# The Dark Pools of Compute — Affaan Mustafa, Alejandro Soumah Fernandez — session 2026-10-14T15:55:00.000Z → 2026-10-14T16:13:00.000Z _39 transcript lines · 47 slides · source: full recording_ ## Transcript **[00:00:38 · 0]** Oh, there we go. All right. I think we can get started here. Hey, everyone. My name is Lafon. **[00:00:50 · 0]** And I'm a lead for Compute Trading at Jump Trading, which is a proprietary trading firm, alongside my co lead over here, Alejandro. So today, we're going to be looking over some of the market structure as it applies to Compute Trading, what Compute Trading even is, what we can see it becoming. And I think we'll start with some macro, look at microstructure, and then talk a little bit about where we see things in policy. So the title of the talk is The Dark Pools of Compute. But some of you guys are in finance, so you might be confused as to why I'm using this terminology. **[00:01:33 · 0]** What do I mean by dark pool here? Essentially, what I'm referring to is an unlit market. Right now, most compute trades and blocks, physical compute here. And these are private quotes. We have a lot of brokers, and these are long bilateral contracts. **[00:01:47 · 0]** Now, at the end of the day, these are really commercial agreements. So we're using the term trading here to look at how it's transacted between parties. These commercial agreements tend to be some mixture of an MSA, an SLA, and an order form, cluster specific with hardware specific language in each order. But when we're looking at pricing and looking at some of these other metrics, a lot of the time, most of the talk is around the SKU, so what the chip itself is, the baseboard. So H100, H200, V300, etcetera. **[00:02:24 · 0]** And so a lot of the negotiation here happens outside the public view. There's no exchange for exchange for physically traded compute that's widely adopted and used for everyone goes there to take a look and trade compute. These are done in chats. Right? They're out of the public view. **[00:02:44 · 0]** There's public marketplaces. There's price references. There's a lot of on demand marketplaces. However, for these actual tenured, so on a given term for a fixed amount of capacity with a certain total contract value, these trade over Slack messages, X messages, Telegram group chats. There's a bunch of OTC channels that are all over in these, I want to call them NeoChat. **[00:03:08 · 0]** And some of you might be familiar with commodities and how a lot of... There's chatting happening on IceChat, BloombergChat, etcetera. This is kind of where the markets here lie. And so I want to look a little bit here on the macro market structure and a little bit on the players here as well. And so this was a recent kind of a16z. **[00:03:31 · 0]** Some of you guys might have seen this. So I'm putting this out there because you can go on a16z's actual X and take a look at this graph yourself. What we're looking at is that the cost of a given capability has fallen sharply. The bill, of course, still depends on the model usage, capacity, and other factors, such as speed is a big thing right now with the new CHaChaPT models. Cheaper tokens unlock more usage, and a team can do more work for the same budget. **[00:03:57 · 0]** So what we're seeing is that a lot of teams are simply just using way more tokens. So we have this latent demand that hasn't really been hit yet, and people are just going to use more with the same budget that they were using before when models cost more, and they're just spending more tokens. So applications that were too expensive become possible. However, as we're talking about compute here, the capacity underneath still has to be built and paid for, even as the cost of a particular task comes down. And so the important distinction is between the price of the intelligence, the number of tasks people want to run, and the amount of hardware needed to serve them. **[00:04:35 · 0]** And so this is where we get to kind of the Javonne's Paradox and a little bit of some of the stuff that we've been looking at lately, where we had a very interesting thing where, you know, intuitively, you'd assume, that Compute and Inference kind of, like, prices together, correlates very closely, as you can see. You can see there's a little bit of a decoupling that happened, and token prices declined further, but compute prices remain robust. And so, like, what can be an explanation for this? Now, there's a lot of, like, potential explanations. One here is the whole Javon's Paradox is that even though token prices and inference is getting extremely efficient to serve on these new baseboards and GPUs, people are just consuming a lot more. **[00:05:16 · 0]** And so the demand for compute rises, and people just consume a lot more tokens even if the cost per cost per million for input and for caching and for output becomes a lot more efficient. And so making the task cheaper, encouraging the additional use to outweigh resource saving for tasks. For Inference Clouds, that's going to be more applications, serving more tokens, bigger market share. And for app layer individuals, that means you guys can serve to a lot more customers, so anyone who's building, utilizing the token infrastructure. And so what really goes into the macro structure now, this is very bare bones here. **[00:05:56 · 0]** But you've got your chips, which we've been talking about. You have power, you