Compute derivatives

Compute derivatives are financial contracts whose value depends on the price of AI computing capacity. In practice the price is almost always the hourly rental rate of an NVIDIA data-center GPU as measured by a third-party index, and the exchange-traded contracts settle in cash rather than by delivering GPU time.^[1][2] The category took shape in 2025 and 2026: Silicon Data and Ornn began publishing rental-price benchmarks, the prediction markets Kalshi and Polymarket listed contracts on them, Architect's Bermuda-regulated AX exchange began trading an H100 perpetual future, and CME Group, Intercontinental Exchange (ICE) and Nodal Exchange announced cash-settled futures.^{[3][4][5][6][7][8][9]} As of October 15, 2026, the CME and ICE futures had not been listed. The Commodity Futures Trading Commission (CFTC) had extended its review of CME's H100 and B200 contracts to November 9, 2026, ICE had tied its own listing to the close of a CFTC request for comment, and that request, open until October 20, asked whether compute is standardized, liquid and observable enough to underlie a regulated derivatives market.^[10][11][2]

Background

AI developers rent much of their GPU capacity from hyperscalers and from specialized neocloud providers, and the rental price has moved sharply since the NVIDIA H100 reached the market. Writing for Latent Space in October 2024, Eugene Cheah put the starting H100 rental rate at about $4.70 per hour in March 2023. He wrote that rates rose above $8 per hour during the 2023 shortage, fell to about $2.85 by early 2024, and cleared at $1 to $2 in some short-term auctions by August 2024.^[12] SemiAnalysis, which surveys contract prices, reported that one-year H100 contracts bottomed at $1.70 per GPU-hour in October 2025 and reached $2.35 by March 2026, a rise of almost 40 percent, while on-demand capacity was sold out across GPU types.^[13] Ornn's H100 index, which Data Center Dynamics reported at $1.70 in early March 2026, settled at $3.21 on October 14, 2026.^[4][14]

The sums exposed to these prices are large. The CFTC's request for comment cites a working paper estimating that the compute stock installed by the fourth quarter of 2025 implied a gross compute-service flow of about $430 billion to $1.3 trillion a year, roughly 1.4 to 4.0 percent of US GDP.^[2] GPUs also became collateral. In August 2023 CoreWeave closed a $2.3 billion delayed-draw term loan co-led by Magnetar and Blackstone and backed by a pool of H100 servers, and other GPU cloud providers later arranged GPU-backed facilities of their own.^[15] Renters paying for training and inference, operators selling capacity, and lenders holding GPU-backed loans therefore all carried exposure to rental rates. Silicon Data said its H100 index, launched in May 2025, was the first daily benchmark for GPU rental pricing.^[3]

Compute also differs from the commodities that usually support futures markets because it cannot be stored: a GPU-hour that goes unused today cannot be sold next month. Federico Bandi and Yinan Su argue that this breaks the usual no-arbitrage link between spot and futures prices, so a compute futures price should instead equal the expected future spot price minus a risk premium. Their preliminary evidence, built from synthetic futures derived from term rental contracts, is consistent with a positive premium that they attribute to hedging pressure from compute providers.^[16] The CFTC's notice draws a comparison with electricity, quoting an earlier Commission statement that electric energy cannot be bought in large quantities ahead of time and stored for later use.^[2]

How do compute derivatives work?

The exchange products announced so far all settle against a price index rather than through physical delivery. The CFTC said in August 2026 that it preliminarily expects most compute derivatives to be cash-settled at first because of the difficulty of delivering the underlying commodity.^[2] NYMEX lists the settlement type of its proposed contracts as "Financial", with the final price set from Silicon Data's index; an expiring position is closed out in cash, and the holder does not receive access to a GPU, a cluster reservation or a service-level agreement.^[1]

Sources: NYMEX, ICE, Architect, Kalshi, Polymarket, FalconX, Auctionomics and eWeek.^{[1][11][17][18][5][19][20][21]}

The arithmetic of the NYMEX contracts shows how a dated future works. Each contract covers 730 GPU-hours, one twelfth of a year of continuous use of one GPU, and settles on the arithmetic average of Silicon Data's daily on-demand index over the contract month. At an average of $2.80 per GPU-hour, one contract would settle at a value of $2,044, and the $0.01 minimum price change is worth $7.30.^[1]

Price indices

The derivatives depend on a handful of commercial index providers, which use different raw data and measure different parts of the rental market.

