Silicon Data’s Carmen Li pushes AI compute futures toward commodity-scale tradeability

Silicon Data’s Carmen Li believes AI compute futures could eventually rival some of the world’s largest commodity markets. That single claim is doing a lot of heavy lifting: it suggests that what currently feels like a bespoke, capacity-constrained scramble to get GPUs and the power to run them could become a standardized, tradeable product.

In other words, the “new oil” is not crude, it is computing power. Li’s argument, at least as framed in the CNBC write-up, is that trading structures could evolve so AI compute behaves more like a commodity complex, where buyers and sellers can lock in expectations ahead of time. If you are a CFO or a board member trying to plan for AI spend, this matters because futures and related contracts are not just financial engineering. They are how markets turn uncertainty into price signals, and price signals into planning tools.

To understand why an AI compute futures market would be a big deal, you have to start with how AI infrastructure currently operates. Demand has surged for compute, and the bottlenecks are often physical and operational: access to GPUs, the data center capacity to house them, the power and cooling to keep them running, and the scheduling systems that allocate capacity to customers. In traditional commodity markets, the product is standardized enough that contracts can be compared and traded. The entire goal of “compute futures” is to make the AI version standardized enough that financial markets can attach to it.

Commodity markets also come with a second-order advantage: they let different players share risk. Instead of every buyer negotiating individually for capacity or building entirely internal reserves, futures markets can allow hedging against volatility. That can change the incentives for both sides. For sellers and compute providers, trading structures can help them forecast revenue and manage capacity planning. For buyers, it can reduce the whiplash that comes from sudden repricing of availability. The market becomes less about who can move fastest in the moment, and more about who can manage exposures ahead of time.

There is also a regulatory and market-structure angle here, even if the CNBC source only gives the headline claim through Li’s perspective. In the real world, turning something into a futures contract usually triggers questions about oversight, transparency, and how contract pricing connects to the underlying physical asset or service. Regulators typically want to understand who is responsible for contract integrity and how trading reflects the thing being traded. Even a “compute” product would have to grapple with verification, delivery, and what “delivery” means when the asset is capacity and performance rather than a barrel of oil.

And that is why the “commodity-scale” ambition is so consequential. If compute futures become durable, the market could attract institutional capital and sophisticated risk managers who are already familiar with commodity trading frameworks. That can increase liquidity, but it can also concentrate attention on standards: performance specs, uptime expectations, capacity definitions, and how quickly the market can settle disputes. Standardization is both the unlock and the battleground. The more futures pricing becomes the reference point for budgets and procurement, the higher the stakes when definitions do not match reality.

For executives and boards at AI infrastructure firms, cloud providers, or large enterprises that buy compute, Li’s belief is a signal to watch how “tradeability” becomes product strategy. If compute starts looking like a commodity complex, leadership teams will likely be measured not only on technical throughput, but on how well they can participate in trading ecosystems that reward predictability. And for companies not positioned to offer standardized contracts, it raises a blunt question: do you want to be the customer negotiating in the dark, or the participant shaping the terms of the market?