have buildings, grid connections, and then of course, you need a lot of capital to get these build outs going. And then that's kind of been the whole talk here. And so chips, powers, buildings, money, they all have to come together before you can even get to an hour of actual usable GPU per hour compute. And so the IEA April update projects that the global data center electricity is going to rise from four eighty five terawatt hours in '25 to 09/15/2030. **[00:06:26 · 0]** And so a lot of the talk right now is around build outs, about getting the Power Grid set up. Will we have enough Power Grid here? Is there other energy sources that we can tap into? Where are we able to get the land to actually be able to do this? And so co location sites and data center operators and people who are doing the financing, the build outs, or the lenders themselves, are all looking into this. **[00:06:47 · 0]** And that type of everything that's upstream, essentially, of the compute is going to then flow down into the compute costs and also usage. And so what does this actual macro map look like underneath a quote? Chip and system vendors supply the GPU, servers, and networking. Sites and operators supply the power, space assembly and management. Right here, GPUs, Memory, Service, and Networking. **[00:07:13 · 0]** Capacity providers, these are hyperscalers, NeoClouds, and any other operator that's, like, operating similar to how a NeoCloud would operate, reselling bare metal capacity. And then the users, so who's the actual end user? The startups, the Inference Clouds, Labs, Neo Labs, and also anyone on the app layer, and Enterprises for anyone doing OpenWAIT stuff as well. And, these boxes, like, I've kind of boxed them out here. But, you know, people can be all of the above. **[00:07:42 · 0]** A chip and system vendor like, someone could do a full end to end build out, and also be the end user of everything. And so there is a little bit of variation that occurs here. And so as I was mentioning about the Inference Clouds before, they might rent a cluster, and they'll serve tokens off of it, and they'll use that revenue to the payoff of the cluster. And so that might be a use case for the actual Inference Cloud themselves. However, for hyperscalers, Frontier Labs, and larger NeoClouds, they can make substantial commitments because they have the ability to get into the large equity, they have the ability to take on more debt, And they have the ability to do long horizon deals. **[00:08:27 · 0]** So ten year deal, OpenAI contracted out, etcetera, so of that nature. However, anyone in the middle market isn't actually able to do this. Who's underwriting a Series B startup with enough capacity to get into $200,000,000 plus contract over three years for Compute? How many equity rounds will they have to raise? Right? **[00:08:46 · 0]** Will they have to use their entire equity round to fund just the actual Cluster, and then they don't have any remaining capital left to use for actual costs. And so we're looking into a little bit of the financing chain here and how that plays in. So lenders are financing the build out against contracted revenue, even, in many cases, for Inference Clouds. And so they want to finance the build out, and they want confidence that their loan will get repaid. I mean, how are you going to underwrite a startup, like a private company that doesn't have much else going for them right now? **[00:09:15 · 0]** They're not publicly traded. There's no that much cash flows, balance sheet, income statement. They might even be pre profit, pre revenue, etcetera. And so what they're looking at is, can the person get contracted revenue over a certain scale? And so, core REITs public filings give large scale example of infrastructure financing supported by take or pay customer contracts. **[00:09:32 · 0]** And so this diagram shows the smaller illustrated example here. And the mismatch is that hardware financing, they want very long terms, but customers are only able to afford or don't really want to get into a long term financing over thirty six months. They have to put up on up to 15% to 25% is what CoreWeave has quoted that they've seen their median prepayment to be 15% to 25 up front on a 100,000,000 PCV cluster. That's $15 to $25,000,000 that a startup has to put up front. That might be their entire round. **[00:10:07 · 0]** And so this is kind of a look at what I'm referencing here. It's like you're raising $40,000,000 you need a $100,000,000 contract over three years you put down 30% prepayments, as an illustrative example you've got $10,000,000 left to run the business. And so you might have this kind of step function, as you can see here. So that's your contracted capacity, and what you can actually and the demand is, like, you know, below that. And so now, every time that you have to get into a new cluster, it's tied to your equity round, or whatever else it may be. **[00:10:34 · 0]** But you might have to cap out and turn away customers. So what are your options? Do you either turn away customers until you can get into your next equity round, so you can put down another 30% prepayment, so you can get the next little bit of capacity going? Or do you use On Demand to close the area under that gap and bridge the shortfall? That's like an option now. **[00:10:53 · 0]** Now, you have to look into the economics. What does that look like? What's the difference