Sources: Silicon Data, Ornn, ICE, Nodal Exchange, SemiAnalysis and Kalshi.^{[22][23][24][14][8][25][9][13][26]}

Silicon Data

Silicon Data, led by former Bloomberg executive Carmen Li, launched its H100 rental index (Bloomberg ticker SDH100RT) on May 27, 2025. It said the index drew on about 3.5 million pricing data points, named DRW and Jump Trading among its backers, and described the index as "a first step toward turning GPU compute into a properly benchmarked, tradeable asset class."^[3] Its methodology standardizes observations for machine specifications, rental terms, platform performance and location.^[22] The company added a B200 index in December 2025, an H200 neocloud index in July 2026 and a B300 neocloud index in September 2026.^[23] CME Group, which plans to list futures on the H100 and B200 series, was among the investors in Silicon Data's $30.5 million Series A round in August 2026.^[27]

Ornn Compute Price Index

Ornn's OCPI takes a narrower approach. Its methodology, first published on July 24, 2026, counts only on-demand rentals and only executed trades in which compute passed between two different counterparties; offered and indicative prices are excluded, and the winsorization percentile is not published.^[24] Ornn's public market page shows series for the H100, H200, B200 and the Ampere-generation A100, and the company also runs a memory index covering DRAM, flash and module spot prices.^[14] The joint ICE and Ornn announcement described OCPI as "the first compute index to be built only from printed transactions."^[8]

COIL, Compute Desk and SemiAnalysis

NATIVX, based in San Juan, Puerto Rico, publishes the COIL Index, which normalizes tokenized GPU compute and connectivity prices to a single energy unit so that regional differences in power costs drop out.^[25] Compute Desk builds daily blended GPU price indexes from verified private over-the-counter transactions, distributes them on Bloomberg and Refinitiv, and calls them the first IOSCO- and BMR-compliant compute indexes.^[9] SemiAnalysis launched its H100 one-year rental index in April 2026 as a monthly survey of contract prices.^[13]

Why do the indices differ?

The indices do not agree with each other. On October 14, 2026, OCPI-H100 settled at $3.21 per GPU-hour, while Silicon Data's H100 neocloud reading for October 15 was $2.81.^[14][22] The two measure different things: OCPI samples executed on-demand trades, and Silicon Data standardizes rental observations from several kinds of provider.^[24][22] SemiAnalysis has argued that most GPU rental volume trades on long-term contracts, that providers rarely change on-demand prices, and that indices built from on-demand listings or posted prices tend to lag the contract market.^[13] Published list prices across providers are compared in cloud GPU pricing comparison.

Venues and contracts

Sources: exchange filings and notices, company announcements and market pages.^{[1][10][28][11][25][9][17][6][29][18][5][19]}

Exchange-listed futures

CME Group announced its partnership with Silicon Data on May 12, 2026. Chairman and chief executive Terry Duffy called compute "the new oil of the 21st century," and the companies described the planned product as a "first-in-class" futures market.^[7] On August 11, NYMEX asked the CFTC to approve two contracts, Silicon Data H100 Rental Index Futures and Silicon Data B200 Rental Index Futures, under Regulation 40.3, which provides for formal Commission approval, and said that it intended to list them for trade date October 5, 2026.^[1] On September 21, the CFTC's Division of Market Oversight extended its review by 45 days, "until the end of November 9, 2026," citing "novel or complex issues" and the Commission's pending request for comment.^[10] The October 5 date passed without a listing. As of October 15, 2026, the CFTC's product register showed both contracts as "Approval Pending (90)", with no approval recorded.^[28][30]

ICE and Ornn announced on May 19, 2026 that they planned dollar-denominated, cash-settled GPU compute futures on OCPI, which could reference series for the H100, H200, B200 and RTX 5090, "subject to regulatory approval."^[8] On September 29, ICE Futures U.S. named the first two contracts, OCPI H100 Futures (symbol HPR) and OCPI B200 Futures (symbol BKL). The notice said they would be listed "pending completion of the request for comment period on compute futures" and that "no listing date or timeline has been set." In the meantime, ICE said, it would publish OCPI data on its trading platform from October 5, 2026, including hypothetical daily settlement prices, so that firms could build systems and become familiar with the products.^[11] ICE's design differs from CME's in three ways: the contract size is the number of hours in the contract month instead of a fixed 730, prices are quoted to a tenth of a cent per GPU-hour, and up to 24 consecutive months may be listed rather than 36. Both settle on the average of daily index values over the month and stop trading on the month's last business day.^[11][1]

