Wayne Nelms, Ornn | theCUBE + NYSE Wired: AI Factories

0:00 Palo Alto Studio Connection Silicon Valley and Wall Street. I'm John B here with Dave Volante, my co-host. Welcome back to the Cube Studio here at the New York Stock Exchange. I'm Jim Allen, co-host of NYC Wired, and today we are talking all things AI factories. Specifically, what happens when our infrastructure becomes its own financial market. Joining me now for a conversation on exactly that is Wayne Nems, CTO and co-founder of Orin.

0:32 Welcome, Wayne. >> It's great to be here. >> So, you've been on the show a couple of times this year. We've met in a couple of continents. We've definitely had a few chats, Wayne, but it seems as though every time I meet you, something new and exciting and revolutionary is happening. So, >> maybe fill us in on what's been going on with you and Kush and the team at Orin since we last had you on in May.

0:51 >> Yeah, so I think we last talked in May. like you mentioned a lot of things have happened since then both in the news and for our company. I guess most notably we had and announced our seed round fundraising led by entre crypto and the Galaxy digital team which is an incredible opportunity for us. I think being able to raise a lot of capital was a great of course for the signaling reasons but also allowed us to really scale the business which is super valuable for us at this stage just in terms of hiring in terms of image and brand but also I think even more important is how the market has evolved since we last spoke yeah >> 40 million seed round right >> roughly 40 million yes >> in around I mean that's a big seed like you know in the world I grew up in that was a very big seed nowadays you know things are changing rapidly but let's talk about where that money is being spent on the I guess how you proliferate your businesses because you actually have two separate businesses in some respects right you have a marketplace >> where you sell GPUs and demand it's like kind of meet meets supply and demand and you also have a data business and I know we're going to talk a little about what's happening with ICE in that business but which is essentially around creating like an index or a commodity for GPUs.

2:03 >> Yes. >> Let's start in the marketplace business. Okay, >> lot happening there. We just today had semi analysis on the show. We had one of the lead analysts, Jordan Nanos. He talked a lot about the hype, the bubble, the fact that there is so much demand, not enough supply, and he, you know, we don't really know when that's going to slow down. >> What are you seeing? Where are you guys sourcing from? Who are you matching?

2:27 Talk about the the beautiful minds that you're connecting here. >> Yeah, so on the exchange side, we have a comput exchange like you mentioned. and that's it's opposite of the data business that we can go into. But I think on the compute exchange side, what we're really focused on is solving the immediate need. And the immediate need in this industry is that there's just like you mentioned so much demand for compute capacity. there's so many smart people raising a lot of capital to train their models to run inference workloads to really deliver AI and the power of AI to enterprise and the individual level right through application layer etc. and there's with that growing demand, there's just not enough supply of compute to keep up, right? It's in today's age, it's not trivial to build a new cloud or to stand up a GPU, right? Not only are there bottlenecks across the supply chain, but in our opinion, one of the biggest unknown bottlenecks or one of the biggest unmentioned bottlenecks until recently is this kind of ability to finance the hardware itself, right? So historically how it's worked is everyone that's financing a GPU call it a neocloud or hypers scale etc needs to finance that GPU against the committed contract of the offtaker right the person that's buying capacity on that piece of hardware and more specifically financers tend to be a little wary of AI risk at least today and they only look to finance projects that are backed by either back stops from Nvidia Google etc. ETA or committed contracts from you know Amazon, Meta, some of the larger hyperscaler players right and if you can imagine in that sort of environment financing can be very difficult so what we're hoping to do in our exchange side business is really grow the market for people that can access this compute >> and who do you think is being somewhat left out or left out in the cold by this cycle we hear a lot about you know the favorites economy right folks are getting into bed together fast in certain space within tech, you know, there's a pecking order even for, you know, GPU access for supply. What are you seeing? Like what sort of, I guess, short tail and longer tail opportunities are you guys considering?

4:39 >> Yeah, certainly. So, there's a huge unserved market in this space and that unserved market is kind of who we target right now. It's those that have raised significant capital to buy compute capacity yet don't have the balance sheets or the credit history to offtake that cap that that capacity quite yet. Right? Right? So you could think of AI labs, neolabs, you could think of you know startups and those that just need access to compute whether it's bare metal or through some virtualized layer.