between, you know, maybe instead what they do is they go to a lender to finance the prepayment component. But if the yield on the prepayment component that they're getting financed is less than them just buying it out on demand at that time, if they're able to get into that amount of capacity, then does the economics make sense there? And so here's some of the operating models that we've seen. **[00:11:20 · 0]** It's like, you can build and operate the full stack. Hardware, site, your team is going to arrange power, permit, the assembly, the management. You can also own servers in a colocation facility with more or less the assembly and operation handled by a partner. On the other end, an offtake contract buys defined capacity, and the provider is going to own and operate the asset. And They might not give you complete root access to the bare metal, or there might be levels to how much access you get, depending on if it's tenant, multi tenant, etcetera, type of configurations that can also affect, then, the pricing. **[00:11:53 · 0]** And so some businesses are going to value the physical access. Some businesses are going to value confidential compute. They might value certain topology changes or the ability to reconfigure the system. And so if that is something for your business, then they might actually want to have more and more control and build and operate the stack. And the economics for them might make sense, given the opportunity cost they would have to give up if they were to just take rental off take contracts. **[00:12:17 · 0]** So they might be willing to pay more there. And so here's how ownership and rental this is a very crude graph how ownership and rental can compare as utilization arises. And so what we're looking at is that, usually, an off take contract, or these off take contracts are getting to the cost of actually almost building out and financing your own cluster. And it's not... Because, I mean, if you're spending $100,000,000 over a three year period, then maybe if the economics make sense, it would have made more sense to simply buy out the compute operator yourself outside of the time cost, and then obviously the operational cost that might exist. **[00:12:56 · 0]** So there's a equilibrium that exists that has to be followed where the ownership costs outweigh the rental costs, and the rental costs should sit below the ownership costs. But if you're utilizing that at such a rate, and in such variable kind of increments, might be best to just own the stack outright. And so we're seeing this new dynamic that compute procurement affects how much runway a growing company actually has. And so this also gives investors an interest in compute procurement. So we're kind of seeing this. **[00:13:36 · 0]** Access and payment terms affect how much runway their port codes are going to have from a VC perspective. So helping a company buy a suitable capacity can improve the underlying investment. And they're very connected. Right? So they're able to procure, and they're able to actually find, compute capacity for their port goes and fund that. **[00:13:53 · 0]** And so that also brings another interesting aspect to the macro structure, into the micro structure is that who helps place the capacity, and how does this offer get compared, and what risks these contracts can transfer. So we have a lot of different players around this GPU hour that are helping shape this market here. You've got your price discovery individuals. So people with, indexes, data, the benchmarks that we're all looking at right now and referencing, which would be used to, you know, in the future, launch the Futures based on the benchmarks, you know, talked about in the associated contracts per spec per pending approval. And then there's the physical matching, Venues, Auctions, Aggregators. **[00:14:38 · 0]** So anyone who can match capacity. We're looking at essentially decentralized marketplaces for on demand compute. You're looking at liquid pools of compute that offer liquid pool services and part it out themselves and have their own standardized contract. You're looking at bare metal resellers. You're looking at NeoClouds themselves doing larger capacity sales. **[00:14:58 · 0]** And then you're looking at individual brokers and individual operators here as well on the bare metal reselling. You have Destin Brokers. They're looking at the bilateral quotes and execution. They might know a lot of people. They might be able to move fast. **[00:15:09 · 0]** They might be able to see quotes at very similar specs in one place and know that there's a buyer that is extremely inelastic to price right now, needs to compute immediately, and they can get the RFS in time, so they can charge a decent spread on it. And then what they do is they're going to have the source negotiating, and then they're going to be carrying and placing the risk. They're going to have to operate that cluster for the person, but it has to be worth that spread. And you have looks in financial markets, the exchanges that are coming that are based on the GPU hour, the futures and derivatives. People are going to be using it to take a look at how to maybe hedge out and look into more financial structures. **[00:15:43 · 0]** So hedge out risk, so hedge out amount of balance sheet risk, maybe get into more interesting agreements. We're seeing agreements such as prepayment payment facilities are coming