ICE announced a second compute product on July 1, 2026: dollar-denominated, cash-settled futures on NATIVX's COIL Index, to trade alongside its natural gas and power futures. NATIVX founder Cole Crawford said that "Compute is now an asset class, and like every asset class, it needs a public price and a market."^[25] Nodal Exchange, part of EEX Group, followed on September 3 with plans for financially settled futures on Hopper and Blackwell GPUs against Compute Desk's indexes, cleared through Nodal Clear.^[9]

Perpetual futures

Architect Financial Technologies, whose chief executive Brett Harrison was formerly president of FTX US, announced on January 21, 2026 that its AX exchange would list perpetual futures on GPU rental prices and DRAM prices using Ornn Data's indices, pending regulatory approval.^[31][32] AX is operated by Architect Bermuda Ltd., which is regulated by the Bermuda Monetary Authority rather than the CFTC. Architect said AX served institutions such as hedge funds, market makers, family offices, asset managers, insurers and lenders in eligible jurisdictions, and The Block reported that the exchange was aimed at non-US institutional traders and described the new contracts as "crypto-style" perpetual futures.^[31][32] Ornn announced on April 3, 2026 that its H100 contract was "now live" on AX.^[6]

As of October 15, 2026, Architect's AX page listed the OCPI-H100 perpetual as live and showed H200, B200 and B300 contracts on its roadmap. It described daily funding payments that anchor the contract to the spot price, leverage of up to 25 times, cross-margining against dollar and USDC collateral, and onboarding that requires KYC/AML and accredited-investor checks.^[17] Architect also said on May 28, 2026 that it would launch a US designated contract market, the American Innovation Exchange, after acquiring IMX Health LLC, and would list futures and options on compute costs there, pending regulatory review.^[29]

Prediction markets

Kalshi, which operates under CFTC oversight, began listing contracts on GPU rental prices in early March 2026, resolved with Ornn data. "You can now trade AI compute prices on a CFTC-regulated exchange," Ornn co-founder Kush Bavaria said at the time.^[4] By July its compute markets covered the H100, H200, B200, A100 and RTX 5090, in weekly, monthly-average and quarterly forms.^[33] The weekly price ladders resolve on the GPU's hourly price at 4 p.m. ET on a Friday "as reported by Ornn," and the monthly-average markets on "the arithmetic mean of hourly values reported by Ornn."^[18][34] Kalshi also lists weekly markets on DDR5 memory spot prices reported by Ornn.^[35]

On July 14, 2026, Kalshi launched "compute forward curves" for the B200, H200 and A100, derived from its weekly and monthly market prices. It said the curves "are not tradeable assets" but could serve as reference prices for swaps and over-the-counter compute deals. Chief executive Tarek Mansour said: "Compute is the new oil. Like every commodity before it, it needs a real derivatives market."^[26] Data firm Allium counted $4.4 million in notional volume on Kalshi's GPU rental markets through July 27, 2026, against $285,000 on Polymarket, Fortune reported.^[36]

Kalshi's market listings show weekly GPU price ladders running through the week ending October 9, 2026. As of October 15, its H100 weekly series page still pointed to that week's market and no later weekly GPU events were listed, while a weekly DDR5 market for the week ending October 16 was open and monthly-average GPU contracts extended into 2027.^[18][35][34] The same H100 pages displayed a chart labeled "H100 Forward" showing the "market expected compute price per hour."^[34]

Polymarket opened its first H100 rental-price market on February 3, 2026. That market resolved on Silicon Data's SDH100RT, traded about $337,000, and ended with the index having touched $2.40 during February without reaching $2.45 or falling to $2.25.^[5] Later Polymarket markets, including H100 contracts for the end of 2026 opened on July 10, resolve on Ornn's H100 index as displayed on Ornn's dashboard.^[37]

Over-the-counter and physical markets

FalconX said on May 27, 2026 that it had executed "the first over-the-counter swap referencing the forward price of compute," a trade on Ornn's H100 index with Robert Leshner, founder of Superstate, as the counterparty. FalconX did not disclose the size of the trade.^[19] Bavaria had told The Block in January that Ornn had "already executed our first compute swap," without naming a counterparty.^[32]

