Baseten CEO Tuhin Srivastava on the AI Inference Crunch, Custom Models, and Building the Inference Cloud

Baseten CEO and co-founder Tuhin Srivastava sits down with Sarah Guo and Elad Gil to discuss the rapid growth of AI inference demand, Baseten’s 30x growth, and why inference is becoming the strategic “last market.” Tuhin Srivastava argues the application layer will persist because companies with unique user signals can encode value into workflows and post-train specialized models, citing examples like Abridge and support workflows. The conversation covers GPU capacity constraints, Baseten’s multi-cloud fabric across 18 clouds and 90 clusters, long-term contracting dynamics, the importance of the software layer for stickiness, evolving workloads, multichip possibilities, and operational lessons at scale. Sign up for new podcasts every week. Email feedback to [email protected] Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @Tuhinone Chapters: 00:31 Baseten growth 01:55 Why the app layer wins 05:57 Serving frontier customers 07:55 Open source model mix 09:21 Chinese models and geopolitics 13:07 Custom inference dominates 14:22 Post training acquisition 17:10 When to invest in custom models 18:35 Supply crunch and data centerse 22:25 Longer GPU Contracts 24:09 What Makes a Winner 26:07 Multi Chip Future 28:19 Runtime Roadmap 31:08 Scaling Edge Cases 33:48 Hiring and Leadership 36:44 Operations Pager Culture 38:19 Efficiency Drives Demand 40:41 Concierge Everything Future 42:34 Conclusion

• 0:05Speaker 00

  Hi listeners. Today, Alad and I are here with Tuhin Strubastava, the founder and CEO of Base10, the AI inference cloud. We're here to talk about capacity constraints for AI compute, why inference is the last market, how the workload is changing, the open source and perhaps multi-chip future, and what 30x scale in a year looks like. Tuhin, welcome back.

• 0:28Speaker 01

  Hi.

• 0:29Speaker 02

  Good to see you. Thanks for having me.

• 0:31Speaker 00

  All right, you are in one of the craziest markets, AI inference. It's very important. There's a lot going on. You guys have grown 30x over the last year. And I think I can say you're expecting to do more than $1 billion in revenue this year. What's going on? Tell us about scale.

• 0:49Speaker 02

  Yeah, no, it's been nuts. I think what's happened over the last, honestly, 24 months, but it just kind of keeps getting bigger and bigger, is that I think everyone is realizing that you can put AI everywhere. You have all these great options available, from closed source to open source models. The open source models have crossed some sort of chasm in terms of their baseline capability. And then I think RL techniques and post-reining for specialized models has become mainstream enough. And there's enough examples of it working. The customers realizing they can kind of own their inference more and more.

• 1:34Speaker 02

  And what that's meant for us is more, the long-tail of models coming true. Customers in-house seeing a lot of that intelligence themselves. And as the application layer just gets bigger and bigger and bigger, and that's growing, we are just someone indexed on that. And we've been around to be able to collect the demand.

• 1:55Speaker 00

  There's an existential question in here that I think everybody is continually asking of, does the independent application layer get to exist at all versus the labs? Like, how do you, you have to believe this. Why do you believe it?

• 2:07Speaker 02

  Yeah, look, I think it'd be, it'd be a sad thing if it didn't exist in general. And I think that's like my, but you know, sadness is fine.

• 2:16Speaker 02

  Sad all the time. Yeah, sadness is fine. But that's not the reason why I think the application layer will exist. I think the application layer will exist for a number of reasons. One is because, you know, I think this idea that what is valuable to a company is, you know, the, the user signal that they can gather, that only they can gather. And to the extent that that is encoded in a model, I think a lot of their business will be at risk. But to the, to the extent that it is encoded in workflows, that is where they will be able to develop most. So a good, a good example of that is, say, a company like Abridge,

• 2:59Speaker 02

  where the clinicians edits off the notes and what they do with those notes after the fact. And the, the thing that happens in, inside the EOMR, three steps down. And that becomes a workflow that only.

• 3:11Speaker 00

  Can you explain what Abridge does?

• 3:13Speaker 02

  Sorry, Abridge is an ambient scribe that is used by physicians in,

• 3:21Speaker 02

  you know, almost all hospitals in the US. I think lots of investors, great, ship's amazing, great, great company, great team, great product. And, you know, they've basically, you know, got this very, very deep integration into hospitals, into clinician workflows. And my argument would be here is that actually, you know, it's very, very hard for a frontier model company to go to eat away at that because they just don't have access to that user signal. And what will happen over time is the folks who have access to that user signal can start to post train models on their reward signal and, and start to get long, long horizon agentic models running that. And I think to the extent that that is possible and that signal is differentiated and unique

• 4:06Speaker 02

  and, and is somewhat.

• 4:12Speaker 02

  Rare to to get access to, there will be an application there. And I think, you know, it's like support companies, another example of that where, you know, a support, a support task isn't one shotted. Usually at a company like Basecamp, when a ticket comes in, there's like, what, like one, two, 10, 20 actions that get taken. And that is where, you know, someone can develop a specialized model.