online. We're seeing people utilize futures and are interested in utilizing futures and other kind of options to be able to... There's demand for this to be able to hedge out the risks that they currently have on the books. We have buy side individuals looking for flex contracts, flexible contracts. **[00:16:11 · 0]** Like, hey, is there ability that if I lock myself into a three to five year contract, that I would be able to at least upgrade when a new chipset comes out? Or maybe I'll be able to roll it over. Or maybe let me sell it back to you. And then there's credit and financing. So equipment lenders, leasers, specialist credit providers. **[00:16:27 · 0]** And so the way we're seeing it right now is that every deal is bespoke. There's different spec sheets. There's formats. The bill of materials may not match the actual delivery, or they might be subject to change if a part is missing or a part is out of stock. And there's very long lead times that some of these OEM and ODM providers have. **[00:16:48 · 0]** And And the same term can mean different things when you're negotiating very quickly. If someone is talking about memory, are we talking about VRAM? Are we talking about the CPU memory? Are we talking about additional storage here? Like, what did you exactly mean there? **[00:17:00 · 0]** It affects quite a bit what type of loads can be run, and the actual end user's needs, and also with pricing and delivery dates. So there's a lot of back and forth before anyone can really compare. And this is where the concept of non fungibility comes into place, and how to partition that out, what parts can be fungible, what parts hold the most of the weight, such as the SKU, etcetera, and what parts are still traded as if it was a product, e commerce, etcetera, where it's only for that particular cluster that this price applies. And so prices can update daily or faster. Quote's still a good starting point. **[00:17:33 · 0]** Bits can get shocked. Creditworthiness changes what's on offer. There's all these latent factors, hidden factors as well, why we're calling it a dark market. These are some public H100 references snapshotted today. I updated them to take a look at what the public... **[00:17:48 · 0]** Those are the references for if you wanted to go on their websites to see. This is just an example here, of, like, what a public on demand or spot pricing looks like for an h 100, today. ## Slides ### 00:00:20–00:00:40 # WE'LL BE RIGHT BACK ## AI Engineer NEW YORK View Full Schedule at [ai.engineer/nyc/2026/schedule](ai.engineer/nyc/2026/schedule) [Silhouette of the New York City skyline at night, including One World Trade Center, Empire State Building, and the Statue of Liberty] ### 00:00:40–00:01:00 # AI Engineer ## NEW YORK ### 00:01:00–00:01:40 # The Dark Pools of Compute Affaan Mustafa · Alejandro Soumah | Compute Trading at Jump Trading ### 00:01:40–00:02:00 # What do I mean by dark pool? - Equities: a regulated, structured venue - Compute: private quotes, brokers and long bilateral contracts - Most of the market happens out of view, in fragments ### 00:02:00–00:02:20 # What do I mean by dark pool? - Equities: a regulated, structured venue - Compute: private quotes, brokers and long bilateral contracts - Most of the market happens out of view ### 00:02:20–00:03:40 # What do I mean by dark pool? - Equities: a regulated, structured venue - Compute: private quotes, brokers and long bilateral contracts - Most of the market happens out of view, in fragments ### 00:03:40–00:04:00 ## Tokens get cheaper while compute demand grows - The cost of a given capability has fallen sharply. - The total bill still depends on the model, usage and the capacity being purchased. [Line graph titled "Costs of New Technology for End-Use Consumers, by investment cycle", showing the PCE Personal Computer Price Index (1980 ICT Investment Cycle), LLM Model Price Index (2022 AI Investment Cycle), and LLM Model Price-to-Intelligence Ratio (2022 AI Investment Cycle) over "Months Since Investment Cycle Start". All three indices start at 100 and decline over time, with the LLM indices showing a steeper initial drop.] Source: a16z State of Markets II (Sep 2026), slide 34; Goldman Sachs / Commerce (Jul 2026); Demirer et al., JEP (2026). ### 00:04:00–00:04:20 # Tokens get cheaper while compute demand grows ## 01 MACRO MARKET STRUCTURE & PLAYERS - The cost of a given capability has fallen sharply. - The total bill still depends on the model, usage and the capacity being purchased. a16z State of Markets II (Sep 2026), slide 34; Goldman Sachs / Commerce (Jul 2026); Demirer et al., JEP (2026). [Line graph titled "Costs of New Technology for End-Use Consumers, by investment cycle", showing three declining lines over "Months Since Investment Cycle Start": PCE Personal Computer Price Index, LLM Model Price Index, and LLM Model Price-to-Intelligence Ratio.] ### 00:04:20–00:04:40 # Tokens get cheaper while compute demand grows - The cost of a given capability has fallen **sharply**. - The total bill still depends on the model, usage and the capacity being purchased ### 00:04:40–00:05:00 ## 01 MACRO MARKET STRUCTURE & PLAYERS # Jevons and the decoupling of token and capacity prices Lower cost per task can expand usage, while capacity prices also depend on ### 00:05:00–00:05:20 ### MACRO MARKET STRUCTURE & PLAYERS ## Jevons and the decoupling of token and capacity prices Lower cost per task can expand