Shawn Wang: evaluation

Transcript context

…There are a lot of things to say. So first of all, there's always somebody who's more GPU rich than you. So inflection is GPU rich by startup standard. I think about 22, 000 H100s, but obviously that pales compared to the, to Microsoft. The other thing is that this is probably good news, maybe for the startups. It's like being GPU rich, it's not enough. You know, like I think they were building something pretty interesting in, in pi of their own model of their own kind of experience. But at the end of the day, you're the interface that people consume as end users. It's really similar to a lot of the others. So and we'll tell, talk about GPT four and cloud tree and all this stuff. GPU poor, doing something. That the GPU rich are not interested in, you know we just had our AI center of excellence at Decibel and one of the AI leads at one of the big companies was like, Oh, we just saved 10 million and we use these models to do a translation, you know, and that's it. It's not, it's not a GI, it's just translation. So I think like the inflection part is maybe. A calling and a waking to a lot of startups then say, Hey, you know, trying to get as much capital as possible, try and get as many GPUs as possible. Good. But at the end of the day, it doesn't build a business, you know, and maybe what inflection I don't, I don't, again, I don't know the reasons behind the inflection choice, but if you say, I don't want to build my own company that has 1. 3 billion and I want to go do it at Microsoft, it's probably not a resources problem. It's more of strategic decisions that you're making as a company. So yeah, that was kind of my. I take on it. Yeah, and I guess on my end, two things actually happened yesterday. It was a little bit quieter news, but Stability AI had some pretty major departures as well. And you may not be considering it, but Stability is actually also a GPU rich company in the sense that they were the first new startup in this AI wave to brag about how many GPUs that they have. And you should join them. And you know, Imadis is definitely a GPU trader in some sense from his hedge fund days. So Robin Rhombach and like the most of the Stable Diffusion 3 people left Stability yesterday as well. So yesterday was kind of like a big news day for the GPU rich companies, both Inflection and Stability having sort of wind taken out of their sails. I think, yes, it's a data point in the favor of Like, just because you have the GPUs doesn't mean you can, you automatically win. And I think, you know, kind of I'll echo what Alessio says there. But in general also, like, I wonder if this is like the start of a major consolidation wave, just in terms of, you know, I think that there was a lot of funding last year and, you know, the business models have not been, you know, All of these things worked out very well. Even inflection couldn't do it. And so I think maybe that's the start of a small consolidation wave. I don't think that's like a sign of AI winter. I keep looking for AI winter coming. I think this is kind of like a brief cold front. Yeah, it's super interesting. So I think a bunch of A bunch of stuff here. One is, I think, to both of your points, there, in some ways, there, there had already been this very clear demarcation between these two sides where, like, the GPU pores, to use the terminology, like, just weren't trying to compete on the same level, right? You know, the vast majority of people who have started something over the last year, year and a half, call it, were racing in a different direction. They're trying to find some edge somewhere else. They're trying to build something different. If they're, if they're really trying to innovate, it's in different areas. And so it's really just this very small handful of companies that are in this like very, you know, it's like the coheres and jaspers of the world that like this sort of, you know, that are that are just sort of a little bit less resourced than, you know, than the other set that I think that this potentially even applies to, you know, everyone else that could clearly demarcate it into these two, two sides. And there's only a small handful kind of sitting uncomfortably in the middle, perhaps. Let's, let's come back to the idea of, of the sort of AI winter or, you know, a cold front or anything like that. So this is something that I, I spent a lot of time kind of thinking about and noticing. And my perception is that The vast majority of the folks who are trying to call for sort of, you know, a trough of disillusionment or, you know, a shifting of the phase to that are people who either, A, just don't like AI for some other reason there's plenty of that, you know, people who are saying, You Look, they're doing way worse than they ever thought. You know, there's a lot of sort of confirmation bias kind of thing going on. Or two, media that just needs a different narrative, right? Because they're sort of sick of, you know, telling the same story. Same thing happened last summer, when every every outlet jumped on the chat GPT at its first down month story to try to really like kind of hammer this idea that that the hype was too much. Meanwhile, you have, you know, just ridiculous levels of investment from enterprises, you know, coming in. You have, you know, huge, huge volumes of, you know, individual behavior change happening. But I do think that there's nothing incoherent sort of to your point, Swyx, about that and the consolidation period. Like, you know, if you look right now, for example, there are, I don't know, probably 25 or 30 credible, like, build your own chatbot. platforms that, you know, a lot of which have, you know, raised funding. There's no universe in which all of those are successful across, you know, even with a, even, even with a total addressable market of every enterprise in the world, you know, you're just inevitably going to see some amount of consolidation.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.