Author: Jae, PANews
Computing power has become the "new oil of the 21st century" supporting the operation of AI globally. The AI-driven arms race in computing power is transcending the physical boundaries of information technology and deeply penetrating the lifeblood of modern financial infrastructure.
Larry Fink, head of BlackRock, a global asset management giant, once pointed out that a futures market linked to computing power might emerge given the scarcity of AI ecosystem resources. This prediction was concretely confirmed in May.
Within a week, the two leading players in the traditional financial markets, the CME Group and the Intercontinental Exchange (ICE), the parent company of the New York Stock Exchange, announced their entry into the GPU computing power futures market one after the other.
Computing power is transforming from an intangible technological resource into a standardized financial asset that can be speculated on, traded, and hedged. The fierce competition among Wall Street giants for pricing power over this new type of macro commodity also marks the official start of the financialization of computing power assets.
GPU futures become a new battleground on Wall Street: ICE dominates the entire market, while CME is vying for a leading position.
In this battle to financialize computing power assets, the two Wall Street giants have each chosen different entry paths.
On May 19, ICE, together with data provider Ornn, made a strong entry into the market, planning to launch a series of GPU computing power futures contracts based on the Ornn Computing Power Price Index (OCPI).
The OCPI introduced by ICE is the world's first computing power index built on real transaction records. Ornn distributes this index to Bloomberg terminals in real time through its subsidiary Ornn Data to ensure the transparency of pricing data and thus avoid the problem of "distorted listing prices".
Ornn co-founder and CEO Kush Bavaria believes that computing power has grown into a trillion-dollar market, and ICE's futures listing will provide a risk transfer layer for institutional buyers and computing power operators.
ICE's computing power futures contracts not only cover mainstream enterprise-grade high-end GPUs such as the H100, H200, and B200, but also include high-end consumer graphics cards such as the RTX 5090, providing refined hedging options for computing power needs in different scenarios. This means that ICE is attempting to seize pricing power across the entire computing power domain, from the cloud to the terminal, and from training to inference.
To further solidify the index's industrial foundation, Ornn has also brought in Hyperbolic Labs, one of the world's largest GPU markets, as an ally. Jasper Zhang, co-founder and CEO of Hyperbolic Labs, pointed out that the current GPU market is increasingly resembling the global commodity market, and ICE's strategy precisely addresses the risk management pain points of new computing power service providers (Neoclouds) and AI labs.
Rather than saying ICE is actively entering the computing power futures market, it's more accurate to say it's accelerating its pursuit. In fact, CME had already taken the lead a week earlier.
On May 12, CME announced a partnership with Silicon Data, a GPU market intelligence and benchmark data provider backed by trading giant DRW, to launch the world's first computing power futures contract. As a benchmark in the global derivatives market, CME's entry signifies that computing power has officially been included in the category of "macro commodities" recognized by Wall Street.
Unlike ICE's broad-based approach, CME's computing power futures are anchored to the "H100 Leasing Index" compiled by Silicon Data. By providing daily standardized tracking of real-time on-demand leasing rates from mainstream cloud service providers and emerging GPU cloud platforms, it establishes a unified pricing benchmark for the highly fragmented and opaque spot market.
To avoid depreciation and transportation losses during the physical hardware delivery process, CME's GPU futures contracts will adopt a cash settlement model. The underlying asset is not the actual chip, but the expectation of future H100 leasing prices.
For large-scale cloud service providers, this provides a necessary hedging tool. When cloud service providers invest billions of dollars to purchase H100, they only need to establish short positions in the CME computing power futures market to lock in the minimum return on investment (ROI) of the servers in advance, thereby resisting the risk of asset impairment caused by a sharp drop in computing power prices.
This approach is very similar to the logic behind turning crude oil, natural gas, and electricity into commodities back then.
Computing power futures ignite a battle for pricing power, bringing both opportunities and challenges to financialization.
Since the wave of large-scale modeling swept the globe, computing power has transformed from an "IT resource" into a "strategic resource" fiercely contested by the AI trio (OpenAI, Anthropic, and Google) and Silicon Valley giants like Meta. In short, whoever hoards the most GPUs holds the ticket to the AI era.
But problems also arise: the computing power market is too expensive and too unpredictable.
Amazon AWS, Microsoft Azure, Oracle, and Google GCP, the four cloud giants, control approximately 78% of global IT power capacity and 69% of H100 supply. Spot leasing prices sometimes surge several times over, and sometimes plummet with chip upgrades. If an AI lab wants to lock in computing power a year in advance, it may have to pay double the premium; if it doesn't, it faces a supply disruption.
