When Computing Power Becomes a Trading Asset - From Computing Power Futures to Computing Power Dollars

Author: Web3 Research of Go2Mars

From Nvidia defining GPUs as collateralizable infrastructure assets, to CME and ICE building pricing infrastructures like computing power futures, and then to OTC swaps, on-chain bulk trading, and forward curves in prediction markets, computing power is gradually being transformed into a financial asset that can be priced, traded, and collateralized.

On August 10, 2026, Nvidia announced a memorandum of cooperation with six top global asset management institutions: Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR, aiming to mobilize over $500 billion in third-party capital and define GPUs as a collateralizable and investable infrastructure asset.

The next day, on August 11, 2026, the Chicago Mercantile Exchange Group (CME Group) announced that it would launch the world's first futures contracts linked to GPU computing power rents on October 5.

Not long before this, across the Pacific, the Shanghai municipal government had already released a document on June 2, clearly stating the need to prepare for the research and development of computing power futures. Even earlier, the spot trading on the Shanghai computing power trading platform had actually been running since 2023.

If we place these three news items side by side, a common signal emerges: computing power is transitioning from an IT resource used through leasing to a financial asset that can be priced, traded, hedged, and collateralized in the public market. Moreover, this transformation is occurring almost simultaneously worldwide: forwards have already been traded, futures are lining up for listing, indices have already appeared on Bloomberg terminals, and for the first time, the term "computing power futures" has appeared in Chinese government documents.

Based on this, this article clarifies this rapidly growing industrial chain:

• How is the GPU gradually being transformed into a financial asset?
• Who is trading it, and where is it being traded?
• When computing power truly becomes a commodity like oil, will it give rise to a new currency narrative, akin to the "petrodollar" of half a century ago, namely the "computing power dollar"?

1. Starting Point: How GPUs Are Transformed into Collateral?

1.1 Jensen Huang's Ambition: Attempting to Redefine GPUs

To understand why the computing power derivatives market exploded in 2026, we must return to a fundamental action: Nvidia's attempt to redefine what a GPU is.

On August 10, 2026, Nvidia's official press release announced that it had signed a memorandum of cooperation with six institutions—Apollo Global Management, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR—to jointly build an independent "computing power financing platform," aiming to mobilize over $500 billion in third-party capital to support AI data centers, chip factories, and supporting power facilities.

The core of this announcement is actually a taxonomic shift proposed by Jensen Huang: GPUs should no longer be viewed as rapidly depreciating tech hardware, but rather as investable infrastructure assets.

In this context, GPUs possess long-term, predictable cash flows similar to commercial real estate and toll roads. Huang's reasoning is that Nvidia's computing power assets are widely adopted by almost all cloud service providers, can be fungible among different clients, and can continuously gain performance improvements through the CUDA software ecosystem, thereby extending their effective lifespan.

"This is truly the first time that tech chips have become an investable asset class."

------Jensen Huang, August 10, 2026

Whether this argument holds is fundamentally contested by one question: Can the economic lifespan of GPUs match the terms required by traditional mortgages?

Traditional collateral—commercial real estate, cargo ships—is accepted by banks because they have mature secondary markets with decades of history, and their price fluctuations are relatively stable. In contrast, the value curve of GPUs is entirely different, determined by the pace of chip iterations rather than physical wear and tear.

According to industry research firms Silicon Data and GPU trading platform GPUSmith, the H100, which sold for about $40,000 at the end of 2023, had its transaction price in the second-hand market drop to between $12,000 and $22,000 by mid-2026, with some auction prices even falling as low as $8,200. This cliff-like, non-linear value reset is the biggest technical challenge of this GPU assetization experiment.

1.2 An Unwritten Guarantee Clause

To alleviate creditors' concerns, Nvidia proposed a key clause: it is willing to provide up to 25% residual value support for some financing transactions, meaning that if the actual residual value of the chip is lower than expected at the loan's maturity, Nvidia will bear up to 25% of the difference, with specific ratios assessed on a case-by-case basis.

One detail worth noting: according to financial analysis platform Electron Economics' line-by-line verification of the official press release from August 10, the original announcement did not actually contain the terms "residual value" or "backstop"; it merely stated that it was cooperating with six institutions to establish a large-scale special fund pool with attractive interest rates for clients, and clearly indicated that the relevant cooperation still needed to be based on the signing of final agreements. The specific expression of the "up to 25% residual value" support mechanism was actually a supplementary explanation provided by Huang in an interview the next day.