Before index-settled contracts appeared, some companies tried to make GPU time itself tradable. In July 2025 OneChronos and Auctionomics, the market-design firm co-founded by Nobel laureate Paul Milgrom, announced plans for an off-exchange market in which buyers and sellers would use bilateral forward contracts to lock in future GPU capacity and prices, matched through combinatorial auctions.^[20] SF Compute, which raised a $40 million Series A at a $300 million valuation in November 2025, sells long-term compute contracts financed by outside investors and lets buyers resell unused capacity on its marketplace.^[21] The CFTC has said that a developed market would likely need standardized terms both for the index used in cash settlement and for compute that is physically delivered.^[2]

Regulation

CFTC request for comment

In the United States, compute futures and event contracts fall under the CFTC, which regulates the exchanges, known as designated contract markets, that list them.^[2] On August 19, 2026, the Commission requested public comment on the listing of compute derivatives. Chairman Michael Selig said: "America cannot win the AI race without a robust derivatives market for compute," and called the request "the first step toward establishing clear rules of the road for American compute markets."^[38] The notice appeared in the Federal Register on August 21, with comments due by October 20, 2026; its voting summary records that Selig voted for it and that no commissioner voted against.^[2]

The notice sets out the Commission's preliminary views. It says compute markets are fragmented, that price formation happens mainly in "opaque bilateral transactions," that dominant participants "may wield significant pricing power," and that prices vary widely across providers, regions and contract structures. It concludes that compute "may not yet exhibit certain of the characteristics of commodities that typically underlie a commodity derivatives market, including fungibility, standardization, and sufficient liquidity."^[2] Among its questions:

• whether it would be appropriate to permit trading in a contract that settles to a price computed from data the Commission "may not be able to observe, verify, or surveil";
• how to treat the possibility that the parties best placed to influence a reference price are the same parties that supply capacity or contribute the transactions or posted rates behind it;
• what customer protections and disclosures are needed for a contract "settling against a geopolitically sensitive commodity," including for retail participants;
• whether perpetual compute futures offer risk-management features that dated futures cannot, and whether they pose unique risks.^[2]

The notice also says the underlier could be a different compute-related commodity, such as access to a stated volume of LLM inference tokens, and invites comment on swaps that settle to the price of compute and on physically settled compute swaps.^[2]

Effect on listings

The request has shaped listing timelines. The Division of Market Oversight's September 21 letter extending the NYMEX review referred to it directly, and ICE's September 29 notice tied the listing of its OCPI contracts to the end of the comment period.^[10][11] Offshore, AX had been trading its H100 perpetual since April under Bermuda regulation while the CFTC was still asking what advantages and risks perpetual compute futures would bring.^[6][17][2]

The delay drew a complaint from CME. At a CFTC Innovation Advisory Committee meeting in August 2026, Duffy questioned why Kalshi could offer compute prediction markets while CME's compute contracts remained under review, in an exchange with Selig and Kalshi co-founder Luana Lopes Lara.^[39]

Reported Commerce Department intervention

On September 15, 2026, Semafor reported, citing unnamed people familiar with the matter, that the Commerce Department had ordered Kalshi in August to unpublish its compute forward curve on national-security grounds. According to the report, Kalshi complied while leaving many underlying markets open, and Commerce also pushed the CFTC to effectively freeze approval of new compute contracts for 60 days. A Commerce spokesperson said the department "has never once asked Kalshi to take down this market or any other markets," and a spokesman called the story false. Kalshi declined to comment and the CFTC did not respond.^[40]

Uses and participants

Exchanges and index providers describe three main groups of hedgers. According to Data Center Dynamics, Ornn hoped a traded market would let AI businesses lock in training and inference costs through long hedges, let data center companies presell capacity and stabilize revenue through short hedges, and let investors hedge against GPU depreciation.^[4] Kalshi aimed its forward curves at neoclouds, data centers and hyperscalers seeking to hedge and price capacity, and at buyers such as inference and training labs that want to lock in prices months ahead.^[26] CME and Silicon Data said their futures would "allow traders, financial institutions, AI builders and cloud-service providers to manage volatility and price risk."^[7]

