hey, let's be their AI Sherpa. Let's start with customer service. We're very automatable. Let's be their AI Sherpa, and now we have these long-running agents with Horizon that can do more and more stuff."

One of the things that I think is most interesting about this to me is that, yes, it's an application company on the surface, but they're doing real AI work. They're experimenting with the models, they're building agents, and they're very close to the metal of the models, the capability, and the harnesses. They're understanding the jagged edge of AI capabilities, which is very different from the human smooth arc that we understand intuitively.

They understand that jagged edge of capability, and they build around it. You saw the evolution of Cursor from the IDE to tab autocomplete to agentic work, over and over and over again. They were obsoleting their work from 6 months ago. It's what Brett Taylor calls sandcastles. We used to be building castles; now we're building sandcastles, and then they get washed away.

Oh, that's great. Yeah.

If you don't think of software that way—

You've got to embrace that.

Yeah, you have to embrace it. If you're an artisan and you're like, "Hey, I built this perfect foundation of a castle, and it was just bricks, and it was perfect, and I really care about this, and it's going to be here for 100 years," you're just not going to make it.

As you get emergent new properties in the models—which is every 4 weeks—you get new capabilities.

Happening every 4 weeks.

Every 4 weeks, you get new capabilities. They understand that jagged edge. They understand the application of that jagged edge, or that valley, to their customer base. They're filling in those gaps and translating it. You can't just be superficially applying things.

I think it's actually a complete inversion of how product development used to work to how product development happens now. In product development, you'd say, "Hey, product manager, you trained. The product manager should understand the technology but really shouldn't be thinking about implementation and shouldn't be specifying implementation and shouldn't be doing this and doing that." The product manager's job is to really understand the customer and translate that problem to the engineer so that the engineers build a solution.

That was traditional product management. Good luck doing that now. That's a horrible way to do it. You can't do it that way. You really need to understand the nuances of model capabilities—what they're great at and where they fail—and you need to understand the customer problem and put those together and bridge those gaps to build valuable solutions.

I think that's one of the things that Michael and the team at Cursor did so well from the very beginning. They really understood the jagged capability and built a product that allowed that translation from developer to that jagged capability, and kept iterating on that as the edge changed.

It's kind of ironic that all that sounds like the returns to being technical are going up—

Totally.

—even as models supposedly are taking away technical edge. Funny conundrum.

Yeah, 100%. Actually, I think there's this weird thing where there was this whole discussion about the traditional roles of product manager, designer, and engineer, whatever it is. I think what there really are are people who understand customer problems, people who have taste, and people who understand the jagged edge of AI capabilities and are curious about it. Those are the 3 things.

If you've got those 3, you're going to do well.

If you have those 3 things, you're going to do great. It doesn't really matter if you are an engineer, a product manager, or a designer. If you have taste, an understanding of customer problems, and an understanding of the jagged edge, you're going to do well.

4. AI Rewrites The Competitive Frontier

When we first did this so many years ago now—which is crazy—it'd be fun to revisit some of the ideas that we talked about the first time with SaaS. You used this term that I've used ever since, which you called it the competitive frontier, meaning the things that will determine the winners and the losers.

My partner Bree has this great idea that stuck in my head, which is that everything is a jump ball right now. I'm really curious, in addition to this idea of sandcastles versus real castles, what else you're seeing among the people that are becoming competitive winners? Are the traits different for winners now—personality-wise, business-strategy-wise, or business-model-wise—versus what you learned in the SaaS era?

I'm very dismissive of this idea that people are going to vibe-code their own shit and whatever.

That isn't the issue at all with SaaS companies. The issue with SaaS companies is that the competitive frontier completely shifted, and everything they thought they were building against—and that would make them win—is not what's going to make them win.

Let's take databases. Databases have been a phenomenal area for software for a long, long time. They had great margins, which is why app developers would build against the specific database interfaces that existed for that database. You would have more and more data over time. Migrating an app from one database to another was a giant project that was very, very difficult to do. So these were unbelievably sticky businesses that you could generate a ton of margin in, right?

Of course, you had Oracle and SQL Server, and a whole slew of smaller players that did really, really well in databases. Well, let's think about that in the context of AI. First, you don't have a developer building against a database interface. You have Claude or Codex building against a database interface. Second, the beauty of database interfaces is that they're very, very well specified. It turns out AI is very good at things that are very, very well specified. Third, agents don't get tired of the monotonous work of translating one specification to another. So it turns out that database migration, which used to be the No. 1 thing you would not do in software, is kind of trivial.

It's just money. Yeah.

Yeah. It's just like, “Put some money against it and move it.”

So what changed? What changed is that the criteria to be an amazing database company changed. It isn't that we don't need databases or that everyone's going to build their own database. That isn't what's going to happen. What's going to happen is that the criteria changed.

Now you're going to have a ton more applications starting, obviously, as we're seeing everywhere. People are going to experiment a lot more because it's much cheaper to experiment. It's much cheaper to start a new application. So now you need databases that scale from basically 0 usage all the way through if it works.

Mm-hmm.

That matters a lot more. Your cost matters a lot more. You want to be able to spin these things up, spin them down, and tear them apart over and over again, so the iteration speed goes up and that's what you need in a database.

Ultimately, I think the cost becomes the arbiter of this. The cost, this 0-to-infinity scaling, transportability, and everything else around that become the arbiters of who wins and who doesn't. That's really different from, “Hey, I spec'd this database for our user thing, and I procured a license, and I ran it on this kind of hardware,” and everything else. There have been elements of these things over time, of course, but I think the criteria just changed.