This sequence, along with the credit market's reaction, is worth examining together:

• According to ICE Data Services, Nvidia's five-year credit default swap (CDS) spread rose in July due to reports of circular trading, jumping 14 basis points in a single day on July 27, reaching a peak of 82 basis points (later corrected), marking the largest single-day intraday increase since the contract began active trading in November 2025;
• Investing.com analyzed that on the day the financing plan announcement was released on August 10, the spread rose to about 77.5 basis points. The guarantee clause did not significantly cool market sentiment.

The pricing in the credit market, to some extent, reflects more honestly the current position of this mechanism: it is still being continuously tested by the market and has not yet fully taken shape.

2. Collateral Needs a Yardstick: Exchanges Enter the Scene

2.1 Why Assetization Cannot Be Separated from Futures

For an asset to truly become a qualified collateral that can be accepted on a bank's balance sheet, it typically needs to meet three conditions: relatively stable value, a mature secondary market, and an authoritative and credible pricing benchmark. A single physical GPU is unlikely to meet all three criteria, which is precisely why established exchanges like CME and ICE entered the scene intensively in the first half of 2026—they aim to establish a recognized pricing benchmark for this emerging asset class.

On May 12, 2026, CME Group announced in conjunction with GPU market data company Silicon Data that it planned to launch the first batch of computing power futures contracts within the year. On August 11, the two parties further clarified that they would launch two futures contracts on October 5—Silicon Data H100 Rental Index Futures and Silicon Data B200 Rental Index Futures, which track the hourly rental price indices of H100 and the next-generation Blackwell B200 chips, respectively. If successfully launched, they will be listed on the New York Mercantile Exchange (NYMEX). Pete Keavey, CME's global head of energy and environmental products, likened it to crude oil futures: "Computing power has become the currency of the AI era. Just as oil powered the 20th-century economy and evolved from spot trading to a global derivatives market, our futures contracts will transform computing power into a standardized, tradable commodity."

Just a week later, on May 19, 2026, CME's old rival, the Intercontinental Exchange (ICE), announced it would collaborate with another computing power data provider, Ornn, to launch GPU computing power futures contracts based on the Ornn Computing Power Price Index (OCPI), covering everything from enterprise-level H100 and H200 to consumer-grade graphics cards like the RTX 5090. The near-simultaneous entry of these two exchanges indicates that the competition for pricing power is unfolding at a pace faster than expected.

2.2 Contract Design: Two Different Paths

From a contract design perspective, the global computing power futures currently mainly diverge into two paths:

• One path is the computing power leasing route, with GPU hourly rental price indices as the underlying asset, essentially pricing from the supply side and hardware level;
• The other path is the Token route, which anchors the demand side of computing power, pricing based on the actual number of Tokens consumed by large models, more closely reflecting the real cost perceptions of downstream AI application developers and end users.

Both CME and ICE are following the former path, while reports from financial media indicate that the Shanghai Futures Exchange in China is currently developing a computing power futures scheme along the latter path, namely the AI Token futures route, which differs from the technological route chosen in the United States and is a divergence point worth continuous attention.

It is worth noting that the claim of CME being the world's first computing power futures needs a qualifier: it is the first computing power futures to enter the regulated mainstream exchange track, while the world's first computing power derivative has another origin—before it officially listed, several transactions had already been completed in the OTC market and prediction market platforms.

3. Beyond Exchanges, The Industrial Chain Combined with On-Chain Finance

Before the standardized futures contracts on exchanges officially list, the computing power derivatives market has actually already run out a complete industrial chain—from OTC forwards to on-chain bulk trading, and then to the forward curves of prediction markets, each link has already had real transaction records, rather than just remaining in the conceptual stage.

3.1 No Index, No Derivatives

All the derivative trading that will be analyzed below—whether it is FalconX's OTC swaps, Polymarket's on-chain trading, or the futures that CME and ICE are about to launch—ultimately relies on one thing: a credible price index.

Without an index, there is no pricing benchmark; without a pricing benchmark, there is no space for the existence of any standardized derivatives.

Currently, there are two parallel index routes in the market:

• Ornn Computing Power Price Index (OCPI): Released by the computing power company Ornn, the OCPI is notable for being the world's first computing power index based solely on actual transaction records rather than quotes or advertised prices. As of April 2026, the OCPI is available on Bloomberg terminals, covering models such as H100, H200, A100, B200, and RTX 5090, and serves as the settlement benchmark for ICE computing power futures;

• Silicon Data Index: Provides an index benchmark for CME computing power futures, tracking the rental prices of GPU chips daily. CME currently plans to launch two independent contracts: H100 Rental Index Futures and B200 Rental Index Futures, with both contracts priced based on the hourly rental prices of the corresponding GPU models, without standardizing different models into a single unit of computing power.

