# Buying GPUs Should Work Like Buying Stocks — Carmen Li, Compute Exchange

Carmen Li, Compute Exchange; Silicon Data

[Talk page](https://ai.engineer/talks/laQ8wMysnkk-buying-gpus-should-work-like-buying-stocks) · [Video](https://www.youtube.com/watch?v=laQ8wMysnkk)

## Buying GPUs Should Work Like Buying Stocks

Carmen Li explains how Compute Exchange matches buyers with verified GPU capacity, then uses Silicon Data’s forward curves and asset valuations to show why procurement, financing and price risk belong in the same conversation.

![Compute Exchange slide describing a neutral compute marketplace, verified providers, GPU supply and quote turnaround.](https://ai.engineer/api/talk-essay-media/5201ee164535297cb23c8afe0b32f76bc7b33dc154f3cc331df4009db26bad44)

Compute Exchange’s marketplace thesis and reported supply metrics, shown at 3:19.

### Three hundred providers create a procurement problem

[0:12](https://www.youtube.com/watch?v=laQ8wMysnkk&t=12s)

If you need reserved compute within forty-eight hours, calling three hundred salespeople is an expensive way to discover what is available. Carmen Li, CEO of Compute Exchange and Silicon Data, opens with that mismatch: a growing GPU ecosystem gives buyers more choices while making the transaction harder to organize. She describes a market with more than three hundred neocloud and private providers globally, alongside token factories and API routers.

The stock-market analogy is about reducing that coordination cost. A buyer can call brokers and banks individually, or use an exchange to find liquidity and complete a transaction through an established process. An exchange does not necessarily produce the lowest price on every trade. Its value is making the search, agreement and settlement more efficient. Compute Exchange applies that idea to finding verified compute resources.

Neutrality determines the platform’s business design: Compute Exchange owns no GPUs. Li’s concern is that owning capacity would give the marketplace a reason to favor its own inventory, cherry-pick customers or keep valuable provider relationships to itself. Without that inventory position, the intended job is to find the resources that suit each transaction. The focus is reserved compute, forward contracts and token packages.

### Verify what exists, what will exist and who will pay

[3:17](https://www.youtube.com/watch?v=laQ8wMysnkk&t=197s)

At the time of the recording, Li reports more than one hundred verified providers and more than eighty thousand GPUs in supply. Li says coverage excludes China. Her slide reports six regions across the US, EU and APAC and a twenty-four-hour quote turnaround. Verification changes with the thing being sold. An operating cluster can be inspected and benchmarked; a future cluster requires evidence that it can become available on schedule.

The checks address three different uncertainties:

- **Existing capacity:** Verify GPU identifiers, inventory and performance, including benchmarking such as MLPerf. The question is whether the functioning resources match what the buyer was promised.
- **Forward capacity:** Examine purchase orders and permits. For a buyer seeking GPUs from January 1, 2027, these documents support due diligence on capacity that cannot yet be tested as a working deployment.
- **The buyer:** Perform know-your-business and know-your-customer checks. Providers also need to establish who is requesting compute and assess whether that customer can pay at the agreed rate.

That two-sided verification supports a market aimed at actual delivery. Li calls the forward business a precursor to physically delivered futures. Her reference point comes from four or five years as an energy options market maker at DRW Trading in Chicago. In her analogy, WTI represents procuring the physical resource, while cash-settled Brent represents managing financial price exposure. “We’re creating the WTI” describes Compute Exchange’s intended role: arrange access to compute that a customer can use.

### Reserve a cluster, fix a token deployment or buy the servers

[5:30](https://www.youtube.com/watch?v=laQ8wMysnkk&t=330s)

A GPU forward contract schedules access to capacity starting in the future. A token forward adds a deployment layer to that reservation: reserve a GPU cluster, then run a dedicated model deployment on top. The resulting agreement can fix the model’s quantization and precision for a period and provide specified service-level agreements. The intended benefit is more predictable service and pricing than a buyer gets from volatile on-demand token APIs.

![Three columns compare reserved GPU rental, token and forward contracts, and an open hardware market.](https://ai.engineer/api/talk-essay-media/233490df616b02cc92f45a03b3ed2c55fda13df0481ed91794e0a9703380c7b9)

Three ways to buy compute: reserved GPU rental, token and forward contracts, and the open hardware market (5:32).

The distinction matters because a token contract commits to more than a hardware model. Two deployments on similar GPUs can differ in the model served and its numerical configuration. Fixing those choices makes the promised service more specific: the buyer is reserving a defined inference deployment rather than assuming that any available token endpoint will behave the same way.