I think one of the big messages to these SaaS companies a few years ago was, “You have a choice: get to AI or be worth 3× revenue.” That was a hard message to hear. You were just like, “Get to AI or 3× revenue. Those are your choices.” We say 3× revenue now because a lot of these public SaaS companies are trading at 6× or whatever. But keep in mind, in 2021, they were going for 30×. Everybody was like, “3×? Wow. I've grown 4× since then.” This whole multiple compression is a bitch.

Yeah.

I've grown 4×. The multiple has gone down by a factor of 6, so I'm worth less, even though I've grown 4× in a few years, gotten to break even, and done all these things.

That was the first message. But I think it became really visceral to me when you were in these meetings and you were like, “Hey, every single day that you are hitting your plan, you are destroying equity value.” Think about that. Our whole careers, we learned that you lay out a plan, execute against it relentlessly and violently, hit your plan or exceed your plan, and keep building that. That's how you build equity value. That's what the whole management team learned. That's what these CEOs learned pre-AI. This is everything we learned.

And now you're in here, and every day that you hit that plan—

You're fucking up.

—you’re fucking up. You're destroying value. The point of saying that to them was to set them free.

Another articulation of this from my friend Anne Lee Skates was about the CEOs who were going through this transitory period and had a business that was at hundreds of millions. They thought they were all ready. They were working on their business from 8:00 AM to 5:00 PM and then trying to do AI from 5:00 to 8:00 in the evenings, and what they needed to be doing—

It's the inverse.

—was the inverse.

Yeah.

And it's so hard to do that because of all the training, muscle memory, inertia, and everything that we learned about the hill that we were climbing. We're all hill climbing in a way. The success model was set the plan, execute the plan, build value, compound value, and it's like, “Oh, no. No, no. Stop that. You've got to completely invert.”

It's a very, very long way to get back to your question of what the profile or mentality of the winners is right now. But if we look at Brendan from Recor, Lin from Fireworks, or Max or Brett—any of these people—they are so nimble about what the eval is and what they're optimizing against. They're so nimble on all of it.

If you look at the evolution of the business of every one of those companies, the business is just constantly evolving, and they've done such an excellent job at that. I think that is very, very different from the way I was taught.

What's your sense of the disorienting nature of model progress? You're one step removed from that as an investor versus being the technical founder with your hands on the metal. How are you behaving differently than you would have 3 years ago because of the pace?

Any time I'm talking to a founder about a problem in their company, what they're doing, a move they're making, or anything else—which is what I spend 80% of my day doing—I'm very much, “This is how we used to do it. This is what we would typically do. This would be the typical readout, this old-school readout, of why this candidate is better than this candidate.”

Let's reevaluate that in the context of today. Let's reevaluate that in the context of an unstable technology substrate. Let's reevaluate that in the context of a business model that's growing this way versus that way. I've really started to question every assumption and every lesson that I learned before: which of it translates and which of it doesn't. That's a huge difference.

I'll give you a really concrete example, which has been very disruptive inside of these scaling AI companies. We've had a lot of leaders from the prior generation, with great experience and everything else, come to these scaling AI companies and completely flame out, and you see it across the industry.

So the question is, why? These are some of the best leaders from 4 or 5 years ago. They had all the lessons; they learned it all. They're excellent, but somehow it's not translating, and there's some impedance mismatch between the AI founders, potentially, the needs of the business, and what these people are bringing.

I saw it really abruptly with a particular sales leader we hired who said, “Hey, we can't hit any of these things because the way software sales has been taught forever is a quota-capacity model.” You have a quota-capacity model: each rep does this. In the early days of a company, the quotas are 1.2 or 1.5 million. Maybe the ISRs are at 750 or 850, and then over time it scales up, and enterprise gets to 2.5 million. That's how all these financial models are built. You start with a quota-capacity model, take a discount on attainment, and say, “This is what we can do.” Boom, boom, boom.

Fundamentally, without realizing it, everybody was implementing something that was based on pushing demand, not pulling demand. For so many of these companies, they're operating here with these customers, and there's a new AI-enabled product that comes along, and it's just fucking magic. These companies are selling magic.

Well, it turns out that if you're selling magic and you're the first one there—

You sell a lot more than 2 million.

—you’re going to sell a lot more than 2 million. The whole notion of a quota-capacity model and having that work—it's not that it doesn't matter. It matters kind of, but it's definitely not the first-order thing or constraint.

So you have these executives come over, and it's just like, “Here's our math, here's the territory assignment, and here's what we would do. First, you do the West Coast, then you do the East Coast, and then you do Central.” All of these things. Oh, wait, no, no, it doesn't work like that at all.

One of the big things I started to realize as I interviewed these folks and talked to them was, “Hey, you need to check everything at the door. Check it all,” which is probably good practice anyway. Check all the baggage. Check everything that you learned, and just learn this from first principles. How is it working? What really are the bottlenecks on delivery? What are the bottlenecks on demand?

It turns out that in a lot of these companies, you have reps doing 10 or 20, 30 million.

I saw 50 recently.

It turns out that's different.

What is the best salesperson that you've seen who's doing it in a de novo way? What are they doing?

Honestly, the best salesperson in any of these companies is the founder. What they're doing is bridging the jagged edge to what the customer's capability is, and that's it.