With price indices like WTI crude oil and Brent crude oil, computing power truly meets the basic conditions to become a commodity.

3.2 Over-the-Counter Swaps: FalconX's First Transaction

On May 27, 2026, digital asset broker FalconX announced the completion of the world's first over-the-counter computing power forward price swap transaction. The counterparty was Robert Leshner, founder of the digital asset platform Superstate, with the underlying linked to the forward price of H100 in the Ornn Computing Power Price Index (OCPI), with FalconX acting as the dealer.

Ornn CEO Kush Bavaria described it as transforming a volatile and unpredictable market into a measurable, tradable, and hedgable commodity.

The amount of this transaction itself is not large, but it proves one thing: while exchange futures are still awaiting approval, institutional investors have already begun to manage computing power price risks through over-the-counter derivatives.

3.3 Polymarket: On-Chain Institutional-Level Bulk Trading

A few days later, on June 2, 2026, the prediction market platform Polymarket also announced the completion of its first institutional-level on-chain bulk trade.

The trading parties were FalconX and AI risk clearinghouse startup AneraLabs, with the settlement linked to Ornn's OCPI index, and the transaction amount reaching six figures in USD, recorded on the Polygon blockchain.

This risk transaction itself was made into a prediction market position on Polymarket, which simultaneously provides a trading venue and on-chain settlement mechanism.

According to CNBC, this is also the first institutional bulk trade explicitly linked to computing power prices (H100 OCPI index) in the prediction market industry—just a month prior, Kalshi had completed its first institutional bulk trade, but the underlying was the auction price of California carbon emission allowances. Viewed together, these two events demonstrate that prediction market platforms are simultaneously enhancing their institutional trading capabilities, with computing power being one specific category.

This transaction, compared to traditional over-the-counter swaps, adds a layer of on-chain execution and settlement infrastructure: after FalconX and AneraLabs completed bilateral negotiations, the transaction was executed through Polymarket's international platform and ultimately recorded on the Polygon blockchain, rather than relying entirely on the backend clearing systems of traditional financial institutions.

There is an easily overlooked distinction here: Polymarket operates two mutually independent platforms.

• One is the international site where this computing power transaction took place, based on Polygon, settled in USDC, requiring no identity verification, and geographically restricted for U.S. users;

• The other is Polymarket US, launched at the end of 2025, operated by the acquired QCX entity, regulated by the CFTC, settled in USD, and requiring full identity verification.

This GPU computing power transaction occurred on the former and is not under the CFTC's regulatory framework, which is fundamentally different from what will be discussed in the next section regarding Kalshi.

3.4 Kalshi: From Prediction Markets to Forward Curves

On July 14, 2026, the CFTC-regulated prediction market platform Kalshi launched the AI computing power forward curve.

According to Kalshi's official press release, this curve currently covers three chip models: Nvidia B200, H200, and A100; while the broader range of computing-related contracts on the Kalshi platform also includes H100 and RTX 5090.

Kalshi's Chief Risk Officer previously worked at CME Group for about 16 years. In an interview with Bloomberg, he stated that Kalshi is utilizing prediction markets to construct the GPU computing power forward curve, viewing it as the foundation for the future development of futures, options, and other products; he also noted that in just 2026, the capital expenditure commitments of hyperscale cloud service providers have already reached $500 billion to $600 billion.

It is important to clarify that Kalshi officially states that the forward curve itself is not a tradable asset but a reference price used for pricing in over-the-counter swaps and structured products; what is truly tradable are the underlying prediction market contracts on the Kalshi exchange.

The data for this curve is also not entirely independent of the Ornn system—some of Kalshi's computing power contracts are similarly settled based on real-time pricing data provided by Ornn. This indicates that while there are many participants in this industry chain, the underlying price index suppliers are actually highly concentrated in Ornn and Silicon Data.

3.5 Summary Comparison: Four Transactions, Four Roles

These four transactions are all computing power derivatives, but their positions are different.

• FalconX's over-the-counter swap and Polymarket's on-chain bulk trade both adopt an inter-institutional negotiated trading format, but the participation structure is not the same: the former involves FalconX as the dealer and Robert Leshner as the counterparty, with FalconX effectively acting as a market maker to complete a hedging transaction for Lashner; the latter involves FalconX and AneraLabs as mutual counterparties, with Polymarket providing the on-chain trading infrastructure (the function of executing trades) and settling on-chain via Polygon.

• Kalshi's forward curve is entirely different; it is not a contract but a reference price derived from a large number of real transactions of small prediction contracts on the platform, with the actual tradable items being those underlying contracts themselves, while the curve merely stitches their prices into a line.