• 4:34Speaker 01

  So there's almost two versions of this then. There's the new companies like Abridge or Decagon or some of these other things that you mentioned that are doing these new types of applications that are using AI and they sell it to customers. The other is enterprises building things in-house or building their own models. What proportion of the market today do you think is these new application companies versus enterprises just adopting AI? And how do you think that looks in a couple of years?

• 4:57Speaker 02

  Yeah, I think that's, that's, we did, I think you asked me the same question two years ago.

• 5:03Speaker 01

  I hate to be repetitive.

• 5:04Speaker 02

  It is crazy.

• 5:05Speaker 01

  At least I'm consistent.

• 5:06Speaker 02

  The answer is just that it's crazy that the answers do, I think, I think if you look by inference count, it'd be 99% the fall.

• 5:15Speaker 01

  Yeah.

• 5:16Speaker 02

  And that is the kind of represents the scope of the opportunity here is that the majority of the market hasn't come online and

• 5:24Speaker 02

  and added AI into the world.

• 5:25Speaker 01

  Most of enterprise adoption is well ahead of us. And I think that's one of the very exciting things about AI.

• 5:30Speaker 02

  Yeah.

• 5:30Speaker 01

  There's just so much still to come and people are underestimating that, I think.

• 5:33Speaker 02

  Hundred percent. And but what's cool is that we're seeing the transition happen, right? Before it was like, hey, are they, are they using AI tools? I don't think that was immediately obvious two years ago, and I think it's obvious now that, yes, they are. Are they using closed source model APIs? I think they're starting to get there. And then, once you do that, and then you kind of see what is possible, then comes the whole custom model adoption. I think that is all that is ahead of us today.

• 5:57Speaker 00

  So if the majority of your customer base today is, as you described, the the former like application companies, AI natives, the fast growing, I mean, some of them are at considerable scale now, like the Abridge Cursor.

• 6:10Speaker 02

  Open Evidence.

• 6:10Speaker 00

  Open evidences of the world. What? You know, what do they, what do they teach you? What does that push the company to do? How do you think about serving them versus evolving for the enterprise?

• 6:21Speaker 02

  Yeah, I, I think firstly, like you just learn a lot by building with the companies at the greatest scale, doing the most interesting things. We, we think of it two ways. Like, I think there's like the, the, the most obvious way, which is just build for the highest scale. You know, most, the, the customers that will push you the most from technologically and everything kind of will fall into play. I think the Stripe evolution as a company showed that was like Stripe now, like so, like so many enterprises. But 12 years ago, that wasn't the case, but they just built for the frontier and kind of went with them. And the second way we think about that, this is to just think about building for companies that are serving enterprises.

• 7:04Speaker 02

  So yes, we don't serve the enterprises, but our customers serve enterprises. Abridge serves, Open Evidence, Decagon.

• 7:12Speaker 02

  All these, right, again, all these companies, all these companies serve enterprises en masse. And what we actually get is like a translation of the requirements from them, which is like, you know, they're like, hey, we need this data retention. We need this, this web models need to be deployed. This is the types of GPUs or the latencies they're okay with. This is the model requirements from like a transparency perspective that they care about. And so I think that is actually the more nuanced answer is that if you listen to what their needs are, we actually get a full translation of what the enterprise was requiring. I would say that by serving companies like Abridge and Open Evidence, we're probably pretty well through to go serve the healthcare system, given that they are selling and latent health, given that they are selling to them.

• 7:55Speaker 01

  How, how much of a shift are you seeing in terms of the types of open source models that are being used? And so I think we've seen an evolution where two, three years ago, I think the main thing was kind of Mistral and then a few other things. And then Meta kind of came along with Lama and then it kind of really shifted.

• 8:09Speaker 02

  Yep.

• 8:09Speaker 01

  In terms of the most performant models or of Chinese origin in different ways. Do you see that sort of mix reflected in terms of what's being used by our customers?

• 8:16Speaker 02

  Yeah, I, I, I think customers, at least the customers we are serving are very, and these are like the fastest growing AI companies in the world, that are very forward thinking. They, they want to use this model and they, they are optimizing. I think there is. There are a, there's a subset of task which I think is small today where people really start to start with cost. Um, but everyone comes from capability first because that's really where the economic growth is being unlocked, where the value is being delivered. And then they optimize. And I think that's like actually being, you know, and so with that in mind, you know, you, you, you name like, you name it, everything from GPT OSS all the way to um, moonshot model to deep seeks um,

• 9:00Speaker 02

  to um, canopy or Ophir, which is like really good text to speech um, models um. Customers generally want to use whatever's at the frontier. And I, I think the um, the difference is just being um, I think we have a lot more visibility into how to run these and how to run these really well. And secondly, that they're good now.

• 9:22Speaker 01

  There've been a number of different concerns raised about the use of Chinese models. Yeah. In particular, security or is there something embedded in the models?

• 9:28Speaker 02

  Yeah.

• 9:28Speaker 01

  Or, you know, Trojan horses or other things.

• 9:32Speaker 01

  A, do you think there's any real concern there? And B, you know, people often talk about how there should be like US counterweights to this. From a geopolitical perspective, do you think that's something that's legitimate or something we should be worried about? Or how do you think about the?