It sounds so simple, but it’s not. But that’s what they’re doing. The market is just so big. Just like we were talking about with cloud, I think the biggest mistake everybody made was that they undersized the market, and it turns out the market’s just really, really big. This market is bigger.

It feels like, right in this moment in time—just this morning—you and I are in this great group chat together where the discussion is that the demand for intelligence seems unlimited. It seems like the smarter the thing gets, the more demand there is. Maybe the bottleneck is just capital. The world just feels like it needs to take a breath for capital to form and evaluate its prospects, and the scale is getting so big.

The RSI concept, if you apply it across technology, doesn’t need to breathe. The agents don’t get tired. But the world feels kind of like, “Oh, man, it sure would be nice to have 3 months just to digest this a little bit.” Does it feel to you like this can just keep going, or are we going to get tapped out of money that can be invested in these things, when it seems like we could consume any amount of money to build any amount of stuff and serve any amount of inference?

I'm not a macro economist. I am worried about energy. If you think about the models as translating compute into intelligence, quite simply, what do models do? They very effectively translate compute into intelligence. What's the demand for intelligence? Well, it seems like a lot. So then it follows that we will continue to have more and more demand on compute. I'm saying compute broadly, not chips, not storage, not whatever. But then, like, what do we need for compute? We need energy, a lot of energy. There's all this topic of distillation and Chinese open source and all these things, but the bigger thing to me is that I think China's bringing on ten times as much energy next year as we are in the US. If energy is what you need for compute and is the bottleneck, and there's unlimited demand for intelligence, then it stands to reason that if we have a lot less energy, then we will have a lot less intelligence or a lot less tokens or a lot more expensive tokens. If we have a lot more expensive tokens, then supply and demand, you're gonna end up with less, and that seems very bad. To me, I think the energy bottleneck, however that is, it'll manifest in twenty different ways. Gas turbines go up and down, natural gas go up and down, and solar, whatever, all these different things, rare earths. That to me is probably more concerning, and if I'm thinking about it from a regulatory perspective or government perspective, and I think the administration is doing some things around this, fostering investment and development of all energy, solar, nuclear, gas, like, whatever, do it all. We should do it all, and it'll work itself out. This is one of these things where, yes, one will be relatively better than the other, and I don't know, and I'm not smart enough to predict which one's which, but it'll all work.

5. Cerebras Exposes Hardware Complexity

Speaking of compute, I would love to hear the Cerebras story. I haven’t heard you tell the full version of this. The reason I’m asking about it is that I’m deeply interested in compute. I have big investments in compute, and I’m fascinated by it. It’s just the most magical thing to watch happen. It’s mind-boggling when you get close to one of these things to see what humans have been able to do on these chips and in these systems.

I think you invested in 2016 or thereabouts. I think it was your first foray into extremely difficult hardware-type investment.

Shit’s so hard.

And now the world is full of opportunities like this, whereas back then—

Yeah.

—it was a one-off. Teach me everything you’ve learned about hardware investing through Cerebras.

Mostly, it’s really hard. It’s an amazing example to me of the naivete required. The company came in in 2016. It was 5 founders and a deck. I did not want to go to the pitch, but it was my job because I was thinking, “Why are we going to make a hardware investment? This is crazy.” It had been 10 years since we’d made a semiconductor investment.

The team was excellent, and the first slide was, “GPUs actually suck for deep learning. They just happen to be 100 times better than CPUs.” You have to remember, this was pre-transformer. OpenAI was this weird research lab at this time. NVIDIA was worth something like $40 billion, not $4 trillion. The TPU hadn’t been announced—none of that.

Mm-hmm.

So this is early. But the whole idea, as soon as he said it, I was like, “Oh, shit. Of course. Of course. Why?” I had spent the last 18 months trying to figure out applications of deep learning, looking at the security thing, looking at this medical-imaging thing, and looking at all this other stuff, thinking, “Hey, there must be something here that’s going to be really transformed by this stuff.”

Anyway, we go through this whole journey. We end up investing, which was amazing. We first met on Wednesday, had a partner meeting on Monday, had a bunch of meetings in between, and built a lot of conviction that this was a great swing. I’ll tell you what I understood: I understood so little. But basically, there are 3 things that we know how to do to speed up deep learning in hardware, still to this day: increase the number of cores, increase the communication between cores, and bring the memory closer to the compute.

Mm-hmm.

That’s it. Those are the only 3 dimensions that we know in hardware. My articulation of what they said to me was, honestly, all I understood was, “Let’s just take all 3 of those things to their logical maximum.”

You have a wafer-scale chip. At that time, you would have 450,000 cores on it. You’d have something like 20 GB of SRAM on the chip, so you never have to go off-chip to get to memory. Because they were all on the same wafer, the communication between cores is maximized. So this is the best you could do on that process. I think the first chip was seven nanometer or something. You’re like, “Okay, that’s it. That’s what we do.”

And it turns out that in software, if you have that logical block diagram of why it works and everything else, you’re 80% of the way there, and it’s a matter of go-to-market execution. In hardware, you’re like 2% of the way there. There are things like physics and an entire supply chain of vendors. There’s, of course, TSMC, which everyone knows, but it’s not just TSMC. There are 30 other—

So many.

—vendors that matter in putting all this stuff together and everything else. I didn’t know any of that.