usage, while capacity prices also depend on efficiency, workload mix and the supply of usable clusters. a16z State of Markets II, slide 34; Ornn, Silicon Data, Bloomberg, compiled by Citadel Securities; data through August 2026. [Line graph showing Omn H100 GPU Rental Price and Silicon Data Token Price Expenditure Index from Dec '25 to Sep '26. A red shaded area indicates "Token and compute prices decline together", and a green shaded area indicates "Token prices decline further but compute prices remain robust".] ### 00:05:20–00:05:40 ### 01 MACRO MARKET STRUCTURE & PLAYERS ## Jevons and the decoupling of token and capacity prices Lower cost per task can expand usage, while capacity prices also depend on efficiency, workload mix and the supply of usable clusters. a16z State of Markets II, slide 34; Omn, Silicon Data, Bloomberg, compiled by Citadel Securities; data through August 2026. [Line graph tracking Omn H100 GPU Rental Price and Silicon Data Token Price Expenditure Index from December 2025 to September 2026, highlighting periods where token and compute prices decline together, and where token prices decline further while compute prices remain robust.] ### 00:05:40–00:06:00 ### 01 MACRO MARKET STRUCTURE & PLAYERS ## Jevons and the decoupling of token and capacity prices Lower cost per task can expand usage, while capacity prices also depend on efficiency, workload mix and the supply of usable clusters. a16z State of Markets II, slide 34; Omn, Silicon Data, Bloomberg, compiled by Citadel Securities; data through August 2026. [Line graph showing Omn H100 GPU Rental Price and Silicon Data Token Price Expenditure Index from Dec '25 to Sep '26, with shaded regions indicating periods where "Token and compute prices decline together" and "Token prices decline further but compute prices remain robust"] ### 00:06:00–00:06:20 # MACRO MARKET STRUCTURE & PLAYERS ## Four ingredients - Chips - Power: the IEA projects global data-centre electricity use rising from 485 TWh in 2025 to 950 TWh in 2030 - Buildings and grid connections - Money Source: IEA, Key Questions on Energy and AI (April 2026); the projection covers all global data centres. ### 00:06:20–00:06:40 ## Four ingredients - Chips - Power: the IEA projects global data-centre electricity use rising from 485 TWh in 2025 to 950 TWh in 2030 - Buildings and grid connections - Money Source: IEA, Key Questions on Energy and AI (April 2026); the projection covers all global data centres. ### 00:06:40–00:07:00 ## Four ingredients - Chips - Power: the IEA projects global data-centre electricity use rising from 485 TWh in 2025 to 950 TWh ### 00:07:00–00:07:20 # How the compute market fits together ## MACRO MARKET STRUCTURE & PLAYERS **Chip and system vendors** - GPUs, memory, servers - and networking **Sites and operators** ### 00:07:20–00:07:40 ### 01 MACRO MARKET STRUCTURE & PLAYERS ## How the compute market fits together - **Chip and system vendors** - GPUs, memory, servers - and networking - **Sites and operators** - Power, space, - assembly - and management - **Capacity providers** - Hyperscalers, - neoclouds - and dedicated - clusters - **Compute users** - Labs, inference - clouds - and enterprises Lenders and investors support the hardware and site buildout. Brokers and marketplaces help connect specific capacity offers with buyers. Schematic: firms can span several layers. Public context: CoreWeave 2025 10-K; IEA (April 2026); SemiAnalysis AI Cloud TCO Model. [Flow diagram illustrating the progression from chip and system vendors to compute users, with intermediate stages of sites and operators, and capacity providers] ### 00:07:40–00:08:00 # How the compute market fits together - **Chip and system vendors**: GPUs, memory, servers and networking - **Sites and operators**: Power, space, assembly and management - **Capacity providers**: Hyperscalers, neoclouds and dedicated clusters - **Compute users**: Labs, inference clouds and enterprises Lenders and investors support the hardware and site buildout. Brokers and marketplaces help connect specific capacity offers with buyers. Schematic: firms can span several layers. Public context: CoreWeave 2025 10-K; IEA (April 2026); SemiAnalysis AI Cloud TCO Model. [Flow diagram showing four stages of the compute market: Chip and system vendors, Sites and operators, Capacity providers, and Compute users] ### 00:08:00–00:08:20 # How the compute market fits together ## 01 MACRO MARKET STRUCTURE & PLAYERS **Chip and system vendors** - GPUs, memory, servers - and networking **Sites and ### 00:08:20–00:08:40 ### 01 MACRO MARKET STRUCTURE & PLAYERS ## How the compute market fits together - **Chip and system vendors** - GPUs, memory, servers - and networking - **Sites and operators** - Power, space, assembly - and management - **Capacity providers** - Hyperscalers, neoclouds - and dedicated clusters - **Compute users** - Labs, inference clouds - and enterprises Lenders and investors support the hardware and site buildout. Brokers and marketplaces help connect specific capacity offers with buyers. Schematic: firms can span several layers. Public context: CoreWeave 2025 10-K; IEA (April 2026); SemiAnalysis AI Cloud TCO Model. [Flow diagram showing the compute market structure from vendors to users] ### 00:08:40–00:09:00 ## 01 MACRO MARKET STRUCTURE & PLAYERS # How the compute market fits together The compute market structure involves a progression of participants: - **Chip and