Ramtin Naimi discusses compute needs in venture capital (Part 3)

Generation prior, let alone current generation Bear Rubin or something. Allocation from Nvidia. And the price of GPUs just looks unidirectional. It seems to me like a very bizarre moment in the market. The thing that we're seeing in our portfolio companies is that we are routinely helping them find compute, because finding compute for them is not easy. A lot of the largest NEO clots and we have relationships, the number of them aren't really interested in engaging with compute deals that are like sub $200 million in scale. And so companies that don't have that level of scale, or you trying to lease some capacity here, some capacity there, but those guys on a relative basis don't actually make up for the scale of this market. The people that are actually moving the needle for how much compute is necessary in this ecosystem are the hyperscalers, the Frontier Model labs, and then the companies that are private Neo Labs raising $1 billion plus seed valuations effectively. I think the only thing that limits demand for compute over time is if there are compute needs from these companies diminish, or the companies that have these compute needs have a harder time raising capital. There's a subset of companies that, you know, I know that are going to continue raising because they're the best assets and they'll have all the cash they need to buy all the compute they need. I think the most interesting segment of the market, which on a relative basis is still small compared to the hyperscalers, but if you aggregate all of them together, I don't think the scale is that small as sort of the neolab segment where people started looking at Neolabs thinking there was going to be like two or three great ones. And now there's probably 80 or 90 NeoLabs that have raised somewhere between 1 to 5 billion dollars. And the funniest thing with them is they all need to raise tons of cash. We met a neolab a couple weeks ago that was raising a $1 billion. I think this was their second round of financing and they were going to spend the entire billion next year on COMPUTE for one training run. And they said they would probably need to raise another 5 billion throughout next year. The tricky thing with these companies is it's very hard to underwrite how well they're doing. The funniest thing with NeoLabs is you can underwrite how well they're doing this and how much money they're raising. But oftentimes they're launching things that are very impressive. Oftentimes it's hard to discern whether they're launching something that's truly impressive on a generalizable scale, or if they launch something that was specifically tuned to perform really well on benchmarks. So that's the whole benchmark maxing concept of whether these things are truly groundbreaking and innovative, or they did something that shows really well on the benchmarks but isn't really commercializable yet. It's not going to be commercializable for the near future, but I think there are a lot of those companies that have massive capital needs, and it'll be interesting to see if all of them continue to raise. If.