Why the CEO of Compute Exchange believes compute will become more valuable than oil—and how his company is preparing for it.
Subscribe: iTunes / Spotify
Apply to be a guest on our podcast here
Show Notes:
In this episode of Mission Matters, Adam Torres interviews Simeon Bochev, CEO and Co-Founder of Compute Exchange, about the explosive growth of AI infrastructure and the urgent need for transparency in computing resources. From leading AI initiatives at Apple to founding a startup tackling inefficiencies in compute marketplaces, Simeon shares his mission to build a more open, efficient, and scalable infrastructure for the AI-driven world.
About Simeon Bochev
A serial AI founder, and former Apple and Lambda Labs executive, Simeon has been at the forefront of the compute market as one of the largest compute users and providers, driving innovation in compute since its early days.
Full Unedited Transcript
Hey, I’d like to welcome you to another episode of Mission Matters. My name is Adam Torres, and if you’d like to apply to be a guest in the show, just head on over to mission matters.com and click on Be Our Guest to Apply. All right, so today I have on Ian Boche and he is the CEO and Co-founder over at Compute Exchange, and this interview is part of our.
Milken series, our Milken Conference, global series, where we’re talking about talking to and for attendees, speakers, panelists, and other individuals that attended the the Milken Global Conference in Beverly Hills, California. And first thing, first, Simeon , welcome to the show. Thank you so much.
I’m so excited to be here, Adam. Thanks for having me. So of course I wanna get into more of what you’re doing over at Compute Exchange and also go talk about your journey from apple to AI and what you’re doing at Compute Exchange. But first thing, first Milken Global Conference. So I understand that you spoke at the conference, is that right?
That’s right. So the Milking Global Conference happened in LA I think about a month ago now. Mm-hmm. And I was asked to, speak in a, private session about the growth of AI infrastructure and the the strain that that’s going to put, not just on our, energy grid, where, you know, AI infrastructure is projected to grow, to consume up to 12% of the US energy grid in just the next few years.
An increase from 4% last year, but also just there’s a lot of, talk about the tremendous amount of money being spent on infrastructure. 500 billion on the Stargate Initiative announced by OpenAI Oracle, and SoftBank. But then you hear it hundreds of billions of dollars each from each of the hyperscalers.
And, you know, the audience there was looking to understand what’s actually going on. Hmm. What were some of the key takeaways that, were there? Obviously everybody listening may not, I’m trying to give people that, you know, behind the scenes of what took place and some of that info, they haven’t had the, privilege of attending or didn’t catch your, talk. What were some of the takeaways? Definitely. And I’ll say that you know, my, personal mission, but also the mission of the compute exchange is to, bring more transparency and openness around this entire computing infrastructure and computing economy piece. So very, much in line with that.
I would say the, the three main insights were one, you can both have too much compute and not enough at the same time. We can unpack that more if you’d like. Which is said more simply, there’s still not enough capacity out there and we need to be investing more money more than what’s already been announced.
In fact. Mm-hmm. There is going to be a strain on our domestic national, but also international resources from anything from semiconductor supply chains to power generation capacity to meet this demand to the human, the, high skilled human labor necessary to build these large data centers and operate them.
Then third is that we’re still very early days despite what some of the rhetoric on online might be in, the news. We’re still very early days and there’s a long runway ahead of innovation and developments in AI and AI infrastructure ahead of us. have you been to the Milking conference before?
Is this your first time? Have you been, have you attended before? No, this was my first time. I believe I was invited to attend one of the regional conferences, but wasn’t able to. I love talking to first time guests. I’ve been going now three years, going on four, I think four years.
Three or four years, I forget. And it’s, it’s one of my favorite times of the year. I just like, it’s my, it’s my Super Bowl conference is what I call it. Some people I’ve heard call it like their Disneyland of, and I’m like, come on, Disneyland. But just ’cause of the brain candy, like the different sessions, the people that are there, such as yourself, that are experts in their field that don’t always, you can’t always just.
You know, pop into a talk and, you know, be educated in that manner. What was your thoughts as, I mean, as a first time attendee and speaker? Well maybe being a little lighthearted. I haven’t seen as high of a bar tab limit at another conference as I did at Milken. But there, are quite a few I think deep pocketed individuals and companies there.
But more seriously, I think the kinds of as you say, brain trust or concentration mm-hmm. Of just. Pure knowledge and very deep, not, not just broad, but deep knowledge and expertise in, in areas was tremendous. And for me it was it was a welcome change from most of the events that I go to, which are here in the Bay Area.