system vendors**: Provide GPUs, ### 00:09:00–00:09:20 ### 01 MACRO MARKET STRUCTURE & PLAYERS ## The financing chain - Lenders finance the buildout against contracted revenue. - The customer's term can end before the financing does. - Credit support and payment timing affect what gets financed. ### A three-year loan with two-year offtake Illustrative term mismatch, not an observed transaction. CoreWeave filings; Liquid Compute; Mittal / Mustafa, Handmade. [Timeline chart showing hardware financing and customer contract durations, with a highlighted "Re-let risk" period from 24 to 36 months] ### 00:09:20–00:09:40 ### 01 MACRO MARKET STRUCTURE & PLAYERS ## The financing chain - Lenders finance the buildout against contracted revenue. - The customer's term can end before the financing does. - Credit support and payment timing affect what gets financed. ### A three-year loan with two-year offtake Illustrative term mismatch, not an observed transaction. CoreWeave filings; Liquid Compute; Mittal / Mustafa, Handmade. [Timeline diagram showing "Hardware financing" for 36 months, and "Customer contract" for 24 months, with a "Re-let risk" period from 24 to 36 months.] ### 00:09:40–00:10:00 ## 01 MACRO MARKET STRUCTURE & PLAYERS # The financing chain - Lenders finance the buildout against contracted revenue. - The customer's term can end before the financing does. - Credit support and payment timing affect what gets financed. ### A three-year loan with two-year offtake Illustrative term mismatch, not an observed transaction. CoreWeave filings; Liquid Compute; Mittal / Mustafa, Handmade. [Timeline diagram showing "Hardware financing" for 36 months and "Customer contract" for 24 months, with a "Re-let risk" period from 24 to 36 months] ### 00:10:00–00:10:20 # The financing chain ## 01 MACRO MARKET STRUCTURE & PLAYERS - Lenders finance the buildout against contracted revenue. - The customer's term can end before the financing does. - Credit support and payment timing affect what gets financed. ### A three-year loan with two-year offtake [Timeline diagram illustrating "Hardware financing" over 36 months, "Customer contract" over 24 months, and a "Re-let risk" period from 24 to 36 months] Illustrative term mismatch, not an observed transaction. CoreWeave filings; Liquid Compute; Mittal / Mustafa, Handmade. ### 00:10:20–00:10:40 # The startup math (illustrative) - Raise $40M - Need a $100M contract over 3 years - 30% prepayment = $30M - $10M left to run the business ### An inference cloud commits capacity in blocks Illustrative, not a forecast. Resale requires buyers, matching specifications and contractual permission. [Line graph showing workload demand increasing over time, with committed capacity represented by a stepped line. The graph highlights unused capacity (blue area) that can serve short-term buyers, and demand above capacity (red area) which is bridged by on-demand services.] ### 00:10:40–00:11:00 # The startup math (illustrative) - Raise $40M - Need a $100M contract over 3 years - 30% prepayment = $30M - $10M left to run the business ### An inference cloud commits capacity in blocks Illustrative, not a forecast. Resale requires buyers, matching specifications and contractual permission. [Line graph showing "Workload demand" (red line) increasing over time, and "Committed capacity" (grey stepped line) increasing in blocks. Shaded areas indicate "Unused capacity" (blue) when committed capacity is above demand, and "On-demand" (pink) when demand exceeds committed capacity.] ### 00:11:00–00:11:20 ### 01 MACRO MARKET STRUCTURE & PLAYERS # The startup math (illustrative) - Raise $40M - Need a $100M contract over 3 years - 30% prepayment = $30M - $10M left to run the business An inference cloud commits capacity in blocks Illustrative, not a forecast. Resale requires buyers, matching specifications and contractual permission. [Line graph showing workload demand (red line) increasing over time, with committed capacity (grey stepped line) and shaded areas for unused capacity (blue) and on-demand bridging a shortfall (pink).] ### 00:11:20–00:11:40 # Rent versus own | Operating model | What you own | Who supplies the site and operations | | :------------------------------ | :----------------------- | :----------------------------------------------------------------------- | | Build and operate the full stack | Hardware and site | Your team arranges power, permits, assembly and management | | Own hardware in managed colo | Servers | Colo supplies power and space; a partner manages hardware | | Own hardware in self-managed colo | Servers or components | Colo supplies power and space; assembly is outsourced and your team operates | | Capacity offtake contract | Capacity rights | Provider supplies the hardware, site and operations | Schematic: responsibility, access and service boundaries vary by agreement. Source framework: SemiAnalysis AI Cloud TCO Model. ### 00:11:40–00:12:00 ## Rent versus own | Operating model | What you own | Who supplies the site and operations | | :-------------------------- | :-------------------- | :-------------------------------------------------------------------- | | **Build and operate the full stack** | Hardware and site | Your team arranges power, permits, assembly and management | | **Own hardware in managed colo** | Servers | Colo supplies power and space; a partner manages hardware | | **Own hardware in self-managed colo** | Servers