5:07 these are all the players that you know have potentially the capital and potentially the interest in compute but again they might have only been around for the last 3 to 6 months or you know under a year. underwriters might not be comfortable you know underwriting a billion dollar plus cloud facility to one of these or a few of these offtakers in a multi-tenant system. So what opportunity there is for us to bridge that gap we try to solve and help and then you know how can we help actually manage this risk. I think that leads us to the data business and a bit of the index side but I'll let you take us there. So let's go there because one thing we know underwriters really don't like is uncertainty, right? They want financial predictability. They want to know if you are underwriting a loan or investment against capex in a data center business that they know it's going to cost them 3 years out and that that CFO or that company know what it's going to cost and there's a lot of uncertainty and volatility in that space. talk about this, you know, I guess movement that you guys are building and the relationship that you're developing with ICE because really what you're saying is GPUs are a commodity, right? There should be a predictive pricing index.

6:18 >> Yeah. So, I think certainly financers hate excess volatility especially in markets where they're underwriting huge deals. I think for us, right, what we've always seen in AI and AI markets is that uncertainty is rampant, right? When we first entered the space roughly last year, what we wanted to solve initially was the fact that no one knew what was happening with AI markets, right? Where is the future of AI? What is the future of course cost of compute? What is the future cost to deploy a GPU? And what is demand and supply look like right in the future?

6:54 And I think we've seen historically since we started the business that demand has been nothing but rampant. it the growth and the adoption of AI has just been on a tear, especially recently. And like I mentioned before, you know, supply just can't keep up. However, that's not that story might not be good enough for a lender in a one-off project, right? They want to see committed capacity in their specific investment, of course. And so rather than necessarily finding a long-term contract, right, what we hope the future looks like is not only finding committed offtakers for a project, but also for the financer, for the lender, for financial players to be able to hedge some of their exposure on financial markets. I think, you know, we took a look at how all financial markets have developed, especially in the commodity space. If you're, you know, let's say a corn farmer, you're able to pre-ell your corn before you've even planted the seed, right? And I think in the future, what we hope to make a reality is the ability to sell future capacity potentially even before deploying the GPU.

7:58 >> well, let's stay on coin for a second and let's use that as a good example, right? For what's funable and what's not because, you know, there's a value to output, right? There's a market price, there's an expectation what you spend versus what you consume. We think about the world of GPUs and compute, it's very different. Like some folks say it's not fungeible and some of the metrics that are being used right now to develop a level of fungeibility like you know GPU cost per hour etc aren't really accurate because they don't take in things like latency performance overall efficacy.

8:30 What is your what are your thoughts? What's your response to that? Yeah, certainly. I think at or we've always believed that compute is, you know, of course it's a commodity in our eyes, but it can be different, right? But I don't think those two things are necessarily mutually exclusive. I think you look at a lot of commodities markets, right? For example, corn. It's hard to say that all corn is the same, right? however, we've implemented benchmarks and kind of standards for what a traded commodity should look like, right? It should meet these grades and should meet these characteristics.

9:00 You know what we try to do at or is something similar right? So when we compile our index people always wonder you know what is our index comprised of how do we calculate it and all of our methodology is available online on our website but what we end up doing is a kind of a very standard volume weighted average pricing metric. It's what you would naively assume an index to be. We ingest so much data on pricing for a certain quality class of compute. We look at only compute capacity that's been sold that hits a certain minimum across a few different specs call it memory networking you know performance etc. And after we compile all that all those prices, we just output the average, right? And effectively what we want to do is really represent what the current market pricing for a GPU hour is across all these Nvidia chips. And I think that is not only something that we are focused on, right? We want to have the most representative index, but of course in order to bring in lots of liquidity, I think a lot of the market participants are looking for such an index that does track reality.

10:02 >> Let's stay on Nvidia for a second, right? Like if we think about a comparison of Exxon and Brent crude right like it provides a level of democratization too for buyers for sellers predictability that that's great right did Exxon want that to develop as it to develop and what are your thoughts in the perspective of you know there's a lot of ambiguity out there in this market it has been very beneficial to some of these titans of industry like Jensen and the team at Nvidia do you think they want to see a level of indexing financial predictability Yeah.