And they’re, tech focused with a very, very smart group of people, but mm-hmm. But very different. I think the Milken lens was. Broader, broader. There was more geo geopolitics there. There was also more around the investing economy and landscape. And I’ll just briefly say, I mean I was walking out, out after a session and I.
Just literally bumped into Eric Schmidt just walking around. Mm-hmm. Yeah. A conversation as we walked to the next thing. And candidly, I was trying to get a meeting with him for a while, and then we just met. Yeah. Wow. That’s amazing. And that, does happen. That does happen. And that’s one of the conferences where something like that would happen.
And that’s why you notice how tight that security was, right? Like that’s one of those things about that conference is that security, I’ll tell you, you ain’t getting in or near if you’re not you’re not supposed to be there. Yeah, I mean, I think there were several secretaries as well as heads of state, especially from the Middle East there.
So it wasn’t a, an easy experience to get in. But they did treat us speakers pretty, pretty well. Absolutely. Let’s get into your journey a little bit more. So, what inspired you to shift from, you know, AI building AI infrastructure at Apple to, compute and launching compute exchange? Like, how’d that take place?
Sure. Fuel permit trip down memory lane just for 20 seconds. I actually had privilege of kind of falling into AI back when I was 15 in 2005 when I was working on some early statistical models. Before, you know, the transformer paper was written, I. Back at Sandia National Labs.
It’s one of our nuclear research labs. But, you know, if we fast forward, I think probably most interesting for your, listeners is my time where I started the first day I researched lab in Japan with a good friend of mine and then transitioned that during COVID, ’cause I got stuck on the other side of the border closures here in the us.
To Apple where I was one of the, the folks leading in the AI infrastructure and platform team there, and we were some of the largest users of computing power in the world at that time. This is 2020 to 2023. And it’s now public that, you know, we work with AWS as well as other cloud providers. And I can tell you that.
Having that front row seat to such large computing workload demands. I mean, we won’t talk about what Siri is today, but Siri was the first at scale AI product in the world. Hmm. Over 300 million users well before chat GBT hit a hundred million and kind of getting to see that I got. Excited but also frustrated by how inefficient and difficult it was to procure and then use compute.
And I’m sure we’ll talk about some of those issues. But to finish the kind of the journey I got approached funny enough on LinkedIn by Lambda Lambda Labs, which is one of the leading neo club providers, folks I’m sure familiar with Core we that went public recently. what was initially a sales call to try and convince Apple to, work with Lambda?
Turned into, I like to say, they got stuck with me instead. Oh, that’s funny. They joined their team. Yeah. That’s awesome. I, I mean, why did I. I joined Lambda because I really believed in their mission of being, you know, building four ml engineers by ML engineers which you don’t often see in, in the cloud provider space.
And I thought that we could really disrupt the paradigm of the incumbents. I. And what I quickly learned is that some of those same structural issues I saw while at Apple were really, really hard for any individual company like Lambda to solve, which is why serendipity happened. Kind of like running Derek Schmidt, I got invited to a dinner with John Wilson, Don Wilson.
Don Wilson is the founder of DRW, one of the largest commodity trading firms in the world. And at dinner we talked about, building a market for compute and how that could solve some of these structural issues. And Don’s built multiple markets in his career and after two months or so of talking about it, we decided to do it.
And so that’s what we’ve been doing since April of last year. what an amazing story is interesting how those, how you connect the dots on that one. And I’d say when you think about the space that you’re going to, like, how did you know the market was ready? And if you think about you know, like just the size, like could it surpass oil, all these different things?
And how, how did you know that the market was ready Or is it ready, I should say? Well, time will tell. I wouldn’t be doing what I’m doing if I didn’t have conviction that the time is right. Yeah. you said something that I like to say. It’s kind of provocative, but you know, the oil and gas market is roughly two to $3 trillion per year, and at least in my life since 1990, I can’t think of a single force that’s led to more.
Conflict as well as opportunity than, than oil and gas in the world. And that’s a two to 3 trillion compute, we estimate is already at half a trillion, if not more. Mm-hmm. And we easily expect it to surpass 10 trillion at a steady state, especially if you believe that AI will become ubiquitous like the internet has become, and that you need computing infrastructure for, AI to be at, at any sort of scale.
Mm-hmm. And so it’s a huge market opportunity. But right now there’s nowhere that you can Fung trade this as an asset. You’re locked into long-term contracts as a user. A provider has to raise very expensive debt capital. I. And from limited options for debt capital raising, as well as costly equity issuances, just look at core weave and the amount of debt that they have, as well as the amount of equity they’ve had to issue to get to where they are, which is a very commendable place and it’s obvious that something’s something’s off.