Fast-forward from 2016 to 2019, and I think they got their first parts back. You go through this bring-up, and then it’s like, “Bring-up. Oh, yes, we got a part back.” And then it’s like, “Bring-up.” Then you have to go through, “Wait, bring-up,” and there are 14 steps of bring-up and everything else. By 2020, we had our first thing that worked. Then it’s just this march of actually getting it to work.

One of the lessons that I’ve learned on this stuff is that you go through all these sims and everything else in hardware, and semiconductors in particular. That is basically your roofline. The best it’s ever going to be is what that is. Then every bit of software in reality—compilers and kernels—takes away from that roofline.

Yeah.

You might start at 10% of the roofline. Once you bring it up, these guys are grinding for months and years to get closer and closer and closer to the roofline.

Mm.

It’s really different, and it’s really hard. I’m astonishingly bullish.

If I rewind, part of the reason we made the investment was that, if you looked at the 4 prior generations of compute in my lifetime, you had CPUs, graphics, networking, and mobile. There was a new workload each time. You had multipurpose compute, and it led to the CPU. You had massive parallelism, which led to the graphics processor. The graphics processor offered massive parallelism, which led to graphics. Then with networking, you needed really low-latency chips, and so you had low-latency chips. With mobile, you needed really power-efficient chips.

In each case, we ended up with a new $100 billion company. The first question, going back to 2016, was: Is AI that big a new workload? There had been many, many other attempts at specialized chips for other things that really just didn’t end up mattering. There were some fine outcomes, but they just didn’t really end up mattering.

Like Warren, et cetera.

Right, exactly. It was just like, well, okay, you need something that’s a really, really big workload. We had a lot of conviction on that. Then the second question was about the nature of the workload: Did it introduce a new constraint or problem?

What I learned was that the AI workload benefited massively from the parallelism of GPUs, but GPUs didn’t solve the core-to-core communication problem—the layers of the network problem. You’re like, “Okay, wait a minute. There is a new constraint, which is communication. It’s a communication-bound problem.”

Then you’re like, “Okay, is AI a new giant workload that is going to have specialized chips?” Everything that I just said was everything I knew at that time.

Mm.

That’s obviously played out. In each prior generation, we got Intel, we got NVIDIA, we got Broadcom/Avago, and we got Qualcomm and Arm in each of these generations.

And there will be these giant winners, standalone winners. Obviously, the TPU itself is a winner. Training is a winner. You've had Groq and Cerebras, Etch. It'll keep getting fought out, but I think that will end up being big.

And actually, I think there's a new sixth one that's coming. I'm really excited we've made an investment that's unannounced in this, but I think that, for the first time in a long time, there's actually room for a new CPU approach.

The thing that's happening right now—and you see this reflected in all the semiconductor stocks and everything else—is that the LLMs, which are running on accelerators and GPUs, generate code. The code runs on CPUs. And right now it's running on classic CPUs we've had around forever. But there's a whole bunch of constraints on CPUs that have existed, and CPUs have dragged all this baggage forward that you might not need anymore. So I'm actually really excited about that possibility.

The next category.

The next category.

Does the experience with Cerebras make you want to do a lot more investing in companies?

Fuck no.

But why not?

In 2019, we're sitting in a board meeting, and this thing is melting. It's like fucking melting, okay? And we've raised $500 million or something, and it's like, wait, what? It's melting or it's burning or something. We're looking at it, and I'm like, “Holy shit.” I realize $500 million isn't that much in today's era, but I was just like, “We're going to lose all this money. This is not going to work.”

What that team did was insane. They're built differently, and I have so much respect and thanks to them for what they've done. There are efforts that make you really proud to be a venture capitalist because you're funding something that makes a difference and matters. I'm an investor because that's a means to work with companies, not because I fundamentally love investing or something like that. I like working with companies. That's my favorite part of it.

Working with teams like that and companies like that is so special on these giant, ambitious efforts. And I said this well before 2018 or whatever: whether Cerberus worked or didn't, I think it was an effort that was worth venture capital. That was the kind of thing you should do. You should try to build something that people have tried for 50 years and have been unable to, but now we think we could do, and there's a reason and application for it and everything else. I love that.

I do like those kinds of things, joking aside, and we do have a robotics company and then this CPU project that we're talking about. I think these kinds of things are actually really fun and interesting and a good use of venture capital, but they definitely aren't easy.

The productive naïveté that you described, where it's probably a virtue that you didn't know more than you knew. Otherwise, you wouldn't have done it. This is certainly my experience with H [?]. In any of these fields, you ask experts, and they're going to tell you, “Don't do it. It sucks. It's too hard. The base rate's too low. Young people can't do it.” The 40 separate reasons.

All these things, yeah.

Is there anywhere where that is just a bridge too far, that could be biotech or something like this, where you just are unwilling to invest if you're naïve?

What's so funny is the hard thing about investing is that most of the time, all of these stereotypical statements are correct. They're not correct three times out of four. They're correct 19 times out of 20. Maybe 99 times out of 100, they are correct.

The thing that Bruce, one of our founders, always says is, “What could go right?” We have to ask ourselves, “What could go right, and do we see that path?” Young people can't build chips, or you shouldn't do a new semiconductor company, or you shouldn't do this, or whatever. All of that stuff is actually totally right, except when it isn't.

Apparently, there's a saying that someone said to one of my partners, which was, “If it doesn't work, it'll be for all of the reasons that your partner said. If it does work, it will be because those reasons didn't matter.” It's just such a good example of this whole thing. Yeah, most of the time when we lay out the reasons that a company won't work, it's right. But then sometimes they just don't matter.