What's more troublesome is that there are no hedging tools in the computing power market.
DRW founder Don Wilson frankly stated that the explosive growth of heavy asset investments such as data centers in the past has been limited by the lack of effective risk management tools, and the launch of the computing power futures market will be a solution to this pain point.
It can be said that whoever controls the pricing power of computing power controls the Bretton Woods system of the AI era.
The battle between Wall Street's two giants for pricing power in computing power reveals that this emerging factor of production is at a historical juncture of "financialization" and "commodification." This evolution is supported by industry cycles, but also accompanied by potential risks that cannot be ignored.
From a supply and demand cycle perspective, the global computing power market is entering a new phase of rebalancing supply and demand. Although the explosive growth of AI applications in the early stages led to a severe mismatch between supply and demand for high-end GPUs and a surge in rental prices, the spot price will exhibit dramatic volatility as data center construction is completed on a large scale and chip manufacturing processes iterate. The market urgently needs forward pricing tools to smooth out risks.
However, the "intangible nature" of computing power means it cannot replicate the delivery logic of traditional commodities. Physical chips have short lifecycles, typically facing technological obsolescence or devaluation within 18 to 24 months, rendering forward contracts based on physical delivery invalid due to underlying hardware iterations. Therefore, using a "standard computing power unit," such as one hour of H100 runtime, as the benchmark for conversion, supplemented by cash settlement, has become the industry-recognized optimal solution. However, this also increases the complexity of the pricing model.
Furthermore, the supply of computing power is highly concentrated, and the spot market is essentially an oligopoly. Establishing a derivatives market within this structure inherently creates a vulnerability in the price discovery mechanism, making futures prices susceptible to indirect manipulation by spot prices.
More importantly, once the computing power derivatives market is fully launched, its leveraged nature may amplify price fluctuations in the spot market. The influx of leveraged funds and the rise in speculative fervor could drive up the procurement costs of computing power, turning small and medium-sized AI enterprises into "victims," or even evolving into a "financial hunt," further exacerbating the uneven distribution of computing power resources.
While Wall Street awaits approval, crypto players have already jumped the gun.
While Wall Street's two major exchanges are still awaiting regulatory approval, players in the crypto market have already taken action.
Back in January of this year, Architect Financial Technologies, founded by the former president of FTX US, partnered with Ornn to launch perpetual contracts linked to OCPI-H100 through its AX platform.
As more platforms follow suit, it's possible that centralized exchanges (CEXs) will gradually launch related computing power futures markets in the future. In addition, they may also introduce structured financial products for computing power or dollar-cost averaging products linked to GPU leasing rates for ordinary users, further achieving seamless integration between the crypto market and traditional financial macro assets.
Compared to CME and ICE, which are subject to strict regulation and lengthy approval processes, Perp DEX (decentralized perpetual contract exchange), which runs on smart contracts, has greater agility and the institutional advantages of permissionless innovation.
Perp DEX also doesn't require the lengthy listing process of CEXs. For example, developers only need to stake 500,000 HYPE tokens (and potentially even lower in the future) to launch perpetual contracts for computing power linked to GPU indices on Hyperliquid's HIP-3 market. This product development capability will allow DeFi to establish a global computing power speculation market that is not limited by geography or barriers to entry, outside of regular Wall Street trading hours.
However, computing power futures are still a relatively new asset class, and they carry a high degree of risk in their early stages. The computing power market is primarily over-the-counter (OTC), making data sources susceptible to manipulation. In more extreme cases, during black swan events such as technological breakthroughs or chip embargoes, computing power indices may experience discontinuous surges and crashes. Both scenarios can lead to price distortions, triggering massive, chain-reaction liquidations of highly leveraged contracts.
In any case, the competition among Wall Street giants for computing power futures marks a turning point in the integration of AI infrastructure and modern finance.
GPU computing power, which was previously regarded more as an IT resource, is now being transformed into a measurable, tradable, and hedging standardized asset, embedding the allocation logic of technological resources into the global financial system.
As computing power assets become more commoditized, their resource allocation logic may shift from solely relying on spot purchases to being more influenced by price signals from the financial markets. In the future, computing power may gradually develop a more mature price discovery mechanism and capital allocation system, similar to basic production factors such as energy and electricity.