The third market grew out of a much more tangible request: someone had twenty thousand H100s sitting in a data center and wanted help moving them. Li’s response was practical—“I can’t store in my home, obviously”—and Compute Exchange developed a marketplace for secondary server sales. She describes roughly twelve months of active trading, including B200 servers and transactions sometimes above the manufacturer’s suggested retail price. Her analogy is used cars during COVID: an already available asset can command more than its nominal new price.

Reserved GPU contracts start at one node for one month and can extend to three, four or five years, depending on provider availability. Compute Exchange does not offer on-demand GPU procurement. Li also describes selling back capacity when a customer has overcommitted on quantity or duration, though the recording leaves the rollout timing and commercial terms unclear. That flexibility should therefore be understood as a described option, without assuming a guaranteed exit price or an unconditional buyback.

### Route one request to the providers that can fill it

[8:01](https://www.youtube.com/watch?v=laQ8wMysnkk&t=481s)

Buyers can browse listings and discuss delivery directly with providers. When the listed inventory does not fit, a request for quote, or RFQ, expresses the requirement. Li describes requests ranging from one node to twenty thousand nodes. Behind the form, inventory tracking records which of the hundred-plus providers has which resources and roughly when future capacity should become available.

The concrete example is a request for 3,300 B200 GPUs in US East, starting January 1. Compute Exchange identifies three providers it believes can meet the requirement and routes the request to all three. Each responds through the platform with pricing, specifications and service-level agreements. Existing servers are benchmarked. The buyer then compares the offers; the platform does not reduce the decision to the lowest price.

The records remain in one place, and the platform introduces the buyer and provider to each other. That follows the earlier neutrality decision: Compute Exchange says it does not keep those relationships hidden. Its used and refurbished inventory spans 4090s through B200s and AMD hardware, with communication through Slack, email or chat. It also tracks transactions end to end and scores both customers and providers, although those scores were not yet visible to users. Li describes the service as a negotiated exchange, with a middleman helping clarify terms and share practices from other transactions.

One requirement, three independent offers
Inventory information guides routing. Providers return separate offers, and the buyer chooses after comparing the terms.
- Buyer’s RFQ — 3,300 B200 GPUs; US East; January 1 start.
- Inventory tracking — Which providers have the required hardware and roughly when.
- Compute Exchange routing — Identify three providers believed able to meet the requirement.
- Matched provider A — Returns its pricing, specifications and SLAs.
- Matched provider B — Returns its pricing, specifications and SLAs.
- Matched provider C — Returns its pricing, specifications and SLAs.
- Buyer compares and chooses — Offers appear together; existing servers are benchmarked. Price is one factor.
- Buyer’s RFQ → Compute Exchange routing (Supplies requirements)
- Inventory tracking → Compute Exchange routing (Identifies plausible suppliers)
- Compute Exchange routing → Matched provider A (Sends RFQ)
- Compute Exchange routing → Matched provider B (Sends RFQ)
- Compute Exchange routing → Matched provider C (Sends RFQ)
- Matched provider A → Buyer compares and chooses (Returns offer through platform)
- Matched provider B → Buyer compares and chooses (Returns offer through platform)
- Matched provider C → Buyer compares and chooses (Returns offer through platform)

### The demo turns a GPU count into a deployment specification

[11:07](https://www.youtube.com/watch?v=laQ8wMysnkk&t=667s)

The [platform demonstration at 11:22](https://www.youtube.com/watch?v=laQ8wMysnkk&t=682s) begins with available inventory. Li describes signup and browsing as free, with no requirement to purchase. The RFQ form provides a way to ask for capacity that the buyer does not find among the listings.

What must a provider know beyond “I need GPUs”? The form collects vendor type, GPU model, deployment type and GPU count, then interconnect preferences, CPU platform and storage. Optional requirements include system and software stacks, regions, head nodes and the intended use case. These fields turn a quantity into a deployment specification: enough detail to distinguish offers that might otherwise look comparable on GPU count alone.

![Lower portion of a Compute Exchange request-for-quote form with interconnect, storage, software, region, head-node and use-case fields.](https://ai.engineer/api/talk-essay-media/00d9198c3af79f300371df0e686073f68cd04fea694a3117034e93de00eb92c4)

The RFQ’s additional requirements include interconnect, storage, software, region, head nodes and use case (12:06).

Li encourages buyers to select multiple regions if the workload permits it. Region is a matching constraint in the same way that hardware and deployment type are constraints. Allowing more than one location gives the platform more possible inventory to match, while keeping a strict location requirement narrows the set. Once those preferences and requirements are entered, the buyer submits the RFQ for routing.

### Longer commitments stopped buying the same discount

[12:29](https://www.youtube.com/watch?v=laQ8wMysnkk&t=749s)

The final part moves from procuring compute to pricing it through time. Silicon Data supplies compute indices; at the time of the recording, Li says CME futures using its H100 neocloud on-demand and B200 price indices were expected to begin trading within a few weeks. That was a forthcoming market. The curves she then discusses are drawn from reserved-contract term structures, with forward rates inferred from those observations.