6. Robotics Needs A Data Flywheel

You and I are both interested in robotics. It's not controversial that if robotics works, it might dwarf what we're currently living through. What do you think has to be true for it to work? Obviously, it's exciting. I want a robot in my house folding my laundry. It sounds great. It's one of these classics that's always 10 years away, and it's been that way for a long time. What do you see happening? What has to happen for this to actually be a thing in the near- to medium-term?

We've had classic robotics forever. And they're all over the place, and they're on assembly lines and manufacturing lines and all these things where you're doing repetitive tasks in controlled environments. Repetitive tasks in controlled environments are more or less solved, and that'll continue to happen.

But having unstructured tasks in real-world environments, that's where you need the AI. That's where you need the AI plus the robots. The trick with it all is you need a model that can do that.

Hmm.

Of course, the first problem is there is no Internet-scale data to bootstrap the whole thing, right? LLMs are all bootstrapped on the Internet, which is a ton of human knowledge. And the equivalent of that for robots doesn't exist.

People are trying different things with videos and simulations and teleoperation, so there are a lot of different ways. But if you think about teleoperation as an example, how much do you need to teleoperate robots to get to Internet-scale data? This first step is getting a good set of data to bootstrap the model.

One of the insights that you can have is that, with the Internet, there's a lot of slop data. Even before AI generated all this stuff, there was also a bunch of junk data, and some data was more valuable than others. You may value certain things on Reddit more than other things. You might value Wikipedia more than other forums. You might value GitHub more than other things. And all of the model companies did that, right? They prioritized data that was more valuable and less valuable and ran that through the model.

I think one of the most interesting things that these AI robotics companies are doing is saying, “Okay, well, let's just go after the high-value data to start. If we go after the high-value data, then we can bootstrap this model.”

Now, once you do that, can you, through the pre-training process, get to a place where you can have very small auxiliary examples of data that you add in post-training, and all of a sudden it works for that? That's the magic we have with LLMs: you have this giant pre-trained base, and then you add a little magic on it in RL and post-training, and you teach it a new thing that wasn't in the pre-training set, and you go from there.

And I think the exact same thing is happening in robotics. We're investors in Sunday Robotics, which is going after this exact pipeline.

It's a cool company.

It's a cool company, and they're doing household robots. But the key part, before you get to household and all that stuff, matters much less, actually, than whether you can get this training pipeline to work and how you do that.

One of the lessons I learned—and I look back, I try to learn from history because it doesn't repeat, but it rhymes—is that if I look at autonomous vehicles as an example, they're basically robots. They're AI plus robots.

You look at Waymo and you look at Tesla. They were both designed to be vertically integrated in their own way. You had a Tesla, you had a fleet of Teslas that everybody owned. It was collecting data on the Teslas, and it was used to train the models to drive the Teslas. The same thing with Waymo.

I think part of the lesson is they gathered very, very high-quality data. They did their pre-training, and then they worked off of that. That was a simplification to allow you to get to a complete product or a complete solution, which, of course, is going to continue to improve and ultimately will generalize. I'm sure it'll generalize in some way, so you'll be able to strap it on any car and everything else.

So I think the same thing is happening in robotics, where you have companies like Sunday and others who are using techniques where they're vertically integrating the robot and the model and the data collection around the robot. Sunday uses gloves that are designed with the robot hands, so they're perfect. So you get very good data transferability from one to another.

You do this pre-training and you have a great pre-training data set. You have a good model, and then you start adding these examples in your RL and post-training on top, and you get cool emergent behavior.

Do you have a most visceral moment?

When we invested in Sunday, the first time we saw them, partner Peter arranged a demo. We went down into the basement at a Stanford lab, and they had this totally janky cardboard glove thing. You see a few evolutions of it.

The last time we saw a demo, we went down in the basement of their new office building, and there were a dozen robots just folding arbitrary laundry. It wasn't in the demo; it was just trial and error. And then they had people taking the clothes and measuring them to make sure that they were folded properly and creating a rigorous baseline and evaluation criteria. And I was like, “Oh my God, this is happening.”

Do you ever worry about how to pick the right customer for these companies? Every one of these things folds laundry, which I don't think anyone likes folding laundry.

So it seems like a good use case, but it feels like we don't actually understand the demand for what these things will be used to do. Do you ever worry that we're building solutions that will then be in search of problems?

I don't, and I'll explain why. It's certainly informed by watching the LLM evolution. Some of the people who were involved in the early LLM work at OpenAI really understood that code was going to be important. Obviously, the Anthropic team had this perspective that you could get to RSI if you got code generation going and automating AI research and whatnot.

If you think about it, the first use cases were very much language-oriented. They were very much essay writing, editing, and marketing. I think the first application that really took off was Jasper, which was just writing marketing copy. I think it's going to evolve a lot.

Basically, I think laundry is a good task because it's arbitrary, it's complex, and it requires dexterous manipulation. It isn't time-sensitive. If it takes 3 times longer, so be it. Who cares? It doesn't matter. Just let it run all day. I think it has some of those properties.

I don't think the task is actually that important. What I think is much more important is whether you're pre-training an amazing model, then being able to post-train on top of it and get that flywheel going. If you get that flywheel going, then the task capability will just keep multiplying.