10:35 >> What are your thoughts? >> So, we really believe that our product is super beneficial for Nvidia specifically. So, you know, in the last two weeks or so, Nvidia and Jensen released a statement regarding the financialization of Nvidia compute. And I think, you know, what you saw in that piece was the introduction of traditionally, you know, financing players, financing giants step in and say, look, we're happy committing capital to help finance this revolution, right? to help finance the clouds that are deploying Nvidia GPUs and hardware.

11:05 And I think that really brings into the forefront of our minds the real value of not only hedging products but the Nvidia ecosystem as a whole. I know there's a lot of conversation about the the strength and the dominance of Nvidia you know across the software across the hardware performance and we certainly agree we see that of course in terms of adoption in the compute markets right a lot of people are deploying Nvidia chips and still but the additional moat that Nvidia has today is that because of their adoption financers are happier underwriting the Nvidia GPU hardware right they've just had more reps they understand potentially how this GPU trades over time and how the compute itself trades over time. And so what we actually see is that when we are launching Nvidia compute indices, what we're allowing people to do is hedge Nvidia exposure and Nvidia GPU compute exposure, which if anything should allow financers to better underwrite this equipment and in theory allow more people to access Nvidia hardware and spread and continue to spread the Nvidia kind of ecosystem. And that will be happening here in with your relationship with ICE. They will be hedging against that Nvidia spend. Talk me through when that will happen, how that will happen and also what kind like the broad spectrum of data points that you use to ensure that that you know index continues to be as accurate as it can possibly be.

12:31 >> Sure. So our partnership with ICE is amazing for us, right? It really helps institutionalize or be one of the first steps to institutionalize our compute index of course and the compute financial economy as a whole. the launch date is in the fall by end of year, you know, pending regulatory approval of course. and then in terms of like the index itself, what we're very committed to doing is compiling as much data as possible. I think what we are really focused on is being as wide breath and depth as possible. We want to effectively allocate and aggregate data from across the world.

13:07 And the reason for that is comput is global, right? Not only is it global, it trades 24/7. And in order to really represent and help hedge risk for the people in this system and in this economy, what we want to do is get as much data as we can across the entire world. so that's what we're committed to doing. >> So circa 40 million raised. talk about where you're going to spend that, where are you guys investing, what does the product road map and the cultural and team road map look like for the next kind of 6 to 12 months.

13:37 >> Yeah, so we're very committed to building one of the best teams in this, you know, not only AI but also the financial spaces. I think one thing that we've learned very quickly as a team is that, you know, it was great when it was just the four of us and then it was great when it was just the six of us and, you know, now we're at 22 people and it's been an incredible ride. and just being able to grow with the team and see how this market has changed over time kind of acting as a tailwind for our thesis has been really incredible.

14:02 So of course one thing is focusing on the team focusing on building talent internally and then also thinking a little bit more about external market dynamics. What we really want to do is continue building partnerships that help accelerate our business and really thinking about the next steps for growth and partnership across the entire compute landscape. And last question, Wayne, growth and partnerships. This is an interesting go-to market strategy, right? You have a lot of different profiles of buyers and stakeholders and advocates here. How do you think about that? Like how do you guys think about building something that's both a marketplace and somewhat of a movement?

14:38 >> Yeah. So, I think there's a lot of interest across like you mentioned so many different kind of groups of people, right? You have the financial players, the banks, the traders, market makers, etc. but of course in our own industry we have all the clouds we have the hyperscalers we have big labs and the frontier labs specifically inference providers etc. there's so many people that have such a vested interest in what happens in our space and so one it's obviously very exciting for us being able to sit in between kind of the most interesting and impactful sectors I think in the economy right now but of course that also gives us the ability to think about strategic partnerships and long-term where we want to take those so for now I think it's still a very open question where we go specifically but I think in general we're very excited for where we are >> well we are certainly excited to continue to watch this journey evolve Wayne, great to have you on the cube.

15:28 >> Yeah, thanks for having me here. >> I'm Jim Allen here at the Cube studio at the New York Stock Exchange. This is AI Factories, one of our programs with NYC Wired, where we connect Silicon Valley to the great minds here in Wall Street. Thanks for watching.