AI Compute Futures: The Next Big Market?

The insatiable demand for artificial intelligence is creating a new frontier in financial markets: futures contracts for AI computing power. CNBC reports that CME Group, a leading derivatives marketplace, has partnered with Silicon Data to launch the first-ever futures market for AI compute, pending regulatory review. This move aims to bring much-needed stability and predictability to the rapidly evolving and often volatile costs associated with training and running AI models.

Companies working on this

StartupHub profiles of the companies this article names, with funding and a one-liner from our database.

Silicon Data

$1.1B

Market intelligence for AI compute infrastructure.

Bloomberg L.P.

$42.6B

Global financial, software, data, and media company providing real-time information and analytics.

Databricks

$190.0B

A unified data analytics and AI platform built on the lakehouse architecture.

HumanX

$23.0B

A company that organizes premier AI conferences and publishes data-driven reports on the AI economy.

The Driving Force: AI Demand and GPU Costs

The exponential growth of AI has placed immense strain on the availability and cost of specialized hardware, particularly high-performance GPUs like NVIDIA's H100. These chips are the workhorses behind complex AI computations, and their rental prices can fluctuate significantly. As Carmen Li, Founder and CEO of Silicon Data, explains, this volatility makes it difficult for companies to forecast and manage their AI infrastructure expenses.

"You have a big cost of electricity, you consume a lot of GPUs, you consume a lot of compute," Li stated. "You will be facing big cost volatility. You want to hedge that away using futures as well." This need for hedging is precisely what the new futures market aims to address.

How Futures Contracts Bring Stability

The concept of futures markets, long established for commodities like oil and corn, is being adapted for AI compute. In traditional futures markets, buyers and sellers agree on a price for a commodity to be delivered at a future date. This locks in a price, protecting both parties from adverse market movements. For AI compute, this means companies can lock in the hourly rental rate for specific types of GPUs, such as the NVIDIA H100, months in advance.

Silicon Data has been tracking these rental prices, identifying over 50 different configurations of NVIDIA's H100 chip. Each configuration, varying in processors, RAM, connectivity, and datacenter location, trades at a different price. The futures market will aim to standardize these offerings, providing a benchmark for the cost of AI compute. As Li noted, "We want to make it easier for people to hedge. You can't do that in the spot market today."

Hedging and Speculation in the Compute Market

The introduction of futures contracts serves two primary purposes. Firstly, it allows companies that rely heavily on AI compute, such as AI startups and large enterprises, to hedge their exposure. By entering into futures contracts, they can secure computing power at a predictable price, mitigating the risk of sudden price hikes or short-term unavailability. This is crucial for long-term AI development and deployment.

Secondly, the market will attract speculators. These are traders who do not necessarily need AI compute themselves but aim to profit from anticipated price movements. Their participation adds liquidity to the market, making it easier for hedgers to find counterparties and contributing to efficient price discovery. The Commodity Futures Trading Commission (CFTC) will oversee this new market, ensuring its integrity and stability.

The Challenge of Standardization

A key challenge in creating a futures market for AI compute lies in standardization. Unlike a barrel of oil, which has established specifications, AI compute is highly granular. Silicon Data’s analysis highlights the complexity, noting that factors like processor type, RAM, connectivity, data center location, and even the specific settlement terms of a contract can influence the price. Standardizing these variables to create tradable futures contracts is a significant undertaking.

The success of this venture will depend on how effectively these complex variables can be distilled into standardized contract specifications that are both attractive to hedgers and liquid enough for speculators. The precedent set by commodities futures, where standardization has been key to market development, offers a roadmap for this nascent AI compute market.