• 9:44Speaker 02

  Yeah.

• 9:44Speaker 01

  Sort of origins of these models versus their uses?

• 9:46Speaker 02

  Yeah, look, I, I, I think these, these models firstly are fantastic. They're amazing. We work with these teams. They're truly awesome. I'd say, look, I, I don't, it, it is hard for me, it's hard for me to see. And I, I could be wrong, but like, you know, if, if I, if I, if I network bound these models that they're not magically, you know, going to be able to cross those network boundaries and to data that are, and you know, I, I don't. And we, I've never seen any real evidence um, except from some very early models that I think people picked up on very quickly that there is some agenda or bias built in these.

• 10:26Speaker 02

  I do think that to so, um, to some extent is, I, I think there is importance to the US that we develop our own models. I think that that would be a massive loss if that there are five companies, you know, five different labs in China that are creating open source models. Um, and we're struggling to get one set up. So it's necessary. I also think it's inevitable. Um, and you know, like the

• 10:54Speaker 02

  deep, the deep sea moment a year ago, um, I remember someone saying to me, you know, I thought it was like very well said, which is like, and the world's changed a lot, but they said, hey, you know, we should kind of just forget that this is a Chinese model. We should just act like this came from, from meta and, and build, and build with that in mind. It's like, you know, I, I think you're kind of missing the forest from the trees. Like there's two, there's two scenarios, right? E- either America does not ever come up with good open source models. And there's probably a fundamental problem there, or we will get there and we need to be ready for that world.

• 11:28Speaker 01

  Yeah, that makes sense. It's interesting because, um, you know, like you, I, I think it's very important for the US to have a strong open source footprint here.

• 11:37Speaker 01

  At least for now, it looks like effectively the Chinese government is subsidizing at least a large subset of these models. And that subsidy or surplus is effectively just being passed on to US enterprises who are adopting these models. In other words, it's a way for the Chinese government to effectively subsidize US enterprise.

• 11:50Speaker 02

  Yeah.

• 11:51Speaker 01

  In an indirect manner. And I think that's a little bit lost right now. But you know, it's always interesting to weigh that against some of the other concerns that are raised. I appreciate your, your comments on this.

• 11:59Speaker 02

  And I, I think the concern also just there just becomes, it's like, what happened if we aren't able to? Like i- i- i- if it is fun, like I, I think if you think of the economics here, which is deep, deep sea by most, deep sea is a good, a very good model. You know, like, and like, you can argue whether it's at the absolute frontier or not, but like, let's, let's go back three months and it's there. And so think about everything. We were doing a whole lot of things three months ago.

• 12:25Speaker 01

  Yeah.

• 12:25Speaker 02

  And so let's just think about that. Well,

• 12:28Speaker 02

  you know, if it, you could run deep sea probably 20% of the cost of running open and drop big models um, in production um, with comparable, better latency, probably better reliability. If we don't have access to that intelligence in that form, I think it's just a massive loss. Um, and as a country, we won't be able to innovate as fast because like the, the cost of intelligence going down and control of intelligence. Well, we have seen just means more intelligence.

• 12:53Speaker 01

  Yeah.

• 12:54Speaker 02

  Intelligence being embedded in more places.

• 12:56Speaker 01

  Yeah. An important note uh, here that we didn't mention explicitly is that the state of the art models, the ones that are most far ahead on the frontier actually still the closed source, anthropic, open AI, Google, et cetera.

• 13:06Speaker 02

  Yeah.

• 13:07Speaker 00

  What has been, actually, maybe you can just characterize like workload a little bit, like how of tokens being served on Base10, like how many of them are from custom models of some kind versus like vanilla open source today?

• 13:20Speaker 02

  It is all custom. It's basically.

• 13:24Speaker 00

  Okay, so like 95%.

• 13:26Speaker 02

  95%, and I think that's really cool, to be honest. And look, we have, we have two businesses, we have, we have three businesses.

• 13:35Speaker 00

  Should we help you count?

• 13:36Speaker 02

  No, no, so we have like dedicated inference, which is basically custom model inference. Your SLA is your SLA. Then we have shared inference, which is shared inference endpoint, shared SLAs, and then we have a training business.

• 13:51Speaker 02

  I'd say 95% of the tokens today are on the first business. And almost all of them, there's probably, yeah, for almost all of them, the customer's making some modifications to the model with their, with their own data, specialized for the use case. And I think what's even more important is they might be compiling in different ways. No one is just running the vanilla open source weights. Like you, you might be customizing it for quality, but you most of might be customizing it for performance.

• 14:22Speaker 00

  You made an acquisition of a research team a few months ago. You've mentioned post-training customization. What was the rationale behind the acquisition? What is that team doing today?

• 14:34Speaker 02

  Yeah. So the rationale around the acquisition was, you know, we are infrastructure and product people. We are product people and now are really good infrastructure people. And the, and we didn't have much of a research capability ourselves. And what we saw was the market moving heavily, heavily, like that we could accelerate the market itself with post-training resources, either productized or onto even just as resources for that market. So Pars was a company that was a Base10 customer. So there were post-training models and running them on Base, on Base10. And I think.