Sources: Ornn, Kalshi, CME Group and Architect.^{[4][26][7][31]}

GPU-backed lending grew after 2023, and Architect listed lenders and insurers among the institutions AX serves.^[15][31] Chris Zeoli of Data Gravity, reviewing those loans in September 2026, concluded that "the real collateral is contracted cash flow" rather than the hardware itself.^[15] Harrison told The Block in January 2026 that Architect was talking to market makers about supporting liquidity at launch.^[32]

The CFTC's notice says a compute futures market could help participants manage price risk and discover prices, and it quotes the view that such contracts would let financial markets "aggregate and reveal information about the future of the AI economy."^[2] Mansour has predicted that compute will become a $10 trillion industry by 2030 and that compute futures could reach $100 trillion to $150 trillion, according to Fortune.^[36]

Criticism and risks

Thin and opaque markets

Trading so far has been small next to established commodity markets; Fortune noted that oil futures trade more than $70 billion a day.^[36] Allium found that two days before settlement, Kalshi's compute contracts had typically been about 10 percent away from the final settlement price, and called them "not ready for prime-time as a serious hedging mechanism." Stanford finance professor Darrell Duffie told Fortune that Kalshi lacked the infrastructure and participant capital for large-scale risk transfer and said he expected large banks to dominate over-the-counter compute swaps. Kalshi's head of research, Nicole Kagan, said: "Compute is very different in that it's very opaque."^[36] The CFTC's preliminary view is that compute price formation happens mainly in opaque bilateral transactions, which limits the availability of current and historical price data.^[2]

Basis risk

A hedge works only if the index moves with the hedger's actual costs or revenue. OCPI covers on-demand rentals only and excludes reserved and long-term contracts, while SemiAnalysis reports that most rental volume trades on long-term contracts.^[24][13] Even a well-matched index leaves gaps. Zeoli wrote that futures "hedge the hour, not the box," meaning that they do not cover a GPU's generation, condition, location or redeployment costs.^[15] Because the contracts settle financially, a hedge pays cash but cannot supply GPUs if capacity runs short.^[1]

Index methodology and governance

Index rules change. In December 2025, Silicon Data removed a divisor adjustment from its H100 and A100 calculations, revised its pricing model and provider weights, and restated both series back to September 2024; it estimated the effect at a 4 to 6 percent reduction for the H100 index and a 35 to 40 percent increase for the A100 index.^[23] Ornn's methodology was first published in July 2026, and as of October 15, 2026 it described the index's oversight committee, which is to have three to five members with an independent majority, as still being established.^[24]

Conflicts of interest

Ornn's methodology says its affiliates "engage in commercial activities involving transactions in and delivery of physical GPU compute capacity," and that although index calculation is operationally separate, "common ownership creates a structural conflict." Ornn says it is developing formal conflicts procedures, that neither it nor its affiliates trade products indexed to OCPI, and that trades on affiliated venues receive no preferential treatment in the index.^[24] The Block reported in January 2026 that HydraHost, a GPU broker, was Ornn's largest data contributor.^[32] The CFTC has asked how to handle cases in which the parties that supply capacity or contribute index data are also best placed to move the reference price.^[2]

Manipulation and surveillance

NYMEX told the CFTC that its reference index is "sufficiently broad in definition and scope and supported by a robust level of trading in the underlying cash market" to resist manipulation.^[1] The CFTC's questions point the other way, asking whether contracts should settle on data the Commission cannot observe, verify or surveil.^[2] Market participants quoted by Semafor suggested that thinly traded compute markets could be manipulated to show a sharp drop in the cost of older chips, which might unsettle AI stocks and debt markets, and noted that thin trading could cause volatility even without bad actors.^[40]

Leverage and retail access

Perpetual contracts on AX can be traded with leverage of up to 25 times, and Kalshi lets individuals open accounts and trade its compute contracts.^[17][18] The CFTC's request asks what protections retail users of compute derivatives need and whether contract terms should disclose risks particular to compute.^[2]

Timeline

See also

• NVIDIA AI compute financing platforms
• AI capex
• List of AI compute deals
• AI bubble
• Compute governance

