AMP launches $100B Grid Trust with Zurich Insurance to give startups access to pooled GPU compute

Speaker 2: Don't just build AI. Own the data platform that powers it. And let's bring in our first guests of

Speaker 4: the show. Welcome

Speaker 2: Boom. To the show. Thank you so much There

Speaker 4: they are.

Speaker 2: For joining. Let's start with some introductions. Knows who's who.

Speaker 4: Hey, guys. Go ahead, William.

Speaker 11: Hi. I'm William Chew. I'm the head of private equity over at Zurich Alternative Asset Management. Thank you guys very much for having us here. Long time fan, first time joiner, diet coke enabled.

Speaker 1: There you go. Cheers. Great to have you. About these other two? Who are they?

Speaker 4: Hey guys. Good to see you again. Some of you know me. My name is pronounced Anjney. Friends call me Anjney. You guys should feel free to as well. And you got my partner Negin over here. Hey there.

Speaker 12: Hey. Negin Oliver. Nice to meet you guys.

Speaker 1: Amazing. Where are guys calling in from? Side of the road somewhere? Data center?

Speaker 4: You can... You Does it

Speaker 2: look good? Is it obvious? Isn't that the Golden Gate Bridge?

Speaker 1: I mean Yes. Your hat was kind of blocking it, I gotta say. At least at the start. But now now I see it. I thought you were gonna be somewhere down in like West Texas or something like that. But it's amazing to have you guys.

Speaker 4: Well, aren't aren't federal land. We're at the Presidio in which is where our campus is. This is where AMP is headquartered. Cool. And and so we just stepped outside for some sunshine and to talk to you guys. Thanks for having us.

Speaker 1: Amazing. Well, why don't you kick it off, Anj? Yeah. Why are we all here today? And why is this... Why will this day be remembered throughout? Why will it echo throughout Yeah.

Speaker 4: Well, as you guys know, compute is getting expensive. And as sheriffs in the AI infrastructure world, we started getting a lot of calls for help last year from startups who wanted access to compute at scale. But because they weren't as rich as the hyperscalers, they were getting told they could not access the compute they needed. So they called us in. We're the cavalry, and I'm excited to announce today that we're kicking off a $100,000,000,000 program called the Grid Trust that's been anchored by Zurich. Mhmm. Zurich Asset Management is one of the biggest institutional investors in the world. They have over 300,000,000,000 of assets under management, and they have anchored the Grid Trust program with a meaningful amount of capital that we are using to bring up capacity... Compute capacity for startups all around the world. And we're really lucky that we had Negin join us from AMD to lead this effort when Negin Negin was at AMD for over eight years and helped build the GPU business over there from scratch. So between Negin and William, we're in pretty good hands.

Speaker 2: That's great.

Speaker 1: Where should we start from here?

Speaker 4: William, I'm gonna hand it over. I'm gonna call on William a little bit to explain

Speaker 2: Perfect.

Speaker 4: What Zurich does. I think the the big question every time you guys have asked me, Anj, do you want to build out two gigawatts? Yeah. Your question has always been, how are gonna pay for it? And so the good news is we've got William to answer your question.

Speaker 11: Well, that's a that's a heavy ask there. But, look, happy to give you some perspective from our lens here and just for some context as well. What I do here over at Zurich Alternative Asset Management. We are the alternative investing arm of Zurich Insurance Group. Obviously, you know, it's a publicly traded company. I will slightly correct Anj. We have a little bit more than 300,000,000,000. It's actually 440,000,000,000. Woah.

Speaker 4: There you go.

Speaker 1: Oh, wow. It's just a rounding error. $100,000,000,000 rounding error.

Speaker 11: You know, what's the... What's that among friends. Right? But...

Speaker 4: Exactly two gigawatts. Isn't that convenient?