• 15:20Speaker 02

  What they realized was that they would eventually need to become an inference company. And what we realized was like, hey, we really needed that expertise because it is, it represents a way for us to get closer to the customer earlier. And, you know, be able to support them all and just made sense as a like pairing them together. And just as, as I said in the opening statement here, which is, you know, as more and more post-trained models have come up, we've realized that the demand for people, for people to either

• 15:58Speaker 02

  for software loops to do post-training or for post-training expertise is very high and we're really, really investing in that. There are also a bunch of Australians, you know, I like to think that we had a bit of alpha there. But yeah, that's been fantastic. They're working with all sorts of customers. And it's also very interesting when you, when you

• 16:22Speaker 02

  start, you know, we were doing a lot of research on the performance side and less so on the post-training side. It's interesting as we've started to do a lot more research on the post-training side, you start to see how linked inference and post-training are. And like, you know, even, even when you think about stuff like quantization and when you should do that and like, you know, how, how, how training, how, how you train the model affects how you need to quantify for inference and how paired these problems are has become like very apparent and more and more we're realizing the post-training inference are kind of both sides of the same problem. So because inference ideally will be get more post-training where inference creates data, you do evals, you can now post-train, post-train on the, on that reward function

• 17:06Speaker 02

  that you, that you found with those evals and hopefully just set up that entire loop.

• 17:10Speaker 00

  Plenty of folks from Ant and OpenAI, Sam, Greg, et cetera, have said in recent months that like inference is super strategic, inference talent is strategic, capacity is strategic. So between that and post-training, these are very difficult to gather like capabilities. I imagine that lots of your customers go to you guys for advice on like how to do this progression of moving to custom models. Like what do you tell people about the lifecycle and when they should invest in that?

• 17:42Speaker 02

  Yeah, I think it's, hey, go find, go prove to yourself with the best in class model that you have something worth optimizing. And I think, you know, a lot of, you know, if a customer comes to us, was that meme, which was like, it was like two years ago, it feels like no GPUs pre-product market fit. It's like no post-training pre-product market fit is what I'd say.

• 18:07Speaker 00

  So people that you're working with here are very, very at scale first.

• 18:09Speaker 02

  Yeah, they have a user signal that they know how to optimize. And they've shown that they can, you know, they can serve customer value and that value and that they have something special around that value. And once you have that value, it's like, okay, now how can I do that better, faster and cheaper? With the idea being that, hey, if you need to be very good at customer support, you can, you maybe don't need to be that good at coding and that a specialized model might be a better fit for that problem and you can do it better, faster, cheaper.

• 18:35Speaker 00

  What about the capacity side? You started with unifying capacity across all the clouds and new clouds. How do you think about this when everybody keeps talking about a supply crunch and a multi-year supply crunch?

• 18:47Speaker 02

  I think, you know, there's so much narrative around the supply crunch. And no matter, like, as much as we hear about it, I don't think people realize how bad it really is. Like, there is, you know, there's very, very little slack compute available. Like, you know, we run pretty large clusters ourselves and we run them in like uncomfortably high utilization. You know, when I'm saying we're like mid 90s utilization, most of the time, there is, we have made, we have, we sit in 18 different clouds now. We have 90 clusters around the world across 18 different clouds. And like, you know, initially, we started, we like built this

• 19:32Speaker 02

  technology to be able to like kind of create one runtime fabric that spans all these different clouds and try to abstract that away from our customers as a way to think about reliability, latency, failover, all these things that we think are going to be very important for very mission critical use cases. That same technology, like just our ability to get compute wherever humanly possible, has been really, really helpful in our ability to get supply. And what I mean by that is, we can be introduced to a new provider. In a different country and have it up and running with the whole base 10 inference stack.

• 20:10Speaker 00

  As part of the fabric.

• 20:11Speaker 02

  Part of the fabric in half a day, half, maybe less. Even and that gives us enormous flexibility.

• 20:22Speaker 02

  Even for us, it is hard for us to grow. We have a, we have a, I think it's, yeah. We have a, a 4pm standing meeting for the company where we basically like, like, how do we, like, how we, how we manage capacity for the demand right now? I think the second part which people don't really, the two, the, the, the second part that people don't really understand is that there are also a lot of suppliers right now that it's kind of grifty. You know, like, I, I, I think, you know, they haven't run, they haven't run data centers before. You know, they don't understand SLAs, especially for inference.

• 21:07Speaker 02

  And so like, you know, even when there is capacity available, there's a lot of, like, there's probably, we, we run a lot more and we have redundancy, so it's fine. But if you, you know, there's probably like a dozen good, like clouds and I'd probably like, put like three or four of them in like the, the gold tier.

• 21:29Speaker 02

  And I think that just means that like supply, like not only are we supply crunch, we're supplier and operationally crunched onto people who can, who can run these data centers as well.

• 21:39Speaker 01

  So how far ahead can you actually buy capacity right now? In other words, like, is there any, any slack in the market if you buy two years ahead or five years?