7. The Partner Comes First

Maybe this question will be annoying or slightly uncomfortable for you, but if you ask basically every founder and, critically, other investors of your type, almost everyone, if I ask who's the best board partner, will say, "You." You come up way more often than anyone else that I've come across, and I'm curious why you think that is—what it is that you're doing that other people aren't.

The incentives are there to do a great job as a board partner. What do you think you're doing on the boards of these companies, partnering with the founders, that's actually different from other really talented investors who are also nominally doing the same job but don't come up nearly as often when asked that question?

One of the things that I've realized is that we're each attracted to different types of entrepreneurs where we have chemistry. I'm an investor second, and I try to be a partner first.

We'll see companies come in. We had one come in yesterday, and it's what I would call an investment-grade opportunity. You can invest, it probably works, you make money, and it's good. An investor would do that. A partner wouldn't, because that's not sufficient for a partner—unless you have real chemistry with that person, where you feel like you're going to be able to work together really effectively, I'm going to learn a ton from them, and they're going to learn something from me. Together, we're going to feed each other's loops.

Unless you feel that way, you can't be a partner, and so you pass on that. That's a really important fit element to me. It starts with this mutual selection, actually, weirdly. They want to partner with us, and I want to partner with them, and I'm really looking forward to working with them together.

I'll give you a really good example of where this comes into play for me. If I take Saji and Benchling, Benchling is life sciences SaaS. The company has absolutely crushed it and done really, really well. Then, of course, you have this biotech crash and everything else, and the company became grindy.

The company had never had any churn for the longest time, such that even on their reports—for every SaaS company, you have gross ARR added, a churn line, and net ARR added; everybody does the same thing—they never had a churn line. They never reported it for the first 6 years that I worked with the company.

Then they got 7 years of churn in 12 months. It turns out life sucks when you get 7 years of churn in 12 months. Through that grind and through it all, and related to the whole, "Wait a minute, the goalposts moved. We have to do something different," they kept thinking about how to apply AI for these biotech and pharma customers, whom they're very close to. How can we make it better for them? How can we apply these models in their world in a way that they're excited about and continue to iterate?

It was grindy, and I was there for it. You're excited to work with that person because, one, of course you think it's a really special opportunity, and there's a way out, there's a path, and we can find it. But two, because of the joy of the game, the relationship, and everything else, that's part of it.

I've seen different models and lots of different models of venture capital work. Moritz was a writer, Dor was a sales guy, Gurley was an engineer, and Peter's a career venture capitalist. They're all different.

When I call these people, I learn something, and they push back on me. Then I ask them questions. What I've realized is that so much of my job is that they know the answer. They know what they want to do. They know the answer, and it's maybe asking questions of them to help solidify their conviction or solidify their articulation of what they want to do and why. You just keep doing that.

Through that process, hopefully we get 1% better a few times a year. We make a 1% better decision, a 2% better decision, a few times a year. If you do that over a decade, that compounds to real results.

One of the questions that I ask myself before making an investment is: There are all these people that I care about through my life, like you care about yours. Could I talk one of them into going to this company and honestly, intellectually honestly to myself, explain to them why this could be their life's work? If I can't do that, I should not invest. It just means that the project doesn't line up in that way.

As long as we have one of those things, it doesn't matter that much what it is to me. It's important. It could make a big dent, and if it can make a big dent and it's a special person, I'd love to work on it.

Are there any other questions like that that you ask yourself before investing? That's a particularly good one.

The other question is: If this person calls me at 9:00 p.m. on a Saturday night, will I pick up the phone? That—

Call that the green button test.

Yeah. It's a chemistry thing. They have to feel the same way, obviously.

One of the other ones is: If it's right, does it matter? Which is different than the first one, but there are so many things that we could be right on as a business.

I looked at one last week, and I told the entrepreneur, "I really think that you can build an amazing company here, and you just shouldn't raise venture capital." There are so many things that you can be right on, but they ultimately just don't matter. Nobody cares. That's a better way to say it. If we're right, will anyone care? If they won't care, then you're just not going to build enough equity value. That's another useful one.

8. Bigger Outcomes Expand Venture Capital

One of the coolest things that's happening right now is that all of what you just described has higher stakes and more leverage attached to it, which is—

Yes.

—manifested most simply in more dollars and higher prices. You and I have talked about this notion of what high-multiple-on-invested-capital investing is like and what it has been like and what it's moving into.

You did something recently, which was that you raised a growth fund for the first time in a long time. I think that is related to this concept. These companies need more capital. The prices are higher. The outcomes are bigger. Maybe we can earn the same multiple on a $1 billion entry price that we could on a $50 million entry price 10 years ago or whatever. Can you talk through that evolution? Talk through the partnership's way of thinking about it and talking about it—what you believe to be true that results in this decision to do this?

Just go back to why LPs, starting with Swensen and everyone else, started investing in venture capital. Fundamentally, it wasn't because they thought they could beat the Nasdaq or the index by 3 percentage points a year, 5, or whatever. It's because there are situations where venture capital could drive these insane multiples on invested capital. From a financial perspective, that's what they're seeking.

For the longest time, for most of the industry's history, 2 things were synonymous: early-stage investing and high cash-on-cash multiples. The way to get high cash-on-cash multiples was to do early stage. That's it. Those 2 circles in the Venn diagram almost perfectly overlap.

The thing that's changed is that recently, relatively recently, in the last few years, because outcomes have gotten so much bigger and these markets are bigger and everything else, the circle of high-cash-on-cash-multiple opportunities is bigger than just early stage.