References

1. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10"NYMEX Submission No. 26-370: CFTC Regulation 40.3(a) Voluntary Submission for Product Approval, Initial Listing of Two (2) Compute Futures Contracts." New York Mercantile Exchange (filed with the CFTC), 2026-08-11. cftc.gov/...ptc08112615405.pdf
2. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12 ^13 ^14 ^15 ^16 ^17 ^18"Request for Comment on the Listing of Compute Derivatives Contracts." Federal Register, Vol. 91, No. 161, pp. 54259-54264 (US Commodity Futures Trading Commission), 2026-08-21. govinfo.gov/...2026-17163.pdf
3. ^1 ^2 ^3 ^4"Silicon Data Launches World's First Daily GPU Rental Index to Bring Transparency to the AI Infrastructure Economy." Silicon Data (EIN Presswire), 2025-05-27. einpresswire.com/...-the-ai-infrastructure-economy
4. ^1 ^2 ^3 ^4 ^5 ^6Sebastian Moss. "Kalshi users able to wager on Nvidia GPU compute prices, based on Ornn's compute derivatives platform." Data Center Dynamics, 2026-03-04. datacenterdynamics.com/...ute-derivatives-platform
5. ^1 ^2 ^3 ^4 ^5"GPU rental prices (H100) hit___ in February?" Polymarket, 2026-02-03. polymarket.com/...ntal-prices-h100-hit-in-february
6. ^1 ^2 ^3 ^4 ^5Ornn (@OrnnExchange). "Excited to announce Ornn-H100 is now live on @Architect_Fi AX exchange!" X, 2026-04-03. x.com/...2040071543993061661
7. ^1 ^2 ^3 ^4 ^5"CME Group and Silicon Data Partner to Launch First Compute Futures." CME Group and Silicon Data (PR Newswire, via Silicon Data newsroom), 2026-05-12. silicondata.com/...to-launch-first-compute-futures
8. ^1 ^2 ^3 ^4 ^5"ICE and Ornn to Launch GPU Compute Futures Contracts." Intercontinental Exchange, 2026-05-19. ir.theice.com/...default
9. ^1 ^2 ^3 ^4 ^5 ^6"Nodal Exchange and Compute Desk to launch Compute Futures Contracts." Nodal Exchange, 2026-09-03. nodalexchange.com/...nch-compute-futures-contracts
10. ^1 ^2 ^3 ^4 ^5 ^6"Extension of Initial Review Period of NYMEX Submission No. 26-370 and Submission No. 26-370S for an Additional 45-Days." US Commodity Futures Trading Commission, Division of Market Oversight, 2026-09-21. cftc.gov/...orgdcmnymexcompcontr260921.pdf
11. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8"New Compute Index Reference Data to Be Made Available on Monday, October 5, 2026." ICE Futures U.S. exchange notice, 2026-09-29. ice.com/...Notice_ORNN_Compute_4Q2026_20260929.pdf
12. ^Eugene Cheah. "$2 H100s: How the GPU Rental Bubble Burst." Latent Space, 2024-10-11. latent.space/...gpu-bubble
13. ^1 ^2 ^3 ^4 ^5 ^6"The Great GPU Shortage - Rental Capacity - Launching our H100 1 Year Rental Price Index." SemiAnalysis, 2026-04-02. newsletter.semianalysis.com/...age-rental-capacity
14. ^1 ^2 ^3 ^4"GPU Market Prices: Compute Price Index (OCPI)." Ornn Data, accessed 2026-10-15. data.ornn.com/markets
15. ^1 ^2 ^3 ^4 ^5Chris Zeoli. "How the GPU Became Collateral." Data Gravity, 2026-09-10. datagravity.dev/...how-the-gpu-became-collateral
16. ^Federico M. Bandi and Yinan Su. "(Early) AI Compute Asset Pricing." arXiv:2607.12156 (v3 revised 2026-09-10), 2026-07-13. arxiv.org/...2607.12156
17. ^1 ^2 ^3 ^4 ^5"AX." Architect, accessed 2026-10-15. architect.co/ax
18. ^1 ^2 ^3 ^4 ^5 ^6"NVIDIA H100 - Hourly price on Oct 09 (Week 41)." Kalshi market page (series KXH100WS), accessed 2026-10-15. kalshi.com/...kxh100ws
19. ^1 ^2 ^3 ^4"FalconX Executes First OTC Compute Forward Trade." FalconX, 2026-05-27. falconx.io/...utes-first-otc-compute-forward-trade
