Woofun AI reports that Vaidik Mandloi, compiled by Block unicorn, identifies a structural shift where computing is being integrated into a comprehensive capital market, mirroring the trajectory of electricity in the 1990s. This transformation is anchored by major financial institutions CME and ICE, which have announced plans to introduce futures contracts for GPU computing time, settled via stablecoin. The thesis posits that this new liquidity forward curve will resolve the financing crisis for the largest infrastructure project since the railroads, involving key players such as Google and SpaceX.

The current state of the GPU capacity market is characterized by significant chaos and inefficiency. Google, one of the world’s three major cloud service providers, currently spends $920 million per month purchasing computing resources from SpaceX, a rocket company. This expenditure highlights the absence of a pricing benchmark in the sector. Lenders are unable to hedge against the risks associated with financing hardware, forcing reliance on blind capital allocation. The lack of standardized metrics means that capital deployment is often speculative rather than strategic, creating a fragile foundation for the rapid expansion of AI infrastructure.

To understand the market structure, one must distinguish between inventory commodities and liquid commodities. Oil serves as a prime example of an inventory commodity; it can be stored in tankers until a buyer is found, allowing traders to stock up when prices are low and sell when they soar. In contrast, computing power is a liquid commodity. Users rent a GPU for a specific period, and any unused capacity during that rental window is lost forever. GPUs sitting idle in racks are not equivalent to "stored computing," just as a disconnected power plant is not equivalent to 'stored electricity.' The valuable product is the flow—GPU hours or kilowatt-hours—not the physical machines that generate that flow.

This distinction is crucial for pricing dynamics. Inventory commodities possess an inherent stabilizer in the form of inventory, which can be released during periods of sharp price fluctuations to offset rising prices. Liquid commodities lack this buffer, leading to wild spot price fluctuations. By mid-2025, the release of NVIDIA’s next-generation Blackwell chips led to an influx of new supply into the market. This surge reduced demand for H100 GPUs, causing spot prices for computing to plummet by 70% within 18 months. The volatility was further exacerbated later in the year when the mass production of HBM chips and a lack of inventory to absorb demand caused H100 GPU prices to surge by another 48% in just four days.

For AI companies, whose training operations cost tens of millions of dollars, such volatility is a matter of life and death. Similarly, for lenders who provide over $120 billion in credit for data centers serving these hardware systems, the absence of hedging tools poses a severe risk. The stakes are high, as the inability to lock in costs or revenues exposes both borrowers and lenders to unpredictable market swings. This financial instability hinders the efficient allocation of capital necessary for the continued growth of the AI industry.

Another critical issue is the lack of standardization and geographic variance. A barrel of crude oil is identical to any other barrel of crude oil anywhere in the world, enabling trade on exchanges without physical inspection.

However, an H100 GPU in Virginia is not the same as one in Iceland. The chip, cluster configuration, and surrounding workload all affect its actual performance. Benchmark test data from global GPU suppliers show that even among nominally identical hardware, performance differences can reach up to 38%. The electricity industry faced a similar problem in the 1990s: electricity in Texas’s grid was very different from that in the Mid-Atlantic region, as transmission and local demand created varying conditions at each node in the grid.

The solution adopted by the electricity industry was to set different prices for each node and quote prices based on reference price differences. This reference benchmark is exactly what the computing market lacks today. SF Compute has created a real-time order book for GPU time, where buyers and sellers can trade time just like they do any commodity in the spot market. The logic is that once there’s a liquid spot market, index prices can be derived from trading activity. These index prices can then be used to create futures contracts settled in cash. Once data centers can sell futures contracts and lock in revenue for months ahead, they can approach lenders and show them that their revenue has been hedged, thereby securing lower interest rates and allowing for expansion.

Woofun AI data shows that another company, Silicon Data, has developed a daily index called SDH100RT, which was launched on Bloomberg terminals last May. It now aggregates 3.5 million data points from global suppliers to form a single benchmark, at a cost of just one hour of H100 GPU usage. The futures contracts recently announced by CME will be settled using this index. Several other companies are competing to develop similar indices, as becoming a reference price means they can capture a small portion of every transaction in the market, as long as the market exists. This competition drives the development of more accurate and representative benchmarks for the computing industry.