It's not so big that there's a gazillion new companies in there that you can generate 100Xs on. That's not true. But there are certainly many outside of early stage where you can generate high returns. That's it. I think that's what we want to go after.

You could argue we're a few years late. I think I'd take that criticism, but I think that opportunity exists on a go-forward basis. We should go do it. Everything else that we represent—the high-conviction, high-commitment partnership—has to still be there.

As part of that discussion, what were the other sides of the debate? Maybe we would've said the same thing in '99 and halfway through 2020. As markets get exciting, the possibilities—we all make this extrapolation error. What were the counterarguments to saying, despite all that, let's still not do it?

Yeah.

I think the biggest counterargument that really made this time right versus 2 years ago or whatever was that you need the team that can do it. It’s just a different mentality. There are differences in how you evaluate and think about things. All the other stuff—why not change and stay within your circle of competency and all those things—are true, but that was the biggest one.

We had several examples over the last couple of years where we had the right intuition on a company or an opportunity, and we didn’t do it.

’Cause it was outside the box.

Because it was outside the box. That’s obviously dumb, and I think it is quite different than a lot of ventures. We are really chasing these very rare, special companies that have very high cash-on-cash opportunities, where we think there can be runaway successes and we can invest in them.

Is there any lesson to be pulled from the many, let’s call it, 20 to 100Xs that you personally have observed? It’s such a crazy amount of return. Obviously, it doesn’t pencil in the beginning. You can’t make something pencil if it was that clear.

Right.

The price would be different. What have the 20 to 100Xs taught you in aggregate, if anything?

Work with really special people. You want to work really hard. You want to work smart and get lucky, and you need it all. It really all has to come together. There are a lot of things that are timing-dependent that you have no control over as a company.

Take the Cerebras example as a good one. This is our second time—we took it public this time in May, but we tried to take it public in 2024, and it would have gone public at a much, much lower valuation. It didn’t work out because of CFIUS and all this stuff. So timing matters. The advancement over those 18 months made all the difference in the world for a bunch of things that were honestly outside of our control.

There were some things that were in our control, like getting inference running and everything else, but there was a lot of stuff that was outside of our control. These are all the classic things. We just have to focus on what we can control.

That’s one of the things that’s really different from software companies. With software companies, aside from building on AWS or whatever, you pretty much own your whole stack, and so you’re really fully in control of your destiny in that way. With hardware companies, you don’t. There’s an entire supply chain, and the HBM thing, yeah—all of a sudden HBM is a thing, DRAM is a thing, and TSMC is a thing. A lot of those cross geopolitical borders, and so geopolitics gets involved, and that makes things complicated.

That’s a really big difference, and so you have to get lucky on the timing and macro and other stuff. But I think it really starts with working with these crazy people with unbounded opportunities.

Yeah.

If you work with these crazy, special people on unbounded opportunities, you get lucky from time to time. You’re bound to.

When I was 20 years old, I was working at an investment bank. Ben Horowitz, Mark then had started Loudcloud. It was still in stealth, and Ben gave me an offer to be his assistant.

I was talking to this associate who seemed like an elder at the time. He was probably 25. He said this thing to me that stuck with me. He was like, “Do you golf?” It was the classic banking question: “Do you golf?” I said, “No, I don’t fucking golf.”

But he said, “With golf, you keep practicing, keep getting the ball on a par 3 close to the pin, close to the pin, close to the pin. You keep getting the ball close to the pin and you keep practicing. That’s hard work. That’s working smart. That’s what you want to keep doing. Getting a hole in one, that’s luck.”

I kind of love that framing. I use it with my kids, actually. What it says is that, yes, there’s luck involved, and there really is luck involved, but there is actually a way to increase your luck. The way to increase your luck is to get a lot of balls close to the pin. Eventually, one will drop. I like that.

Each of us invests in 1 to 2 companies a year. I think in my 12 years, I’ve invested in 18 companies total.

Crazy.

Which is a relatively small number, so it’s very high conviction and very high commitment. I have a lot of skin in the game. I believe in these companies. If you keep working with these very special people in these opportunities, magic can happen.

What have you learned about the best reasons and conditions for going public?

Ben Horowitz does these Monday night dinners, and we had a CEO last night from a multi-hundred-billion-dollar private company. We had this whole conversation, so it’s kind of fresh.

I think that ultimately, when you go public, you have a range of new opportunities in what you can do. You have public trust, because there’s some transparency that comes with being a public company. You obviously have a currency that you can do things with. That ends up being there.

You have an unbelievable ability to raise capital, which I think is why the labs will ultimately go. Although I think the trust thing is actually a really important element of why they should go, and it’s beneficial to the world and to America if they do go public.

See what’s going on.

Yeah, see what’s going on. Everyone can see it. I think that’s a really beneficial setup.

There’s another element of it, which is: What does a collegiate athlete want to do? Go pro. They want to play at a higher level. Is it harder? Yeah, it’s harder. Is the competition tougher? Yeah, the competition’s tougher. They move faster. They’re tougher. They’re bigger. They’re stronger. The stakes are bigger. The stage is bigger. The scrutiny is bigger. All of that’s true.

It’s kind of the same thing with companies. There are a handful—and it really is a handful, 3, 4, whatever—that can get to this tremendous scale without going public because of their execution and excellence.

Yeah. Lots of free cash flow.