• 21:47Speaker 02

  You, you, you mean like actually like contract length or actually like, hey, I want this in January 28?

• 21:54Speaker 01

  Yeah, either one. Yeah.

• 21:55Speaker 02

  Yeah.

• 21:56Speaker 01

  I mean, it's more the, I want this in January 28 or at least I have some visibility into my future supply.

• 22:00Speaker 02

  Yeah. Um, you could buy that, but you would also remember how quickly the market is, how quickly the market is moving. And like, you know, that gets balanced somewhat off like the fact that the H100 is such a great chip. And like, and then, you know, it's crazy, it's four years, four and a half years old and the price is going up still. Maybe it has to use for like nine years. So, you know, that, that's, that's good, but at the same time. At the same time, you know, yes, you can do that.

• 22:32Speaker 02

  But you're making a lot, like you're making a lot of bets.

• 22:35Speaker 01

  Yeah.

• 22:36Speaker 02

  As part of that. And then in terms of, I think that's the big thing that's changed over the last six months is that the term length that people want has just gone up. So if you, if you wanted

• 22:48Speaker 02

  a thousand, a thousand, thousand twenty four B two hundreds, which is, you know, from a good cloud, right now, you're not getting that less than a three to five year contract. Right now with a probably a twenty to thirty percent T C V, T C V prepay. So like actually what becomes important when acquiring capacity is you need to have enough demand to supply it to serve, but then you also need like a low cost of capital, which is, which is actually changing the dynamic pretty significantly.

• 23:20Speaker 01

  Does that, does that impact how you think about going public as a company because arguably?

• 23:24Speaker 02

  Yeah, I think you'd go sooner.

• 23:26Speaker 01

  Yeah, exactly.

• 23:26Speaker 02

  Yeah, I think you need, like, I think the, and I think there is demand for that. But I think, you know, the pool, the, you know, and also, you know, one of our, one of the, one of the realizations that we had recently and with, with software people. And so we don't, we don't think like this all the time is that, you know, our, our business has like very interesting working capital. Um, requirements like we don't, um, and, and I think, you know, even, and that as a result of that, it has very interesting financing.

• 23:58Speaker 01

  Yeah.

• 23:59Speaker 02

  Requirement and we're not, at least right now, we're not even going down to the, down to the dirt.

• 24:03Speaker 01

  There's also things you could do in terms of debt or other structures that, yeah.

• 24:06Speaker 02

  Yeah. And, yeah, I've learned a lot about debt.

• 24:09Speaker 01

  Yeah.

• 24:09Speaker 02

  Recently.

• 24:10Speaker 00

  Given the supply crunch inference being one of, you know, the top couple markets you're going after, you have plenty of people who understand this problem and therefore, you know, some competition. How do you, uh, how do you think about like, what are the factors that create a dominant player here or a winning player? Is it, as you mentioned, cost of capital? Is it access to supply? Is it software? Is it demand?

• 24:36Speaker 02

  Yeah.

• 24:37Speaker 00

  Just being excellent, everything.

• 24:39Speaker 02

  Yeah, it's, um, look, I think what's so interesting about inference is.

• 24:44Speaker 00

  Is it operations?

• 24:45Speaker 02

  Yeah, I think, yeah, I think GPUs as a service is not sticky. I think that's been seen. Customers generally just see that as commodity. Inference with the software layer included is incredibly sticky. You know, like, just, just like, you know, none of our top 30 customers have ever churned. You know, we're talking like 400% annual NDR around our business. And so it's like, very, it's, um, it's very, very sticky. So I think that software layer is very important. The optimist in me is like, oh, there's so much value in the software and I, we will build the best software layer for inference that exists. I think, you know, as I think is becoming clear now, access to inference computers.

• 25:29Speaker 01

  Yeah.

• 25:30Speaker 02

  Is a strategic, strategic advantage. And I think that is like the, I think that is the strategy that even the labs are going after, which is like, if we have the, if we have all the compute, good luck running inference.

• 25:43Speaker 01

  Yeah, yeah. In a world of constrained compute, the number one thing to own is compute.

• 25:48Speaker 02

  Yeah.

• 25:48Speaker 01

  And so, you know, just owning it in and of itself as an asset. And I think people underappreciate that.

• 25:52Speaker 02

  Yeah, you can't, you can't make a good hot chocolate without milk. And you know, the.

• 25:56Speaker 01

  Unless you're a vegan.

• 25:57Speaker 02

  Unless you're a vegan. Say no one wants a vegan inference.

• 26:02Speaker 00

  Well, I'm going to ask you, people might want, um, they want, they might want alternative milk, right? So, okay, like when you, the H100 is a great chip. People, you know, want to be 200, they want to be 200, they want, of course, tons and tons of Nvidia. When you think about making a bet, you know, several years in the future, do you believe that there is a like multi-chip world? Like, what do you, what do you think happens from a compute perspective on the chip side?

• 26:28Speaker 02

  Yeah, um, I think.

• 26:32Speaker 02

  I think, you know, like diversification everywhere is a same way I want to water many models. I think, you know, we want to water many most things. And I think.