How did the reward for signing a longer H100 contract change? In the [term-structure discussion at 13:00](https://www.youtube.com/watch?v=laQ8wMysnkk&t=780s), each colored curve shows reserved rates observed on a different date. Moving right along a curve means committing for more months. The vertical axis shows the rate in dollars. On the earlier yellow curve, longer commitments generally bought a lower rate. Li describes a shift into contango early the following year, followed by higher, flatter pricing. The later curves sit above the yellow one, though their slopes vary across the term.

![Silicon Data dashboard with B200 selected and several term curves; a hover tooltip covers part of the chart.](https://ai.engineer/api/talk-essay-media/86ba3bfd7fac23a7bd7a75459fee9ec4582ca37c27947ff37018250bc440ab5f)

B200 reserved-contract rates by contract length in months (14:45). The colors compare observation dates. The tooltip samples one duration and covers part of the curves.

An observed multi-year contract price and an implied forward rate answer different questions. The contract price describes the quoted commitment over its duration. The implied forward curve uses the term structure to infer rates for future periods within that horizon. Li describes a methodology similar to Treasury-curve calculations and interprets the result as a market expectation of rising prices over the next thirty-six months. These are calculated rates rather than traded GPU futures prices, and a market expectation can be wrong.

Li says B200’s earlier curve was already in contango and that rates later rose at every term as the curve flattened. The B200 dashboard shows how the comparison depends on duration: several curves dip through the middle of the range before turning up at longer terms. Li separately reports that the implied forward rates pointed toward higher future prices.

### A three-year reservation leaves a six-year financing problem

[15:00](https://www.youtube.com/watch?v=laQ8wMysnkk&t=900s)

Term structure matters because a provider’s financing and customer contracts can end at different times. Li’s example has servers on the balance sheet with financing obligations extending five or six years, while reserved customer contracts cover only the first three. Years four through six still depend on the rates the provider will be able to earn after the current reservations expire.

That creates two distinct exposures:

- **Uncontracted later years:** Debt obligations continue after the reserved revenue ends. Li describes swaps and block trades as ways market participants seek to hedge the resulting future price exposure.
- **Default during contracted years:** The first three years also carry risk if the reserved customer contract defaults. Li reports trading around indices used as floating rates, raising the question of how dependable the promised contract revenue is.

The earlier physical-procurement analogy now has a financial counterpart. A reservation arranges a period of access or contracted revenue; a hedge addresses exposure to changing prices. Neither the existence of a reservation nor a favorable-looking forward curve removes the mismatch between a server’s financing horizon and its contracted earning period.

### What earning power says a used GPU is worth

[15:55](https://www.youtube.com/watch?v=laQ8wMysnkk&t=955s)

The same term structure supports another calculation: valuing the servers themselves. Li says Silicon Data provides appraisals for several banks holding GPU servers on their balance sheets and that NVIDIA cited its data in a September investor presentation. The valuation method she explains is forward-looking discounted cash flow: estimate the cash an asset can earn, then translate those future receipts into a present value.

The example is an A100 server with a new three-year lease. The lease rate supplies an earning assumption; utilization and costs affect how much cash the server produces; discounting brings that cash back to time zero. The resulting value depends on those assumptions. This is an estimate of earning power, so a change in rental rates can change the valuation even as the hardware gets older.

How does that estimate compare with depreciating hardware over five years? Silicon Data compares normalized residual-value estimates for A100, H100 and B200 servers with a five-year straight-line depreciation curve. Li describes three different outcomes:

- **A100:** Estimated value declines, but remains above the accounting depreciation curve. Getting older and retaining more value than the accounting schedule predicts can both be true.
- **H100:** The described value has risen. Li calls the pricing “going bananas” and returns to the analogy of used cars during COVID.
- **B200:** The described value has moved above MSRP, consistent with the earlier discussion of secondary transactions above the nominal new-hardware price.

The ending makes the procurement market consequential for anyone financing or holding the machines. A five-year depreciation schedule reduces book value with time; a cash-flow valuation responds to what compute can earn. In the examples Li presents, those paths diverge: A100 value declines but remains above the accounting depreciation curve, while H100 and B200 values rise. The rates negotiated for access to compute help explain why the servers supplying it can retain—or gain—economic value.