or components | Colo supplies power and space; assembly is outsourced and your team operates | | **Capacity offtake contract** | Capacity rights | Provider supplies the hardware, site and operations | Schematic: responsibility, access and service boundaries vary by agreement. Source framework: SemiAnalysis AI Cloud TCO Model. ### 00:12:00–00:12:20 ### MACRO MARKET STRUCTURE & PLAYERS ## Rent versus own - **Operating model: Build and operate the full stack** - What you own: Hardware and site - Who supplies ### 00:12:20–00:12:40 ## 01 MACRO MARKET STRUCTURE & PLAYERS # Rent versus own | Operating model | What you own | Who supplies the site and operations | | :------------------------------ | :---------------------- | :---------------------------------------------------------------------- | | **Build and operate the full stack** | Hardware and site | Your team arranges power, permits, assembly and management | | **Own hardware in managed colo** | Servers | Colo supplies power and space; a partner manages hardware | | **Own hardware in self-managed colo** | Servers or components | Colo supplies power and space; assembly is outsourced and your team operates | | **Capacity offtake contract** | Capacity rights | Provider supplies the hardware, site and operations | Schematic: responsibility, access and service boundaries vary by agreement. Source framework: SemiAnalysis AI Cloud TCO Model. ### 00:12:40–00:13:00 ## How ownership and rental compare as utilization rises ### 01 MACRO MARKET STRUCTURE & PLAYERS With $25,000 of net fixed ownership costs, $1/hour of operating costs and $3/hour rental, the curves meet at 12,500 used hours. The assumptions determine where they cross. [Line graph comparing "Total cost per GPU over the horizon" for "Rental" (red line) and "Ownership" (grey line) against "Used GPU-hours". The lines intersect at 12,500 used hours.] Hypothetical, ### 00:13:00–00:13:20 # How ownership and rental compare as utilization rises ## 01 MACRO MARKET STRUCTURE & PLAYERS ### Total cost per GPU over the horizon - Rental - Ownership With $25,000 of net fixed ownership costs, $1/hour of operating costs and $3/hour rental, the curves meet at 12,500 used hours. The assumptions determine where they cross. Hypothetical, undiscounted same-workload example. It compares pay-as-used rental; a term contract can charge for idle allocated hours. [Line graph comparing total cost per GPU for rental versus ownership, showing the ownership cost starting higher but increasing slower than rental, with the lines crossing at 12,500 used hours] ### 00:13:20–00:13:40 # How ownership and rental compare as utilization rises ## Total cost per GPU over the horizon - **Rental** - **Ownership** With $25,000 of net fixed ownership costs, $1/hour of operating costs ### 00:13:40–00:14:20 # A new dynamic - Compute procurement affects how much runway a growing company has - Investors have an interest in the terms as well as the company's access to capacity ### 00:14:20–00:14:40 ## The players around the GPU-hour ### Price discovery - Indices and data - Benchmarks, transaction data and comparable price references ### Physical matching - Venues, auctions and aggregators - Matching capacity offers with suitable workloads ### Decentralized networks - Distributed supply and settlement - Coordinating distributed resources and access to available capacity ### Desks and brokers - Bilateral quotes and execution - Sourcing, negotiation and carrying or placing risk ### Financial markets - Price risk and reference contracts - Exchanges and derivatives - Separate from physical supply ### Credit and financing - Funding hardware and commitments - Equipment lenders and lessors - Specialist credit providers ### Insurance and operations - Protection, telemetry, metering, billing and collateral services Functional view adapted from a public September 2026 compute-market map. Roles can overlap; not exhaustive. [Diagram showing seven categories related to the GPU-hour, each with bullet points describing its function] ### 00:14:40–00:15:00 # The players around the GPU-hour ## Price discovery - Indices and data - Benchmarks, transaction data and comparable price references ## Physical matching - Venues, auctions and aggregators - Matching capacity offers with suitable workloads ## Decentralized networks - Distributed supply and settlement - Coordinating distributed resources and access to available capacity ## Desks and brokers - Bilateral quotes and execution - Sourcing, negotiation and carrying or placing risk ## Financial markets - Price risk and reference contracts - Exchanges and derivatives - Separate from physical supply ## Credit and financing - Funding hardware and commitments - Equipment lenders and lessors - Specialist credit providers ## Insurance and operations - Protection, telemetry, metering, billing and collateral services Functional view adapted from a public September 2026 compute-market map. Roles can overlap; not exhaustive. ### 00:15:00–00:15:20 ## The players around the GPU-hour ### Price discovery - Indices and data - Benchmarks, transaction data and comparable price references ### Physical matching - Venues, auctions and aggregators - Matching capacity offers with suitable workloads ### Decentralized networks - Distributed supply and settlement - Coordinating distributed resources and access to available capacity ### Desks and brokers - Bilateral quotes and execution - Sourcing, negotiation