• 26:41Speaker 00

  It'd be sad if it didn't happen.

• 26:42Speaker 02

  Yeah. And I think everyone would be sad. I will say, to some extent, which is, yeah, and I think there will be inference specific chips. I think you have like decode specific chips, I think. And we're looking.

• 26:55Speaker 00

  And Nvidia said this.

• 26:56Speaker 02

  Yeah, yeah. I mean, that was a whole Grok LP thing. It's like, you know, I think I think that is very straightforward and it makes sense. I think people really, really, really underestimate supply chain stuff in Nvidia, like how good they are at that. Kuda, how good Kuda is, the developer ecosystem around it. And, you know, we, the ability, like, to me, like, one of the most important things as an infrastructure company in this moment is how fast you can move. And you can move fastest with Nvidia today. And I think that is the reality. And like, it just like, given the scale that they operate at, given the scale that they operate at, it's, it's hard to, it's hard to see, it's hard

• 27:41Speaker 02

  to see the, and I'm not saying it won't happen, like, the short term, like, in the next couple of years, how anyone's gonna be able to compete for that. Especially with, you know, so much of the other, the other players, like, what you need to be able to compete here is the ecosystem to form around you. And if you tie up all your supply with one buyer, which, you know, a bunch of the other chip providers have done, it's actually hard for that ecosystem to form. You know, like, if you think about, if you're a big lab and you have a proprietary deal with one chip type where you get 90% of the supply, it's actually in your best interest to make sure you get 95% of the supply and everything that's good built for you and no one else to ever use it.

• 28:19Speaker 00

  When you think about reacting to the market, what do you think is like happening with the actual workloads that you have to go invest in, right? Like, obviously, code agents and long horizon agents over time have become a big deal. People talk a lot more about CPU compute, video inference is different. I don't know if it's that, sandbox this, like, what what's important for you guys to invest in now?

• 28:42Speaker 02

  Yeah, look, I think for us, all the runtime stuff is obviously very important. And what that means is like, what chips we run on, how we run, what kind of workload we support. Like, do we get very good at diffusion transformers? Yes, coding agents need sandboxes, we should go with sandboxes. There's all sorts of new speculation techniques to get faster inference. We need to do that, even stuff like KV cache aware routing. And you know, that stuff's a bit old now, but like getting, continuing to be very good at that. And somewhat disentangling preview and decode. I started to treat them as separate problems. I think that's, you know, something we are very focused on and we're seeing massive gains there. That's at the runtime level.

• 29:24Speaker 02

  I'd say, you know, beyond that, you know, everything we think about is how to create more of that loop between inference, post training, because we think that just begets more inference. And so like, we will build a partner in almost everything there. So, you know, we're going to work with, you know, the best evals company in the world to make sure that's very well integrated, like brain trust into and around base 10. You know, we will partner with all on the sandbox side, build the best sandbox experience that will exist. And then we'll create the best training APIs to make it so continual learning becomes somewhat of a solved problem. It's not just like a discrete thing.

• 30:08Speaker 02

  That's, I think, the core base 10 product thesis. It's like, how do we build that loop? And then everything out around that becomes how do we make sure that we can do everything we can to ensure that gets the biggest possible? That's access to compute, that's an infrastructure, make sure we can get compute anywhere, make sure we have access to our own compute. And then I think it's all the primitives that come off of that. That just become incredibly like margin agreed of both for us and our customers, which is, you know, stuff like, you know, sandboxes and like the async batch inference, like how do we drive utilization by having a first-class batch inference experience? To me, this is like what an inference cloud looks like.

• 30:50Speaker 02

  It's that you are very good at inference and then you start to do all the things tangential or that loop into inference and partner and where necessary and build where necessary. But we really do want to own, like start with that core inference story and then go down. To unblock supply and create margin and go off the stack to unlock value.

• 31:08Speaker 00

  What would surprise people about some of the issues you discover only at scale? I'll give you an example. I was surprised when you guys ran into scale limitations, like fundamental limitations with some of the hyperscaler products that you were consuming.

• 31:25Speaker 02

  Yeah.

• 31:25Speaker 00

  And I'm because I kind of think of, you know, the AWS GCPs of the world as supporting infinite scale.

• 31:31Speaker 02

  Yeah. I mean, I think you just, and like, again, like, I think very, very large companies that run services of big scale, it's probably the same stuff.

• 31:39Speaker 00

  Mm hmm.

• 31:39Speaker 02

  Is that all the edge cases just become-

• 31:42Speaker 00

  You actually experience.

• 31:43Speaker 02

  You experience them. And like, you know, and you, I like, I'll give you a few examples here. Like you see, you know, you start seeing, you know, yesterday we had, for the first time ever, we saw some kernel panic. And that only happened because some fluent bit worker was creating too many logs and and the scale was too big and it was all into one node and it was happening two, two, two turns at the same time by two different workers. So you see all like the systems level and kernel level problems. But then you start to see, I think the craziest stuff is that you start to see with LLMs

• 32:22Speaker 02

  that these runtimes are pretty immature. Even how we use KV cache is, you know,

• 32:30Speaker 02

  you know, probably a little less sophisticated than most people see, than most people see. And we are starting to see the limitations of the current and the next set of primitives that need to be built from a scale, security, a performance perspective. But I think it's really at the runtime level and the systems level and then, but the edge cases are, I'd say a lot more systems level than they are LLM specific.