## Key takeaways

- Compute Exchange’s procurement model combines inventory matching with different checks for operating clusters, future capacity and buyers. It owns no GPUs to avoid competing with its providers.
- A token forward reserves a dedicated deployment on a GPU cluster, fixing model configuration and service commitments for a period to make service and pricing more predictable.
- The RFQ process routes a deployment requirement to plausible providers and returns pricing, specifications and SLAs for the buyer to compare. GPU count and price alone do not determine the choice.
- Reserved-contract curves had moved higher and flattened. Silicon Data’s implied forward rates suggested rising future prices, but those calculations were neither traded futures prices nor a guarantee of future outcomes.
- Financing can outlast reserved revenue, leaving providers exposed in later years. Discounted cash-flow valuations also depend on future earning power, which can diverge substantially from straight-line depreciation.

## Related talks

- [Covalent Launch: The GPU Cheatcode: Fine-tune 20 Llama Models in 5 Minutes](https://ai.engineer/talks/zHYQZFy0UVk-covalent-launch-gpu-cheatcode-fine-tune-20) — Follows the operational side of acquired compute: assigning remote GPU and CPU resources to fine-tuning, evaluation and inference workflows.
- [Homa: The End of TCP for AI Clusters — John Ousterhout, Stanford](https://ai.engineer/talks/eZ8WWZzoaR0-homa-end-tcp-ai-clusters-john-ousterhout) — Explains why cluster interconnect and communication latency affect useful GPU work, adding technical context to the deployment requirements collected in an RFQ.

## Slides

### Compute Exchange

[0:12](https://www.youtube.com/watch?v=laQ8wMysnkk&t=12s) [The title introduces a marketplace for AI infrastructure and three ways to obtain GPUs.](https://assets.ai.engineer/sites/aiecode2025/sha256/78a8a20d34863964094dbbd90e29919e0959ba21fe09781f97148034993543b0.webp)

Compute Exchange

### A marketplace for compute

[3:15](https://www.youtube.com/watch?v=laQ8wMysnkk&t=195s) [A headline sets out the commodity-market thesis above four figures about the exchange.](https://assets.ai.engineer/sites/aiecode2025/sha256/443c2a91ea87641a3b29ab4fa73c77170637d18530edbc8218d77e67c183fe02.webp)

Compute will become a commodity. We are building its marketplace.

### Three ways to buy compute

[5:28](https://www.youtube.com/watch?v=laQ8wMysnkk&t=328s) [Three columns distinguish renting reserved GPUs, contracting for future capacity and buying hardware.](https://assets.ai.engineer/sites/aiecode2025/sha256/1096a3946e814c767a21af4682cfc39a745a4bcaa4ecb7c296acfc87103dfd1f.webp)

Reserved GPU rental · Token & forward contracts · Open hardware market

### Reserved GPUs without a rate card

[7:20](https://www.youtube.com/watch?v=laQ8wMysnkk&t=440s) [The slide describes a competitive quoting process for a chosen rental term.](https://assets.ai.engineer/sites/aiecode2025/sha256/7de63c69b8fbf9fda27dbcf13211c4f19ce0e058eea93cc4ad7f8a7090e17be6.webp)

Reserved GPUs, without the rate card.

### From requirements to a request for quotes

[8:11](https://www.youtube.com/watch?v=laQ8wMysnkk&t=491s) [Two screens show how a buyer specifies a GPU model, region, deployment type and term.](https://assets.ai.engineer/sites/aiecode2025/sha256/a7ec610205d7b45e0846eef2923827508895576d4dee7d8844709b89579e32bc.webp)

Requirement to RFQ in two minutes.

### One view of competing quotes

[9:31](https://www.youtube.com/watch?v=laQ8wMysnkk&t=571s) [A comparison table sits next to a provider-quotation screen with individual offers.](https://assets.ai.engineer/sites/aiecode2025/sha256/0d1a98fe4719f598116fcbf10f019d2fd4077f3c2deb7ba2908098e9e4dfa878.webp)

Every quote in one place.

### Buying future capacity

[9:44](https://www.youtube.com/watch?v=laQ8wMysnkk&t=584s) [A forward-contract slide explains a buyer’s ability to reserve tomorrow’s capacity at an agreed price.](https://assets.ai.engineer/sites/aiecode2025/sha256/29111ab67ff2633737f285a8d21a0322a53d060ba6e4b83b4be35aae0ebc45ec.webp)

Buy tomorrow’s compute at today’s price.

### The hardware market

[9:47](https://www.youtube.com/watch?v=laQ8wMysnkk&t=587s) [A chip drawing accompanies options for pre-owned, refurbished and new hardware.](https://assets.ai.engineer/sites/aiecode2025/sha256/54fe53dc638dc68ac76a58fffa364a71fffe16b932a63175706432da16f237d9.webp)

Own the silicon.

### Tracking an RFQ end to end

[10:17](https://www.youtube.com/watch?v=laQ8wMysnkk&t=617s) [An exchange dashboard shows requests for quotes and their progress through routing and review.](https://assets.ai.engineer/sites/aiecode2025/sha256/74f83b5dd11cc10c3300ebcd85ef0bcc70b8c3685ec7ba72cd50fa823e6051b6.webp)

Every RFQ, tracked end to end.