Speaker 11: Yeah. Look, we are super thrilled, super excited with our partnership here with AMP. And Anj is a a dear friend of mine and the team as well. I've got to know them for some time and really excited about this ambition endeavor. Super proud to have this opportunity and privilege to kind of showcase with you guys live today, as you said. This is kind of a a more... A very memorable day for us and, obviously, for hopefully the future of compute. I mean, there's a number of reasons why we're super excited about this opportunity. And, you know, I think from a perspective of what really kinda got us over over the hump, so to speak, is a a few reasons. One, I would say is Anj has been kind of the lead catalyst around the whole world of compute, he's he's looking to change the game in venture, in in the very core and what the essence of it is. I cannot be more excited about partnering with someone like that who is not just to kinda be yet another venture firm, because I think the firm... The reality is, you know, we've seen a lot of those folks out there, but I think for those that really kinda have an ambitious plan to really change the DNA of what it means to be a venture capitalist today in today's environment, I think that's one of the reasons why we're super excited. And I think from the specific opportunity set here, I think a few reasons. One is I'll break it down into a few categories, so I'll try to be brief. Right? But, you know, the spending curve. One is people often think about the frontier AI labs and the compute. It's always speculative. But, frankly, the infrastructure side behaves a little bit more like a utility than a start up. And that's, I think, a testament to what Anj has been kind of evangelizing, and we're a believer as well. I've... Obviously, some data and research has kinda shown that. I think model capabilities have improved. I mean, we know the obvious ones, particularly on the frontier side. And likewise, higher compute consumption has followed. Right? So labs may compete, models come and go and change, but the demand, frankly, for the compute has followed consistently. So as investors, institutional investors at that, we're looking for a pattern of spending that's durable and observable. Right? Mhmm. A few other things I'd just highlight, you know, I I think that we've seen what Jensen's been doing, particularly around the NVIDIA chips, and really just kind of making a big splash around info. All the various different parts of infrastructure being now financeable because of this pattern of spending. Right? And there's demand, and you get reasonable confidence in that. So that helps. We're not relying on any single model provider to win the game, and this is going back going now into my third piece of it, why we're super excited, which is pooling. Pooling actually reduces the risk. So instead of kind of putting all the eggs in one basket where you have a single tenant build out, you know, you're basically going for the Hail Mary, one customer run more roadmap, one growth trajectory, pooling actually changes the economics. It actually helps diversify. And from a risk perspective, as you can imagine, I work at an insurance company, so risk is a a natural big theme. This is actually an incredible diversifier here. So and last but not least, you know, I would say, why does this matter? I think we all know, like, look, AI is not fat. AI is long standing here, and we're we're investing here not just in in the promise of compute and what it can bring, but, you know, in terms of the high utilization that it, of course, compute can deliver over a long period of time. And so we're super excited, and, of course, it starts with AMP and our partnership with them. They are... I mean, Anj doesn't need yet another plug. You know, if you if you follow Anj, he's successful in his own right.

Speaker 13: He's well

Speaker 11: spoken, and he's got his x account for those that don't follow, they should. It's quite entertaining on many levels. But I think what's important as well is there's a lot of players in this space, but you have to have the right one. And I think AMP is the partner of choice for us, and they're one of the very, very few that we are doing business with.

Speaker 1: Yeah. Very cool. Anj, how are you thinking about timelines here? Do you have a bunch of portfolio companies? You're going to have a lot more in the future. You also have, I'm sure, companies reaching out to you all the time. Anj, we don't work together yet, but I need compute. You're looking to bring two gigawatts online. That's not going to happen instantly. But how are you communicating the timeline today?

Speaker 4: Typically, I introduce them to Negin, and then she takes over. So why don't I turn to her? Because, you know, at AMD, the last thing Negin helped assemble over there was the the two gigawatt partnership each with OpenAI and Anthropic. And I think those will be looked on as templates for scaling across the ecosystem. But, know, Negin, would you answer that question?

Speaker 12: Yeah. I mean... And I think it is gonna take some time. And as Anj mentioned, my focus is really to figure out how we get the two gigawatts online in a reliable way. You know, one of the things that William talked about, pooling for the end customers, there's also value in having multiple cloud providers and neo clouds on the grid. So our customers are now, you know, have a little bit more security and they're less dependent on a single cluster and have have a more of a reliable vision of where the capacity is gonna come online. And whether it's one or the other, it kinda doesn't matter. When people come to us looking for compute, I mean, we have compute live today. The grid works. We have people running on it. It's been unbelievable. We have NVIDIA and AMD. Our hope is to get all chips onto the grid because that's what's really going to make this fungible for for people. And, you know, we've got capacity coming online between now and, say, twelve months from now that we have very clear line of sight to where it's gonna be, how much it's going to be. And it's a a process to educate our customers on how best to utilize that capacity because our goal is not to have them lock it up and just sit idle. The goal is to make sure we get the most utilization out of those GPUs that we have on the grid. So that's really what my focus is. I've spent my entire career in the semiconductor industry focusing the last eight years just on the infrastructure side for AI, and so that's what's complementary with what Anjney's bringing to the table and looking, you know, looking forward to getting this built out and getting our labs access to GPUs at the way they need it, really, when and where they need it.

Speaker 1: Amazing.

Speaker 4: So next time any of our friends listening in wanna know exactly when their next cluster is gonna come online, all you have to do is message nagin@amppbc.com. You no longer need to go... I'm no longer the bottleneck as many will testify. I hope you don't mind me sharing your email.

Speaker 12: Most people won't know how to spell it.

Speaker 14: We'll be good.

Speaker 12: We're good.

Speaker 4: There'll be a lot of A of big ones will find out.

Speaker 12: Yes. Some of

Speaker 15: the big ones will find me.

Speaker 1: Yeah. John, what were you going to say?