They have lots of free cash flow, and they’ve done really well over a really long time. I think that’s fantastic. Good for them. But in general, for everyone else, get out there.

The other thing that I would tell you is that there are windows for a particular type of company. The SaaS companies that went public in 2021—a whole boatload of them have struggled, and it’s been tough in the public markets because their stocks ran into this multiple-compression issue. They were trading at 30 times. They 4X’d in size, but now they’re trading at 6 times. It turns out you’re still under.

That’s a tough place to be. I will also tell you that there are probably 500-something SaaS companies between 100 million and 500 million that are private. What happens is that those employees never got a chance to sell. Those employees don’t have annual tenders. Those employees don’t have an opportunity to exit. Those investors don’t have an opportunity to exit. They’re stuck.

I don’t think they’re all going away, and as I said, I don’t think they’re all getting vibe-coded and everything else. But ultimately, the AI natives with their growth rates have sucked all the oxygen out of the room and all the interest, and the window was missed. That’s tough.

9. AI Value Is Not Zero Sum

What are the biggest debates right now inside of the partnership? I always love coming here and talking to you guys when there’s something interesting going on because you debate. It’s healthy. It can be really fun to watch, and I learn a lot from it. What are those debates today?

There’s a ton of debate around AI infra, apps, foundational models, and the infra/apps ecosystem: Where does value accrue, and how does it accrue? Where are the moats, and how do we think about that?

There’s also the business-model innovation. I think one of the things people don’t understand about why SaaS did so well versus traditional software is that it wasn’t just that it was a better delivery model and everything else. There was actual business-model innovation in it.

You really did have this subscription element that ended up being fantastic for both the company and the customers. It was a win-win situation. That same thing actually exists in AI in selling by outcome and that piece of it.

How much value just accrues to the labs?

Yes. How much of the value just accrues to the semis? That’s a real discussion.

What do you think?

I am of the view that it all works. It’s a very weird thing. Will the CSPs do well?

Yes.

Yes. Not all of them, but will some of these neoclouds do well? Yes. Will the Fireworks of the world do well? Yes. Will Nvidia do well? Yes. Will these chip startups do well? Some set of them, yes.

Are we going to have edge inference on our phones? Yes. Are we going to have near-edge inference on POPs? Yes. Are we going to have big models in data centers? Yes.

There’s so much zero-sum thinking, which is just: “Okay, how do we cut up this pie? They’re going to eat this much.” And, “Oh, no, no, no, no, Anthropic or whomever is going to eat 98% of the value, and they’re going to do all the drug discovery.”

I was like, “Come on. No. That’s not what’s going to happen.” When I say I think everything’s going to work and I list off all these other things, it’s really important to understand that doesn’t mean that every company that’s doing every one of those things is going to work.

It actually means quite the opposite of that. Most companies in each of those areas are not going to work, and it’s actually more important than ever to have real differentiation—to really take each of these thoughts to their logical extreme and understand, “Wait a minute, you have to go all the way on these things and really be differentiated on it.”

Is there anything that you have your eye on, whether it's in the funding market, in the technology world, or anything at all that you're really watching carefully?

It's actually funny to me. Some of these people are so smart, and yet they're in this tech world where they're reaching these deterministic, almost conclusions of mass unemployment and all of these different things. I'm going to give you a concrete example, which I think is so good. Take Geoffrey Hinton in radiology. I think it was 2016 when he was like, "We should stop training radiologists. AI's going to do it all better." Geoffrey Hinton is 3 orders of magnitude smarter than I am. He could not have been more wrong, but the actual thing that led him to make that statement or conclusion was 100% correct.

If you look at radiology images, we should be able to train AI to do a better job reading these things than humans, and that's probably true. Actually, I think studies in the area have shown that to be true. We have an investment in a company called New Lantern, which is approaching this, but the big hurdle and the big thing that it articulated was, "Wait a minute. First off, all of the aggregated training data set doesn't exist anywhere." What you see are companies going after chest CTs or very specific elements, but your typical radiologist looks at a whole variety of things every single day, from X-rays to CTs to MRIs of all parts of the body and everything else. AI tackling specific areas like chest CTs is only marginally helpful because it's only doing that one thing, which could be 1 of 20 things or 40 scans that they read that day.

Problem number 1: You don't have the data, just like we talked about in robotics and everything else, to train the AI. Problem number 2: The whole healthcare industry is oriented around reimbursing doctors for making readouts.

Mm-hmm.

How is that going to work? There's liability associated with that, and there are repercussions of getting something wrong or missing something, and there's medical malpractice and everything else. How are we going to avoid that, and how are we going to get around that?

Problem number 2 is real-world stickiness. We're going to end up with this application where AI really does help radiologists. It helps radiologists get higher and higher throughput because the AI can do some parts and the radiologist does some parts, and they're checking each other and everything else. You do weirdly end up in this co-pilot situation for some time. Then you're going to slowly have the AI read more and more of the scans and build up and build up and build up.

But the actual duration to get from here to there is going to take a long time. In the ensuing time, we need more radiologists, not less, because everyone's getting more imaging than they used to get because the cost of imaging is going down in a Jevons paradox kind of way. My point is that you have someone very, very smart who really understands the capabilities, really understands what's happening, and has the right data, but by not thinking about that data in the real-world application, comes to the wrong conclusion. That's how I think of the unemployment thing. I think it's almost the exact same setup.

Eric, I love talking about markets and companies with you. It's an absolute blast. Thanks for the time.

Thank you.