### How the exchange verifies participants

[10:46](https://www.youtube.com/watch?v=laQ8wMysnkk&t=646s) [The slide connects participation to corporate registration, ownership and operating history.](https://assets.ai.engineer/sites/aiecode2025/sha256/72d139d1b89aff1c3c26c6b8ac20be94f20a1378ff473dc136014a129181756d.webp)

A verified network, not a listing site.

### Negotiating inside the exchange

[10:50](https://www.youtube.com/watch?v=laQ8wMysnkk&t=650s) [An interface screenshot shows the exchange’s in-platform negotiation workflow.](https://assets.ai.engineer/sites/aiecode2025/sha256/52f9eb2ba236a2020887ae3206abaa92538949e4a676e89351fed77b96e8fe5f.webp)

Negotiation stays on the exchange.

### A provider’s view of the market

[11:03](https://www.youtube.com/watch?v=laQ8wMysnkk&t=663s) [Four stages take a provider from listing capacity through buyer review, terms and tracking.](https://assets.ai.engineer/sites/aiecode2025/sha256/489f1cd07281fe88962e49cb61f93fccc3953c8519f800bbeeb95657fb4e86fd.webp)

Providers list capacity. Buyers come to them.

### Term-implied forward curves

[13:21](https://www.youtube.com/watch?v=laQ8wMysnkk&t=801s) [A market-data screen plots several colored forward-price curves on the same axes.](https://assets.ai.engineer/sites/aiecode2025/sha256/78cf31d0729b6b25125374bc981a5e44152183d0c198b5c81ccd524516d4360b.webp)

Term-Implied Forward Curve

### Accelerated depreciation versus retained value

[16:09](https://www.youtube.com/watch?v=laQ8wMysnkk&t=969s) [A chart compares the per-GPU value of an HGX system with an accelerated depreciation schedule.](https://assets.ai.engineer/sites/aiecode2025/sha256/3b826d7160334005c3711f888f0c79a1febfb95e1c92b2eef7157e688781e672.webp)

NVIDIA AI Infrastructure Retains Value Beyond Accelerated Depreciation Schedules

## Full transcript

[0:12](https://www.youtube.com/watch?v=laQ8wMysnkk&t=12s) Happy Wednesday. My name is Carmen Li. I'm the CEO for Compute Exchange, as well as Silicon Data. Um, I love the, you know, the previous session, um, we talk, they talk about a lot about the whole ecosystem, the financiers and the spot market traders and the hedgers. So what we're sitting in is really cool. I have a company on Compute Exchange. The world's actually the only neutral marketplaces for GPU procurement. We started in twenty twenty-four when there

[0:42](https://www.youtube.com/watch?v=laQ8wMysnkk&t=42s) were, I don't know, five neoclouds, three hyperscalers. Um, essentially, we sort of s- will foresee, interestingly enough, the proliferation of the GPU ecosystem. So you think about stock market today, right? We try to buy and sell stocks. You can call brokers, obviously. You can call a few banks to say, "Hey, I want to buy and sell X amount of Tesla or actually SpaceX." Right? And you can do that, or you can transact through your app at

[1:12](https://www.youtube.com/watch?v=laQ8wMysnkk&t=72s) NYSE or NASDAQ. Doesn't mean you get the best price at NYSE or NASDAQ, just because it's more efficient. And you know the liquidity will be there. You know the trade will be settled, it will be cleared. You can get what you're looking for at the prices you're looking for in a very efficient manner, right? For capital markets and resource allocation, it's about efficiency. So in compute world, obviously now we have a huge proliferation, which is great. We have over three hundred neocloud private provider

[1:42](https://www.youtube.com/watch?v=laQ8wMysnkk&t=102s) globally. We have a lot of token factories. We have a lot of API routers, which is all good for end customers. But at the same time, introduce a very interesting variable is, how do we make transaction efficient, right? Do you want to call three hundred salespeople? I mean, it's great to make friends, right? You want to talk to them regular basis. They know a lot of things. But same time, when you need compute in the next forty-eight hours on reserve contracts, it's the easiest way is go to

[2:12](https://www.youtube.com/watch?v=laQ8wMysnkk&t=132s) an exchange and then go to the verified compute resources. Just like go to NASDAQ or NYSE to buy stocks, right? So that's one of the things Compute Exchange does extremely well, is we have this neutral platform. And by design, we do not own GPUs. Once we start owning GPUs, that would translate to just like NASDAQ owning stocks, translate to I have a co- conflict of interest with you and with the providers. I could cherry-pick clients, right? I could hoard

[2:42](https://www.youtube.com/watch?v=laQ8wMysnkk&t=162s) relationships. Because Compute Exchange does not own anything by design, we work really well with our providers globally. Um, and one very important thing is we treat every single transaction individually, meaning we want to find you, as an end customer, the most suited compute resources. And we focus on reserve compute, forward contracts, as well as token packages. So give you some, um, high

[3:12](https://www.youtube.com/watch?v=laQ8wMysnkk&t=192s) level.