• CME and ICE futures are closest to traditional commodity futures, with standardized terms, exchange listings, and centralized clearing, resembling the structure of crude oil futures.

Putting the above content together shows a clear division of labor in the industry chain: over-the-counter forwards are responsible for capturing risk management needs during the approval gap of exchanges, prediction markets are responsible for probing a market-recognized price level through high-frequency small transactions, and exchange futures are responsible for standardizing this price into contracts that anyone can place orders on.

4. China’s Side: Early Spot Trading, Slow Futures

4.1 Shanghai: From Trial Operation to Inclusion in Government Documents

Few have noticed that China actually started building infrastructure for computing power spot trading earlier than the recent wave of U.S. futures enthusiasm, and it has been rolled out along several lines simultaneously.

On the platform level, Shanghai is leading the way. The Shanghai Computing Power Trading Platform began trial operations in April 2023, launched version 2.0 in December of the same year, and upgraded to the China Computing Power Platform (Shanghai) in May 2025. However, this is just a local node; national-level expansion is also gradually beginning: the China Computing Power Platform, led by the China Academy of Information and Communications Technology, achieved full connectivity at the China Computing Power Conference in August 2025, with ten provincial and municipal sub-platforms including Shanxi, Liaoning, Shanghai, Jiangsu, Zhejiang, Shandong, Henan, Qinghai, Ningxia, and Xinjiang completing access. Shanghai is one node in this national network, not an independent entity of this achievement.

In terms of pricing benchmarks, the China Securities Index Company released the China Securities Intelligent Computing Power Supply Index Series on December 24, 2025, comprising 16 indices, covering one total index, seven regional indices, and eight national computing power hub node indices.

At the national policy level, the turning point occurred later and more clearly in two steps. In April 2026, the Ministry of Industry and Information Technology first proposed exploring innovative businesses such as computing power banks and computing power supermarkets, indicating conceptual loosening.

The real breakthrough came at the end of May: the Shanghai Municipal Government Office issued "Several Opinions on Deepening the Construction of Shanghai as a Global Asset Management Center" on May 28, publicly released on June 2, which explicitly stated the need to prepare for the research and development of electricity futures and computing power futures. ------ Thus, the term "computing power futures" officially appeared for the first time in a public document from a provincial government in China.

Prepare for the research and development of electricity futures and computing power futures, and develop more new types of futures that represent the direction of new productive forces. ------ Shanghai Municipal Government "Several Opinions on Deepening the Construction of Shanghai as a Global Asset Management Center", June 2, 2026

Interestingly, the wording of this document is "research and development preparation" rather than "upcoming listing." Compared to specific listing dates like CME's October 5, the pace is clearly more cautious and forward-looking. This document from the Shanghai Municipal Government also sets a larger goal: striving for Shanghai's asset management scale to reach 55 trillion yuan by 2030, accounting for one-third of the national total.

4.2 A Previously Undermentioned Technical Route Fork

According to an exclusive report by Reuters, the Shanghai Futures Exchange is designing a futures scheme linked to AI Tokens, which represent the smallest unit of information processed by large models, used for pricing AI services. This is completely different from the GPU leasing hour model adopted by CME and ICE. The report also notes that this research is still in its early stages: it did not disclose whether pricing will be based on the number of Tokens or the price per Token, and the timeline for regulatory approval has not yet been determined.

Even so, the two routes for computing power futures in China and the U.S. have already diverged from their design starting points: the U.S. anchors its model on the rental costs of hardware supply, while China is exploring pricing methods related to demand and Token consumption.

This divergence is not hard to understand—CME and ICE primarily serve data center operators and cloud service providers, who are concerned about how much the rental cost is per card per hour; Token pricing is closer to the real cost experience of downstream AI application developers, who care about how many Tokens are consumed to complete a task and how much it costs.

It remains to be seen which route is superior, but if China successfully navigates the Token futures path, it will correspond to a downstream application market that is much larger than the rental market for selling cards, and it has considerable potential for product performance (such as transferability, convenience, etc.), although there are significant challenges in standardization.

5. Computing Power Dollar: A Concept Under Discussion but Not Yet Established

Once computing power begins to be priced, traded, and collateralized like oil, a question naturally arises: oil trade is uniformly settled in dollars, giving rise to the oil dollar system that has lasted for half a century. Will computing power follow the same path?

5.1 Proposal of the "Computing Power Dollar" Concept

This analogy is not unique to this article. Over the past six months, several institutional-level studies have proposed similar frameworks, one of the more influential being a commentary article titled "Turning the AI Revolution into Dollar Dominance" published by Navin Girishankar, head of the Economic Security and Technology Program at the Center for Strategic and International Studies (CSIS), on December 8, 2025.