• 32:54Speaker 00

  What are the things that keep you up at night?

• 32:56Speaker 02

  Capacity. I think, you know.

• 32:58Speaker 00

  Quick answer.

• 32:59Speaker 02

  Yeah, I think capacity. I think the other one is probably just this market's so big and it's so like it represents a moment when you should be as aggressive as possible. And, you know, really, you know, we've grown a ton obviously over the last 12 months, the last few months, but the answer is always just go, you know, go bigger, go faster. And I think that's really, really fun. It's also a little exhausting and it's also like we are, we are all in somewhat uncharted territory in terms of how fast and how big you can go and how things can get. But I, but I think the big one is compute. I think like there's no world in which there's enough compute to, you know, get the amount

• 33:41Speaker 02

  of, the amount of value that we want to get out of our limbs in the next five to 10 years.

• 33:45Speaker 00

  Or we have to invent a lot of new stuff.

• 33:47Speaker 02

  Yeah.

• 33:49Speaker 00

  Maybe if we just talk a little bit about what you're learning scaling, you know,

• 33:56Speaker 00

  30X is like an aggressive thing to go through as a company. You've brought in a lot of really amazing talent like Danny and Samir and Stephen Day, folks on both the, the technical and the go to market side. Like what do you, what do you think is working about how you are recruiting and scaling? Or what's your philosophy on that?

• 34:19Speaker 02

  We were very, very flat, like until, I know 12, 13 months ago. I remember I went on walk with a lot, actually, and a lot of it's like, you just need leaders. And like, it's actually like so contrary to everything, you know, as engineers, you're like, oh. All overhead. It's all, everything is overhead. Everything is overhead.

• 34:41Speaker 00

  You once told me, I think that you, you didn't, you're like, hey, Sarah, Sarah, what about we just have engineers instead of salespeople?

• 34:47Speaker 02

  Yeah. Yeah. Yeah.

• 34:49Speaker 01

  Everybody learns it.

• 34:52Speaker 02

  I remember like, you know, you, you, you said it so clearly at the time a lot. And I, I think that's what we've noticed. We're just like actually having a leadership team that you can trust, that you can trust is, is, is so important. I, I think the, the two or three things that I would say is like, you want people where you can give them whole problems. And so like, you know, if, if, if you are, if you feel like you are micromanaging, if you feel like you need, if you feel like, you know, you, you have to be involved in everything. I think that's a bit of a cop out as a founder. Cause you're just like, I just need to be involved in everything. It's like, no, you, you probably don't have the right people. Um, I think the second thing is, um, be very, very clear what you're optimizing for. Because I think when

• 35:37Speaker 02

  you're very, very clear what you're optimizing for, the people on, and like, if it's something generic, like we want the smartest, hardworking people, like you can't do much with that. Like with us, what we cared about was, hey, actually, we don't care about a lot of people who have done this before. We care about first people who think first principle, first principles, work has to be, um, a high priority, but they also have to be very kind and nice and, you know, care about the collaborative environment. We don't have a hero culture. Um, you know, very low ego. Um, and you know, if you need, if you need a manager, like, um, it's probably not, it's probably not the right place, um, to be. But I think once you have, when you have that clear rubric, the, the people become very apparent that will fit into it. And the

• 36:22Speaker 02

  people that don't, um, fit into it also become very apparent. And I think what's more like, we've hired amazing people, like you mentioned, but I think what's a lot more interesting is like, I think we've, we haven't had a ton of like turnover there unnecessarily. Like people tend to work, um, cause we, cause we have a very, we are very clear in what we want. It took us a while to get there though.

• 36:44Speaker 00

  What about the idea of like an operations culture? And we were talking to Alyssa and Henry about this. And she's like, well, the hard thing about cloud is actually just operations. I slept with a pager under my pillow for a decade. I don't think I've seen you detached from your Slack channel.

• 36:58Speaker 02

  Yeah. My phone is buzzing right now. The, um, the, um, as I, I'm getting anxious. So, um,

• 37:06Speaker 00

  And, and you've, you've been concerned before, like, do people get it? Like, you know, what is distinctive about that?

• 37:12Speaker 02

  I, I think, I think. Like one, I think if you've worked at an infrastructure company, like we were once in a meeting with a bunch of AWS execs and this was, you know, like very senior AWS folks, all their pages went off multiple times during our forty five minute, you know, like it's a, I, I, I think like it's, it's, it's very much like it's a cultural thing. Um, but yeah, like I, I don't, you know, our like inference can't go down and like, you know, we, you know, you, you, you, you learn to like, you know, what was this? Like, I think me and my co-founder, when his pager goes off, his seven year old said, is that a P zero?