[3:17](https://www.youtube.com/watch?v=laQ8wMysnkk&t=197s) So today, we have over a hundred verified providers, um, and over eighty thousand GPU in supply. We cover all regions, obviously ex-China. Um, our commitment is twenty-four hours turn around. The difference between us and other so-called brokers is we do verify GPUs. If it's existing reserve capacity, meaning things are actually in a functional stage, we'll verify the UIDs, the

[3:47](https://www.youtube.com/watch?v=laQ8wMysnkk&t=227s) inventories, the performances. We run NVIDIA DGX, InferenceX, MLPerf, make sure you get resources you're promised to get. Um, so we're the middle, middle sort of trusted platform. So if it's forward capacity, let's say you need GPU starting Jan first, twenty twenty-seven, we actually look through the purchase orders, all the permits they get, make sure, you know, we help you do a due diligence level. Now, on the flip side, right? If you're a neocloud,

[4:17](https://www.youtube.com/watch?v=laQ8wMysnkk&t=257s) your concern is there's a zillion clients looking for compute. How do you know who has the ability to pay at a scheduled rate? That's where Compute Exchange comes in. We do KYB, KYC of the clients. Obviously, you want to make sure your clients is the clients you think they, who they are. So we step in and do that kind of verification, just like any exchange would do. And everything we do in the forward markets is actually the precursor to physically deliver futures. Also, give you some context who I

[4:47](https://www.youtube.com/watch?v=laQ8wMysnkk&t=287s) am, what I do. Sorry, I probably should say that. I started my career as energy options market maker, um, at DRW Trading in Chicago. So I did that for four or five years. So as option, energy market maker, everything's very intuitive. You have WTI, you have Brent, right? WTI is a physically delivered commodity contracts. Brent is cash-settled. They also have very distinct use cases. If you're looking for hedging in the financial sense, right, the, your price volatility,

[5:17](https://www.youtube.com/watch?v=laQ8wMysnkk&t=317s) your cost volatility, you're going to Brent. You're looking physically procure resources, use WTI. So we're creating the WTI at Compute Exchange.

[5:30](https://www.youtube.com/watch?v=laQ8wMysnkk&t=330s) So the ways to buy compute through, uh, Compute Exchange right now, you can get reserve contracts, you can get forward contracts on both GPU and token. Let me explain. The forward contracts on GPU is very reasonable, right? You want something starting in the future, nothing available right now. So that's something we're doing on regular basis. Token forward, I think, can be a new-ish concept. Basically, you can think about as, um, a dedicated deployment, on-premise deployment, right? Similarly, you reserve a bare-metal

[6:00](https://www.youtube.com/watch?v=laQ8wMysnkk&t=360s) GPU clusters. You reserve a cluster, but you have dedicated token deployment on top of that. So you get guaranteed, um, SLAs for that particular models. You get, um, quantization, precision, everything's fixed for you for a period of time. So we try to solve the on-demand token API volatility pricing problem here. The third is we have a huge market, open hardware market. It's really interesting. The

[6:30](https://www.youtube.com/watch?v=laQ8wMysnkk&t=390s) secondary server transactions started last year when people just pinging me say, "Hey, Carmen, I have like 20,000 H100 sitting X data center. Can you take away from me?" So I was like, "I don't know. I can't store in my home, obviously." So we started having the marketplace for secondary server sales. This has been a very vibrant market, um, for the past, I would say, 12 months. Uh, we cover, again, all servers, um, even B200. You see secondary transaction is happening,

[7:00](https://www.youtube.com/watch?v=laQ8wMysnkk&t=420s) um, sometimes trading above MSRP, right? Can make sense. Think about like COVID times when secondary car is trading above, right, its, uh, brand new car price. So we cover the three major, um, layers of compute.

[7:17](https://www.youtube.com/watch?v=laQ8wMysnkk&t=437s) So reserve GPU, obviously, I think the previous speaker covered extremely well with the terms. The minimum we do is one node starting at one month, and then we can go all the way to three, four, five years, right? It's all depends on the availability, the providers, but you get a chance to state your preferences. Um, we don't do on-demand.

[7:40](https://www.youtube.com/watch?v=laQ8wMysnkk&t=460s) And another thing really works really well for Compute Exchange is we do provide a flexibility. Let's say if you overcommit i-in terms of quantity or duration, you can sell it back to Compute Exchange, right? So that's something we're coming to do, so you don't have to worry too much if you overbought.