20. ^1 ^2 ^3"Auctionomics and OneChronos Partner on First Tradable Financial Market for GPU Compute." Auctionomics and OCX Group (Business Wire, via Yahoo Finance), 2025-07-29. finance.yahoo.com/...tner-first-tradable-130000332
21. ^1 ^2"SF Compute Raises $40M to Build Marketplace for AI Compute Capacity." eWeek, 2025-11-27. eweek.com/...sf-compute-ai-funding
22. ^1 ^2 ^3 ^4"H100 Rental Price Index." Silicon Data, accessed 2026-10-15. silicondata.com/...h100
23. ^1 ^2 ^3 ^4"GPU Index Announcements." Silicon Data documentation, accessed 2026-10-15. docs.silicondata.com/...gpu-index-announcements
24. ^1 ^2 ^3 ^4 ^5 ^6 ^7"Ornn Compute Price Indices (OCPI): Methodology." Ornn Data, 2026-07-24 (accessed 2026-10-15). data.ornn.com/methodology
25. ^1 ^2 ^3 ^4 ^5"ICE and NATIVX to Launch Energy-Normalized Compute Futures Contracts." Intercontinental Exchange, 2026-07-01. ir.theice.com/...default
26. ^1 ^2 ^3 ^4 ^5"Kalshi Announces Compute Forward Curves." Kalshi News, 2026-07-14. news.kalshi.com/...compute-forward-curves
27. ^"Silicon Data raises $30.5M to price the compute economy." Dealroom, 2026-08-13. dealroom.co/...-30-5m-to-price-the-compute-economy
28. ^1 ^2 ^3"Designated Contract Market Products: Silicon Data H100 Rental Index Futures (Filing 62544)." US Commodity Futures Trading Commission, accessed 2026-10-15. cftc.gov/...62544
29. ^1 ^2 ^3"Architect Financial Technologies to Launch U.S. Futures Exchange for Compute and AI Commodities Following DCM Acquisition." Architect Financial Technologies (PR Newswire), 2026-05-28. prnewswire.com/...lowing-dcm-acquisition-302784564
30. ^"Designated Contract Market Products: Silicon Data B200 Rental Index Futures (Filing 62545)." US Commodity Futures Trading Commission, accessed 2026-10-15. cftc.gov/...62545
31. ^1 ^2 ^3 ^4 ^5"Architect Financial Technologies Partners with Compute Index Provider Ornn to Launch Exchange-Traded Futures on GPU and RAM Prices." Architect Financial Technologies (PR Newswire), 2026-01-21. prnewswire.com/...-on-gpu-and-ram-prices-302666613
32. ^1 ^2 ^3 ^4 ^5James Hunt. "Former FTX US president Brett Harrison's Architect expands crypto-style perpetual futures into AI compute markets." The Block, 2026-01-21. theblock.co/...ual-futures-into-ai-compute-markets
33. ^"Compute has a Market-Implied Forward Curve Now." Kalshi News, 2026-07-14. news.kalshi.com/...gpu-compute-forward-curve
34. ^1 ^2 ^3"NVIDIA H100 - Average hourly price in October." Kalshi market page (series KXH100MS), accessed 2026-10-15. kalshi.com/...kxh100ms
35. ^1 ^2"DDR5 16Gb (2Gx8) 4800/5600 - price on Oct 16 (Week 42)." Kalshi market page (series KXDDR5WS), accessed 2026-10-15. kalshi.com/...kxddr5ws
36. ^1 ^2 ^3 ^4Jeff John Roberts. "Kalshi's CEO is racing to build a futures market for AI's most precious resource, which could be worth $100 trillion by 2030." Fortune, 2026-08-12. fortune.com/...ompute-futures-cme-intercontinental
37. ^"GPU rental prices (H100) hit___ in 2026?" Polymarket, 2026-07-10. polymarket.com/...00-hit-in-2026-20260709165909063
38. ^1 ^2"CFTC Requests Comment on the Listing of Compute Derivatives Contracts (Release 9286-26)." US Commodity Futures Trading Commission, 2026-08-19. cftc.gov/...9286-26
39. ^"CME's Duffy Clashes With Kalshi and the CFTC Chair Over Prediction Market Manipulation." Unchained, 2026-08-21. unchainedcrypto.com/...diction-market-manipulation
40. ^1 ^2 ^3Reed Albergotti and Liz Hoffman. "Exclusive: Trump administration orders Kalshi to scrap AI price tracker over national security concerns." Semafor, 2026-09-15. semafor.com/...ake-down-ai-compute-futures-product