Speaker 2: I'm I'm interested to know about how do you think about the diversification, the spreading of risk across different customers at different levels. That seems like it's important to the philosophy here. Obviously, there's a number of labs and hyperscalers and neo clouds, but then there's also start ups. But how do you think about actually spreading out the long tail of demand across all the potential customers? Is there, like, a portfolio strategy This in

Speaker 4: is a great question. So if you go to the website, nationalcompute.com, which is where the grid is hosted, you'll see it says the grid is now in public access. Public sector access. You know, we we started with a public mission Mhmm. Which means the the compute, the infrastructure that's being built out has to go on the grid to people who need it the most. Mhmm. Today, that means for us .users, ..gov users, .milusers, we're... As far as we're we're we're one of the only places

Speaker 2: Mhmm.

Speaker 4: Where these folks can access live, like hundreds of chips on demand at subsidized rates. You can get an AMD chip there, an MI three fifty five for $3 an hour. And you can get a b b 200 from NVIDIA for about $4. The commercial pool we have, there's over 60 different commercial labs also on the grid. They pay a much higher price. So we've had several millions of dollars being spent on m I three fifty five at about $12 an hour as of yesterday. And so just to give you a sense of the spread, commercial customers do pay higher for on demand access, but public sector users get their compute at subsidized rates. And and the benefit of pooling is we get to essentially load balance across these different customer types and ensure folks who need access the most can get it at prices that don't sort of exclude them from the market. Does that make sense?

Speaker 1: Yeah. Totally. Last question from my side for now. How... Like demand planning. It feels like the greatest challenge for application layer companies, even obviously the labs as well. But is this something that can be solved? Or like it feels like this is the solution where you just get the most number of chips in a single sort of platform all over the world. And because it just feels like it feels like demand planning for these companies is entirely unpredictable. You have like an instinct moment where you go from obscurity to, overwhelming demand. Or you've seen labs release a new model and it's you get that new capability increase and suddenly you have customers that you weren't even imagining. And so this feels like the only solution, but what are you guys seeing the smartest companies do around around forecasting? Is it really just napkin math and sort of like guessing? Or or is this a new kind of core skill set?

Speaker 12: I I think each company on their own is doing the napkin math and guessing. And they're looking at their neighbors. Like, they're looking at what is this guy doing? What is that guy doing? And trying to, like, extrapolate that. I think the beauty of what we're doing is is that our grid members are sharing their napkin math with us, so we get visibility across a larger number. And then we're the ones taking the bets on some of the unknowns. Right? Like like the instinct moment you mentioned, you know, we're gonna plan for that so that there is capacity for those that didn't know better than to plan. And the idea is that if you have enough users across it, some may plan too high, some may plan too low, but you should be able to sort of smooth out that demand curve and make sure that the people who need it most have access to GPUs at the time they need it.

Speaker 4: And and like everybody else, we also use AI. So the first thing you'll see when you log in to nationalcompute.com is our agent called Marshall.

Speaker 2: Mhmm.

Speaker 4: Marshall is named after, you know, the the the general who helped unite The United States and Europe post World War two after the Marshall plan. Mhmm. And Marshall is an automated system that input ingests all of that data that Negin was talking about and then actually does demand forecasting for us. And every good member can go talk to Marshall, do... You can kind of think of Marshall as your chief AI compute officer. And we're using reinforcement learning to improve it. And over time, I think this will really unblock a lot of bottlenecks for folks who don't have their own infrastructure teams, their own procurement teams. You know, it's not uncommon today for the the largest labs to have a 100 plus people doing this job. But I don't see why we can't relieve some of that stress through automation.

Speaker 2: Fantastic. Thank you so much for coming on the way down.

Speaker 1: Milestone and and so excited that a lot more companies will will get access to this to this compute. And and, yeah, it feels like we had this beautiful sort of hacker moment, hacker culture emerging and then the sort of like financial realities of like how do you access massive amounts of compute as a small team kind of you run into that. So Yeah. I'm very excited about this and and congratulations on the partnership.

Speaker 4: Fantastic. Well, big shout out to Zurich. It takes a village to raise AI infrastructure. And as you can see between William with his background and us over here in San Francisco. William's actually based in New York. We think this has to be a cross country scaling approach, and we're really lucky to have Zurich believe in us at scale.

Speaker 2: I love it.

Speaker 1: Amazing. Well, looking forward to having all three of you back on the show soon. Yeah. And good luck out there.

Speaker 2: Have a good rest of your week.

Speaker 1: We'll wear our cowboy hats next time.

Speaker 9: Thank you.

Speaker 1: See you. Cheers.

Speaker 2: Let me tell you about Cisco. Critical infrastructure for the AI era unlock seamless real time experiences and new value with Cisco. Our next guest is Neil Stephenson, the fantastic author. Know and love his latest book is d heavy water. Here he is.