[8:01](https://www.youtube.com/watch?v=laQ8wMysnkk&t=481s) All right. Second-hand live. So I will actually do a quick demo later. Um, so we have a lot of listings, marketplace listings. You can browse and chat with the providers and figure out the best delivery mechanisms. You can also RFQ if you don't see anything in like-- you liked. We actually seeing over a billion RFQ-ed every month. Sounds big, but we all know it's full compute, it's not that huge, right? But still you see the vola-- the, the, the, the, uh, activities, which I love, is

[8:31](https://www.youtube.com/watch?v=laQ8wMysnkk&t=511s) we see a big range of RFQ requests from one node all the way to 20,000 nodes. So our clients are, you know, all different shapes and forms, different use cases, which is great.

[8:45](https://www.youtube.com/watch?v=laQ8wMysnkk&t=525s) Yeah. So you set timing, submit, and we route it. So in the behind the scene, we have inventory tracking. So with over 100 providers globally, we know who has what, roughly when, right, the forward capacity. And then let's say you, you submit a request saying that, "Hey, I need 3,300 GPUs, B200, US East, starting Jan first." We say, you know, we have three providers we think they have those. We'll route it to the three. They will respond to you through the platforms, and they'll give you, um, a rubrics

[9:15](https://www.youtube.com/watch?v=laQ8wMysnkk&t=555s) back, pricing, specs, SLAs. We benchmark them if it's existing servers, and then it's your choice, right? Price is not only indicator. We believe give you as much transparency as we can, and then you will make a decision.

[9:32](https://www.youtube.com/watch?v=laQ8wMysnkk&t=572s) So all records in one place. Again, it's rubrics coming back. You will see everything, and we introduce you to each other. We never hold relationships, right? We'll make sure you know who you're dealing with. Um, so buy tomorrow's compute today's price. So it's a physically procured platform.

[9:49](https://www.youtube.com/watch?v=laQ8wMysnkk&t=589s) This is the used, refurbished, uh, GPUs. So again, we have extremely, um, exciting, uh, ecosystem participants in this ecos-- uh, in this, uh, secondary GPU sales. We cover from 4090s to B200, AMD, everything.

[10:10](https://www.youtube.com/watch?v=laQ8wMysnkk&t=610s) Inventory is live, and we do connect with you through all different channels, Slack, email, chat, however you wanna do this. Um, we track end-to-end. We actually have scoring for everybody. We score you, our customers, we score our providers. You can't see it quite yet, but you know, will be really fun. Um, we're gonna have end-of-year celebration for our providers. We can r- we're gonna give out the best provider in APAC, best provider in Europe, best provider in US,

[10:40](https://www.youtube.com/watch?v=laQ8wMysnkk&t=640s) most responsive providers. So we're gonna have a lot of provider awards coming up. Stay tuned.

[10:48](https://www.youtube.com/watch?v=laQ8wMysnkk&t=648s) Um, verified network, so everything's, uh, you know, verified, which is, you know, negotiated exchange. So we help in terms of be the middleman. Sometimes things are not very clear. We try to make sure people understand what's the best practices we have seen in the ecosystem.

[11:07](https://www.youtube.com/watch?v=laQ8wMysnkk&t=667s) I'm gonna quickly show you guys the platform. I think that's more fun.

[11:19](https://www.youtube.com/watch?v=laQ8wMysnkk&t=679s) Um, here.

[11:22](https://www.youtube.com/watch?v=laQ8wMysnkk&t=682s) So this is the platform. You can see bunch of available inventories. By the way, anyone can, can sign up. It's completely free. There's no downside using us. You don't have to buy, you can browse. You can request for quote if you, you know, interested something we don't have. You can start with your vendor type. Again, we're agnostic of GPUs. You can select your model. You can select your deployment type. You can select number of GPUs. You can select your compute interconnect preferences, CPU platforms,

[11:52](https://www.youtube.com/watch?v=laQ8wMysnkk&t=712s) storages, everything you need, and then some optional items with system, software stacks, regions. Obviously, we encourage you to select multiple regions if you're open to that. You can add head nodes, tell us your use cases, anything you want, and then submit RFQ. So this has been extreme robust, and then we, you know, we obviously we want to make sure, um, if you have any feedback, let us know. So I'm going to pause on Compute Exchange side. Um, I think I have some time left,

[12:22](https://www.youtube.com/watch?v=laQ8wMysnkk&t=742s) which I'd love to show you some of the numbers we've seen in terms of term structure.

[12:29](https://www.youtube.com/watch?v=laQ8wMysnkk&t=749s) Also, context. I'm CEO for Compute Exchange. I'm also CEO for another company called Silicon Data. So I'm a founder CEO for Silicon Data, which is the Bloomberg for compute. So we provide indices for CME to settle their futures on, which will be happening in a few weeks. W- I'm very excited about this. So the ones going to be start trading at CME will be the H100 neocloud on-demand rates, as well as B200, um, prices.

