Short answer: Dov Gertz says it runs exactly like the LLM market, though compute is not yet the bottleneck; one breakthrough could change that. He says pharma token budgets are rising and molecular-level demand will explode.
Watch this answer from Episode 32 (starts at 16:30) · Episode 32 notes and transcript
From Episode 32: From DNA to Drugs: How AI Is Rewriting Human Biology, starting at 16:30.
Jordan Metzner [16:29]: Awesome. And just one last question here. Not to know too much about how these foundation models work, but are these all running on top of NVIDIA? Is, is NVIDIA's innovation kind of a downstream impact to you guys as new NVIDIA chips come to market? It allows you step function improvements. Is it NVIDIA chips are not optimal for this type of learning and transformation? Or just tell us a little bit about how it may differ or be the same as kind of how the LLM market is.
Dov Gertz [16:58]: Exactly the same. No difference at all. I'll just explain that within this the foreseeable future, that isn't the bottleneck for this industry, but it might tip very, very quickly. So this is an extremely dynamic market. We're one breakthrough away for it to become compute bottleneck and see the scaling laws that we see in text image and audio. But we're we're in my opinion, we're not there, and no no no modeling organization is there yet, but we're a moment from there.
Jordan Metzner [17:29]: Yeah. I mean, I know for us I mean, sometimes it's hard for us to get GPUs to run trainings and things like that as the market is, like, insatiable demand. I can imagine as these pharmaceutical companies start to really hit the GPU consumption, start really eating a lot of tokens, you can see kind of an additional, kind of player taking chips out of the market as they're using them as well. Right? So more people all with the same, supply chain, I guess, so that'll kinda increase demand and decrease supply maybe.
Dov Gertz [17:57]: Yeah. True. We're talking about one of the largest industries in the world, and this is what we believe to be the largest financial opportunity in the history of life science. So once I mean, we're already seeing this in consumption by pharma of textual and image models within their context. Okay? So for filing better regulatory affairs and analyzing data and whatnot and designing clinical trials, a lot of that is through textual and image related models. So we're seeing a massive increase from budgeting in a on a token level from pharma companies. But I'm talking about the molecular level, that's gonna explode, and we're gonna see a whole new, demand emerging from pharma there.
Jordan Metzner [18:40]: Yeah. Totally. It just seems like there's not enough chips to go around, or at least there won't be for a while.
Dov Gertz [18:44]: I don't think that's gonna be the massive bottleneck going forward. But yeah. I think it's obviously a bottleneck, but
Jordan Metzner [18:52]: Yeah. Yeah. Well, the data centers will get built. The power will come. All those things will happen. So, yeah, I'm sure that it'll it'll fill the market's needs as they come along.
Dov Gertz [18:59]: I think we're gonna have to solve this as a species because it's just so so necessary for our day to day currently. We need to build those fun and computer games in a minute. So yeah.