The core idea of the article is that the U.S. is exporting advanced AI chips to allies and partners to help these countries build computing power infrastructure, which will continuously produce AI services for global export. Whoever controls the settlement currency for these AI service export revenues replicates the position of the oil dollar from back in the day. (Note: This is a commentary article representing the author's personal views; CSIS officially states that its research is nonpartisan and does not represent the institution's position or established U.S. government policy.)

5.2 Core Flaw of the Concept: Computing Power Has Not Yet Established a Strong Binding with the Dollar

The biggest flaw in this concept, as pointed out by the author himself, is that these chip export agreements do not require recipient countries to settle their AI service export revenues in dollars.

He provided a calculation logic: a country might spend only $10 billion to build data center infrastructure once, but these chips could generate $50 to $100 billion in AI service export revenue annually, and there are currently no constraints on what currency this ongoing, long-term revenue will be settled in—the oil dollar relied on invisible pricing agreements, while the computing power dollar has not even taken that step yet.

To fill this gap, the author proposed three policy recommendations:

• Tie chip export licenses to a commitment to settle in dollars or dollar-pegged stablecoins;
• Sign the GENIUS Act by July 2025, using dollar-pegged stablecoins as payment settlement tools;
• Provide an economic security umbrella, as a contemporary version of the Cold War defense umbrella.

Based on this, the computing power dollar is currently more accurately positioned as a policy initiative written in a think tank commentary article, rather than a currency system that has been established and is operational.

5.3 Historical Reference: The Cycle from Proposal to Verification of New Narratives

A new asset narrative typically requires a complete cycle from proposal to actual market validation. The two bubbles around the year 2000 can serve as a reference. One is the internet stock bubble that the public is more familiar with; the other is the telecom equipment bubble, which structurally resembles the current computing power narrative—Cisco, Lucent, Nortel Networks, and other equipment vendors provided large-scale vendor credit to cash-strapped telecom operators and internet service providers to help customers buy equipment, forming a self-reinforcing loop of loans driving procurement and procurement driving up stock prices.

The entire industry expanded to over $1.5 trillion before the bubble burst, accumulating about $1 trillion in debt, resulting in over $500 billion spent on laying fiber optic and other infrastructure. For instance, Nortel Networks completed a $19.7 billion acquisition almost entirely through stock in 2000.

Interestingly, there were also attempts to make bandwidth a tradable commodity back then, which ultimately failed. The famous Enron Corporation completed its first bandwidth trade in December 1999, targeting a monthly incremental contract on the Global Crossing network from New York to Los Angeles, explicitly stating that it wanted to price this contract as a benchmark and proposed a series of subsequent expansion plans. This is the same methodology as today’s Ornn, Silicon Data making indices, and CME standardizing contracts.

The outcome was that the bandwidth market collapsed before sufficient liquidity could be established. After Enron's bankruptcy in 2001, follow-up companies like Williams, Dynegy, and El Paso gradually shut down their bandwidth trading departments, and independent bandwidth trading platforms like RateXchange exited the market.

Looking at these two bubbles together illustrates the same point: telling the story early does not mean the demand is false. Internet traffic has indeed continued to grow as expected, and the overbuilding of fiber optics has actually accelerated the decline in costs for broadband, video, and cloud computing; the bubble burst punished leverage and stock prices that outpaced real demand, not the direction itself. The current signals of AI computing power demand are equally strong, but whether they can translate into sustainable profits remains uncertain. Furthermore, historical regulatory reforms occurred after the bubble burst, whereas today, institutions like the U.S. CFTC and the Bank of England have already intervened early to observe, which is different from the near absence of regulation back in 2000.

Conclusion: The Long Road Ahead

The rapid development of computing power trading and other industrial chains does not mean they have already been fully realized. The progress of forward transactions, index launches, and futures listings can easily lead one to believe that the assetization of computing power is a done deal. However, what truly determines whether this system can stand firm is whether demand can support leverage, whether regulation can keep pace with innovation, and whether different paths can ultimately converge into a single standard. Currently, there are no answers to these three questions.

Whether the assetization of computing power will repeat the internet financial bubble of 2000 depends on several unclear variables: whether the residual value assumption of GPUs can withstand the impact of each new generation of chips, whether the computing power dollar can truly be written into contractual terms, and whether China’s Token futures and the U.S. GPU futures will ultimately converge on the same standard.

Computing power is being transformed into a financial asset, and this direction is likely irreversible. As for who will ultimately price this new system, what currency will be used for settlement, and what standards will govern its operation, the answers are still on the way.

This is a new story.