and carrying or placing risk ### Financial markets - Price risk and reference contracts - Exchanges and derivatives - Separate from physical supply ### Credit and financing - Funding hardware and commitments - Equipment lenders and lessors - Specialist credit providers ### Insurance and operations - Protection, telemetry, metering, billing and collateral services Functional view adapted from a public September 2026 compute-market map. Roles can overlap; not exhaustive. ### 00:15:20–00:15:40 # The players around the GPU-hour ## 02 MICROSTRUCTURE & PLAYERS ### Price discovery - Indices and data - Benchmarks, transaction data and comparable price references ### Physical matching - Venues, auctions and aggregators - Matching capacity offers with suitable workloads ### Decentralized networks - Distributed supply and settlement - Coordinating distributed resources and access to available capacity ### Desks and brokers - Bilateral quotes and execution - Sourcing, negotiation and carrying or placing risk ### Financial markets - Price risk and reference contracts - Exchanges and derivatives - Separate from physical supply ### Credit and financing - Funding hardware and commitments - Equipment lenders and lessors - Specialist credit providers ### Insurance and operations - Protection, telemetry, metering, billing and collateral services Functional view adapted from a public September 2026 compute-market map. Roles can overlap; not exhaustive. ### 00:15:40–00:16:00 ## The players around the GPU-hour ### 02 MICROSTRUCTURE & PLAYERS - **Price discovery** - Indices and data - Benchmarks, transaction data and comparable price references - **Physical matching** - Venues, auctions and aggregators - Matching capacity offers with suitable workloads - **Decentralized networks** - Distributed supply and settlement - Coordinating distributed resources and access to available capacity - **Desks and brokers** - Bilateral quotes and execution - Sourcing, negotiation and carrying or placing risk - **Financial markets** - Price risk and reference contracts - Exchanges and derivatives - Separate from physical supply - **Credit and financing** - Funding hardware and commitments - Equipment lenders and lessors - Specialist credit providers - **Insurance and operations** - Protection, telemetry, metering, billing and collateral services Functional view adapted from a public September 2026 compute-market map. Roles can overlap; not exhaustive. ### 00:16:00–00:16:20 ## The players around the GPU-hour ### Price discovery - Indices and data - Benchmarks, transaction data and comparable price references ### Physical matching - Venues, auctions and aggregators - Matching capacity offers with suitable workloads ### Decentralized networks - Distributed supply and settlement - Coordinating distributed resources and access to available capacity ### Desks and brokers - Bilateral quotes and execution - Sourcing, negotiation and carrying or placing risk ### Financial markets - Price risk and reference contracts - Exchanges and derivatives - Separate from physical supply ### Credit and financing - Funding hardware and commitments - Equipment lenders and lessors - Specialist credit providers ### Insurance and operations - Protection, telemetry, metering, billing and collateral services Functional view adapted from a public September 2026 compute-market map. Roles can overlap; not exhaustive. ### 00:16:20–00:16:40 ## MICROSTRUCTURE & PLAYERS # The players around the GPU-hour ### Price discovery - Indices and data - Benchmarks, transaction data and comparable price references ### Physical matching - Venues, auctions and aggregators - Matching capacity offers with suitable workloads ### Decentralized networks - Distributed supply and settlement - Coordinating distributed resources and access to available capacity ### Desks and brokers - Bilateral quotes and execution - Sourcing, negotiation and carrying or placing risk ### Financial markets - Price risk and reference contracts - Exchanges and derivatives - Separate from physical supply ### Credit and financing - Funding hardware and commitments - Equipment lenders and lessors - Specialist credit providers ### Insurance and operations - Protection, telemetry, metering, billing and collateral services Functional view adapted from a public September 2026 compute-market map. Roles can overlap; not exhaustive. ### 00:16:40–00:17:20 ## Every deal is bespoke - Different spec sheets and formats - Bills of materials that don't match - The same term means different things - Lots of back-and-forth before anyone can compare ### 00:17:20–00:17:40 ## 02 MICROSTRUCTURE & PLAYERS # Every deal is bespoke - Different spec sheets and formats - Bills of materials that don't match - The same term means different things - Lots of back-and-forth before anyone can compare ### 00:17:40–00:18:00 # Pricing is a negotiation - Prices can update daily or faster - A quote is a starting point - Bids get shopped to push the price - Creditworthiness changes the offer ### Public H100 references, 13 October 2026 **Silicon Data** $2.85 Standardized neocloud daily index **SemiAnalysis** $3.32 H100 spot-contract composite **Compute Desk** $2.96 US one-year reserved reference **Ornn** $3.17 H100 SXM on-demand transactions USD per GPU-hour; the definitions and methods differ. Sources: Silicon Data; SemiAnalysis; Compute Desk / General Index; Ornn. Published benchmarks are not quotes.