[13:00](https://www.youtube.com/watch?v=laQ8wMysnkk&t=780s) What I want to show you is implied forward curve. So as I said, Compute Exchange, we do a lot of forward capacities, so essentially a lot of longer-term contracts is going to be very critical. You think about e- even in your cloud or your, your consumers. Let me just make it larger. So the yellow line was last year, right? The, the X, the horizontal line is the duration, and the dollar amount is the Y-axis. So last year was

[13:30](https://www.youtube.com/watch?v=laQ8wMysnkk&t=810s) complete downward sloping curve from the yellow line, right? So meaning the longer term contract you signed, the cheaper it, you know, was per GPU power. And this year, starting early this year, the whole curve shift up and became contango, if you see the, the, the red line, right? So it translate to people not getting a discount, enough discount for longer term contract, implies that people assume the price is going to rise down the line. And now you see the price, the ter- whole term keep shifting up but got

[14:00](https://www.youtube.com/watch?v=laQ8wMysnkk&t=840s) flattened, right? So meaning people will now either get penalized or get, you know, awarded by signing longer term contract. This is reserve contract we observe, and if you look at implied forward rates, this is not-- this is implied forward rates, right? Because we don't have a future trading, so this is calculated based on our term structure. Um, and we use similar methodology as Treasury futures curve. So you can see market consensus view is the price could go up next in, in the

[14:30](https://www.youtube.com/watch?v=laQ8wMysnkk&t=870s) future, next thirty-six months. Again, market consensus doesn't mean it's right. If you look at B200, again, last year was contango, and this year it's-- the whole curve shifted up. Every single term is more expensive, and it's got flattened, right? Implied forward rates, similarly, people think the price is going to go up. I want to point this out, the term structure out, is this shows you that GPU is behaves just like every commodity,

[15:00](https://www.youtube.com/watch?v=laQ8wMysnkk&t=900s) have a very clear term structure risks, which data futures are excellent at hedging. So we have seen a lot of block trading happening around the term structure. So you think about this, right? If you have servers on your balance sheet, you have financing obligation for next five, six years, and you have reserve contracts for first three, then you're exposed essentially year four or five or six, right? So you want to hedge that risk out by entering, you know, swaps or block trading.

[15:30](https://www.youtube.com/watch?v=laQ8wMysnkk&t=930s) Um, so those are things. Or, or there are, there are trades happening even for the first three years, which you have reserve contracts on. What if the reserve contract defaulted, right? What's your, um, guarantee from some of the players? So we saw block trading happening around the indices trade as a floating rate.

[15:55](https://www.youtube.com/watch?v=laQ8wMysnkk&t=955s) This chart, um, actually is being, uh, used differently by NVIDIA team. Um, NVIDIA's latest, uh, um, September investor day deck, they said us were the only data provider they actually cited in their deck. Um, this is very interesting. Silicon Data becomes the appraisal for a few banks when they have GPU servers on their balance sheet. Um, this, everything you see here are forward-looking view, right?

[16:25](https://www.youtube.com/watch?v=laQ8wMysnkk&t=985s) So discounted cash flow. So as I said, we have a term structure, so assume A100s, right? A100s have been around for six years. If today you sign a three-year long lease on A100s and you have a rate, and we discount the, the cash back to times zero, you have how much the asset's worth, right? Obviously, there's a lot of things baked in, utilization rate, what is your cost? There's a bunch of as-assumptions which we're pretty transparent about. Um,

[16:55](https://www.youtube.com/watch?v=laQ8wMysnkk&t=1015s) we actually calculate the residual value for normalized value, obviously, across all A100, H100, B200 servers globally. And then we compare that with the straight-line depreciation, which is everybody using the five-year depreciation. What's interesting is the A100, right? Even though it is downward sloping, uh, you know, it means the things are depreciating, but it's above the accounting depreciation curve, straight line. And the H100 obviously is, you know, it's going, going bananas now on the

[17:25](https://www.youtube.com/watch?v=laQ8wMysnkk&t=1045s) price actually went up. My as- my analogy will be just like secondary cars during COVID time. The price went up, right? And B200 obviously went above its MSRP by this point. So this is gave some, give you some views about what's going on in the residual value side. I think I'm out of time, but, you know, feel free to grab me, Carmen, cli@, uh, computeexchange.com, cli@silicondata.com. Feel free to grab me on Twitter, carmenli is my Twitter handle.

[17:55](https://www.youtube.com/watch?v=laQ8wMysnkk&t=1075s) Just Carmen Li. That's it. Thank you.