We can also talk about how for seemingly the same thing, let’s call it a an H 100 GPU one company might be. Three to five times what another company’s charging. And they rightfully so, might, might deserve to charge that, but it’s next to impossible. Even if you have, as I say, a, a PhD in AI infrastructure, by the way, that degree isn’t offered anywhere.
It’s earned through, I think, hard work in the industry. Mm-hmm. Even if you know the industry, it’s very hard to compare Apple samples. What these different providers are offering. Hmm. what’s your next steps with compute exchange, like, as you can and as the market continues to, obviously the need is gonna just continue to increase and increase like the numbers of the numbers, but what’s your next step like as you continue to scale this and, do your work?
I. Well, I don’t know if you allow repeat guests, but well, if possible, I’ll have to come back because we’re gonna be announcing two major press releases in the coming two to three weeks. Mm-hmm. One is around the launch of a compute standards body that’s going to to work with some of the largest companies and most impactful companies in the, in the AI industry to define the standards for how compute should be deployed, because today.
Too many companies differentiate themselves just by granting you access to compute as opposed to having access be open and ungated. Hmm, and instead profiting off of the value add that they add on top. Think of the analogy of the internet. The internet had the open T-C-P-I-P protocol as well as certain technical standards like ethernet that were agreed to by the main participants at the time.
Compute is very much going in a different direction and ultimately. I have conviction that that hinders the adoption of ai. So that’s one major initiative. Be announcing. And then the other one is really pushing for continued transparency in the market. We launched a pricing calculator on our site, which to my knowledge is the most transparent and trusted dataset for compute in the world that shows folks exactly what the price should be, not what some provider lists on a website, but exactly what the training price should be for particular spec under certain con conditions.
And. We’re projecting what that price could be one month out and eventually we’ll project it up to six months out. So provide and plan for what they want to deploy in terms of capacity and buyers can plan for when they make their next purchase. Hmm. That’s exciting, man. You got a lot going on and this is you’re right.
I, I feel like year, I feel like year over year things are gonna continue to grow. Things are gonna move faster and to me it’s just, I, I’m still a beneficiary in my business. Still a beneficiary of just some of the more common things that I think a lot of businesses out there are just. Workflows and just like, and productivity tools and just, you know, things that make our lives easier.
And I feel like as they go through, there’s a, there’s been a ripple reflect in our, in our organization of, productivity. In a good way. In a good way. And, and it’s actually allowed us, which is kind of, I, I always like to bring this up when I can, and I guess now’s a good time to to kind of fit it in is, it’s actually allowed us to grow and to hire more, like our workforce actually grew based off of the implementation of different tools that were, you know, AI based, based because we, we, we picked up more business, we are more effective, we are more profitable, and we’ve been able to grow because of that.
So, contrary to, you know, some of the things and maybe some other, you know, is it gonna take jobs? Is it gonna do this, gonna do that? So. I’m excited, I’m excited to continue to see what, you know, individuals like yourself doing the heavy, heavy lifting right. That allows, that allows us to, to benefit from that on the, on the, you know, downstream we’ll say.
We, I would definitely say the team here is doing some heavy lifting, but if we want to be even more precise about it, the folks that are in those data centers racking and stacking those servers and chips and literally lifting these things that weigh hundreds to thousands of pounds, yeah, they’re actually doing the heavy lifting and people may never see them.
But I’m not call them 10 day. Come on mean you’re a loud one. Come on, man. Okay. Fair enough. Enough. I had get it you on. Oh man. This is wonderful. Well, well, Simian, this has really been great having you on the show. If somebody wants to follow up, if they wanna learn more about compute exchange or follow your work or your journey, how do they do that?
It’s 3 1, 1 of three easy ways they can email me. It’s my first name, SIM as in Mike, EON as in [email protected]. That’s my, that’s my work email. On LinkedIn or they can go to our website. It’s compute exchange. Perfect. But I, I appreciate the chance to be on, on, on your podcast. Absolutely. And for everybody listening, just so you know, we’ll definitely include the links in the show notes, so you can just click on the links and head right on over.
And speaking of the audience, if this is your first time with Mission Matters and you haven’t done it yet, hit that subscribe or follow button. This is a daily show. Each and every day we’re bringing you new content, new ideas, and hopefully new inspiration to help you along the way in your journey as well.
So again, hit that subscribe or follow button. And Simeon, thanks again for coming on the show. Thank you, Adam.