• 37:52Speaker 02

  Oh, is that, is that, is that a P zero? And so, you know, I, I think that is, you just have to get used to it. And that's the culture you live in. And it, it, it just changes the speed. Um, but also it's, you know, becomes like a, you know, a cultural thing. And I think it's very, very, it reject, it rejects people that don't fit into it very, very quickly.

• 38:10Speaker 00

  Like engineers who avoid pager.

• 38:13Speaker 02

  Yeah. You know, when we, when we have P zeros, we're like everyone on the call. Like, you know, like there's been a joke, there, there may as well be a siren that goes off in the office.

• 38:21Speaker 00

  So people have been talking ad nauseam in the AI community about Jevon's paradox.

• 38:26Speaker 02

  Yeah.

• 38:26Speaker 00

  Um, where if you decrease the cost of, it's, it's a, it's really a question around price elasticity and availability. If you decrease the cost of a good, say, intelligence as a good, um, people actually consume more of it.

• 38:41Speaker 02

  Yeah.

• 38:41Speaker 00

  Um, like the personal or business ROI of it, the demand for it goes up, not down. Um, do you see this? And are you, are you working against yourself trying to make these models more efficient? Do people just use them more or less?

• 38:54Speaker 02

  Yeah. I, I think you think about this from a developer's perspective and a consumer perspective. I think like, I think consumers just want the best answers and the, and the, and the best experience that's somewhat governed by, you know, more intelligence to some extent. I think when you go to the developers, from the developer's perspective, um, they would insert more intelligence if you make it cheaper. Like that's, you know, and they will, they will, they will insert more intelligence anyway. But if you make it in more cheaper, they'll, they'll insert a hell of a lot more intelligence. And you see this with agents that agents are just longer running. Now, and I think that's what we have seen with the cost of inference going down, which is, you know, folks are just like, okay, we can, we can run this for longer.

• 39:37Speaker 02

  We can make it do a bit more work and we'll get to a larger end. I think like if the compute scales from inference perspective as well. And, you know, I think we are seeing that with almost all our customers, which is, you know, they either, they either start with like, this is the quality of answer I need to get to. And this is the amount of inference like I need to do to get there. Or this is the base level model that I can start with, that I can work with to get there. And I think the more we drive down the costs, what they realize is more intelligence just means better.

• 40:11Speaker 00

  I just want a better answer.

• 40:12Speaker 02

  Better answers, better experiences, more dollars, more dollars, even more revenue. So, yeah, I think inference going down just begets more. But I, it is truly like, I think we're kind of in a world that is, you know, it is the last market. Right? Like even if there's AGI, all that's left is inference.

• 40:28Speaker 00

  Yeah. So you do not see in your customers a like this, this answer is enough and this action is enough dynamic?

• 40:38Speaker 02

  No.

• 40:39Speaker 01

  Yeah, it's going to keep going for a long time.

• 40:40Speaker 02

  Yeah.

• 40:41Speaker 01

  How do you view all this kind of evolving as a future? So basically this is one of the, it seems like it's going to be one of the biggest markets of all times. We have this massive shift where we're moving from software and seats and digitization into actual intelligence, selling units of cognition.

• 40:54Speaker 02

  Yep.

• 40:54Speaker 01

  Selling agentic workflows. What does this all look like in a couple of years? Like, what is your view of this future world?

• 41:01Speaker 02

  I think for the, for consumers, it's the best possible thing, right? Like every, everything is somewhat smarter, you know, your doc, you get better care because your doctors have access to better, better tools. There's more, you know, like there's all this stuff about there being less software engineers. And I think we just build more software. And we just build a ton more software and like, you know, I, I see, you know, we're not slowing down hiring of software engineers. We're just building more things. And for the consumers, that just means better tools, more software, all those, all those good things.

• 41:34Speaker 01

  It's almost like everybody has their own team for everything, right? You have an agent which helps with your doctor. You have an agent that helps you learn stuff. You have an agent that helps you organize your life.

• 41:41Speaker 02

  It's concierge. It's concierge.

• 41:42Speaker 01

  Yeah, concierge is everything for everyone.

• 41:44Speaker 02

  Yeah. And, and, and I think like what that means, well, that, that's amazing. I think that's great. And I think the, and education, same thing. You have concentrated education, like you get personalized access to everything. I think then you go one step. Back in how it affects developers, I think, you know, and companies, I think if you don't embrace this, I think it's the extension moment for a bunch of folks, which is like, you know, everything needs, and I don't think that means that, you know, for design needs to figure, I don't think that's a thing. I think like what's more, what's more interesting is to say, you know, all these workflow and software companies need to figure out what is the intelligent or intelligent inserted versions

• 42:28Speaker 02

  that that drive the amount, the all that user value for those and consumers that we talked about.

• 42:33Speaker 01

  Yeah, very exciting. Thank you so much for joining us today.

• 42:36Speaker 02

  Thanks, guys.

• 42:39Speaker 00

  Find us on Twitter at NoPriorsPod. Subscribe to our YouTube channel if you want to see our faces. Follow the show on Apple Podcasts, Spotify, or wherever you listen. That way you get a new episode every week and sign up for emails or find transcripts for every episode at no-priors.com.