The electricity market went through a similar phase: in 1993, Nord Pool established the first electricity futures exchange, followed by more than 200 new electricity marketing companies. It took industry insiders a decade to debate whether electricity was a commodity in a legal sense, but today it’s a market worth $6 trillion annually. The computer market is currently going through a similar journey, with the establishment of spot markets and the announcement of futures contracts marking significant milestones in its evolution. This historical parallel suggests that the computing market will eventually achieve a similar level of maturity and liquidity.

Intermediaries play a crucial role in bridging the gap between index prices on Bloomberg terminals and a well-functioning capital market. The operation of computing futures markets is different from stock exchanges, where standardized stock transactions take place between anonymous buyers. Computing futures markets will be dominated by traders who act as intermediaries between GPU owners (who want to lock in revenue) and AI companies (who want to lock in costs). For example, suppose a data center in the U.S. has a large number of H200 servers available starting in October. A startup needs 500 GPUs but only cares whether the interconnection method is InfiniBand (a GPU communication medium), not the specific location of the servers. This is a very specific requirement that needs someone to handle this custom order while hedging against the risks posed by standardized indices.

This isn’t anything new; in the past, every commodity required such an intermediary to help relevant parties sort out the complex relationships of physical products and convert them into interchangeable units that can be traded on exchanges. An H100 contract on a shelf is just a custom contract that no one else can price. It can only generate income for one party through private agreements, and other parts of the financial system can’t even access it. But if it can be combined with index prices and a public settlement layer, it can become a tradable asset that lenders can use for hedging. In 2023, CoreWeave borrowed $2.3 billion using only NVIDIA GPUs as collateral—this was the first time H100 hardware received financing. Its most recent funding round received an investment-grade rating from Moody’s, based on Meta’s creditworthiness rather than CoreWeave’s own, because Meta signed a 'pay-as-you-go' contract that required payments regardless of whether computing resources were actually used.

This is where the cryptocurrency track plays a vital role. Buyers and sellers of computing resources are spread across the globe, but none of them can obtain approval from the CFTC to open accounts on U.S. commodity exchanges.

However, crypto wallets can settle payments in stablecoins, and any wallet can hold tokenized computing resources. GPU export controls have revealed the geopolitical stratification in accessing computing resources—NVIDIA, for example, cannot export cutting-edge chips to China and dozens of other countries. A computing futures market settled in stablecoins allows researchers and startups outside of regions subject to export controls to get access to pricing for computing resources and hedge costs through infrastructure that bypasses these restrictions, just as stablecoins have already done in Argentina and Nigeria.

Currently, building a GPU cluster means borrowing millions of dollars using unsecured revenue as collateral, as there are no corresponding tools in global financial markets. But a liquid forward curve allows companies to borrow at lower interest rates than those for unhedged positions, using secured revenue as collateral. This means lower costs per computing hour. So, who will build the settlement layer for the forward curve? The only solution needed for now is to establish a settlement layer that allows anyone to verify collateral and make the forward curve a public product. Currently, we can’t verify the condition of the collateral hardware, whether it’s double-pledged, or its actual utilization rate. But if GPUs and their revenue streams are tokenized as on-chain assets, every lender can verify the collateral in real time, making the forward curve publicly visible instead of remaining trapped in bilateral negotiations.

Additionally, the next generation of AI agents will purchase computing resources based on the number of inference calls they make, and they won’t be able to open bank accounts. Cryptocurrencies are the only payment gateway that can complete microtransactions between agents in Tokyo and GPU racks in Virginia in less than a second. There are currently strong checks and balances, as GPU supply is highly concentrated. The world’s top hyperscale data center operators control 78% of global IT hash rate. NVIDIA holds over 80% of the market share in high-end AI chips, and its product release schedule is enough to influence the entire market. Standardization is a bottleneck, but financializing an asset class during a boom period may make it more contagious. Over $120 billion in debt related to AI infrastructure has been shifted from balance sheets to special purpose vehicles (SPVs) funded by Wall Street, with most of it ending up in corporate bond funds within target-date retirement products, without the individuals holding these bonds knowing about it. I believe the financing models used to build this infrastructure likely contain assumptions about the residual value of hardware, and existing data isn’t sufficient to support these assumptions. The electricity market doesn’t stop at generators; it extends throughout the entire system, all the way to wall outlets, affecting the prices of all electrical devices. There’s still a lot of work to be done in the computer market!