The Missing Market for Compute
Highly liquid markets exist for commodities like oil and for financial assets like stocks. No efficient market exists yet for GPU compute. Over the next 5 years, no gap matters more than efficiently pricing compute and power and financing those assets. OneChronos is focused on this.
Everything here is at an early stage. The first step is a funding market for compute: how to bring projects online, add capacity, and efficiently price both the capacity and the cost of capital for what gets built. The starting point is funding markets.
Why the Opaque Compute Market Costs You
Investors looking at Nvidia (NVDA) usually see a hardware play. The underlying compute market stays opaque. What is lost is efficiency in raising capital and the ability to underwrite larger and larger projects and recycle capital back into the underwriting system. Crude oil futures let more people lend to and finance energy projects. The AI buildout needs the equivalent.
Three Key Players
A working GPU compute market has three interconnected participants:
- Offtakers - people buying and using the compute.
- Builders - people building the compute projects.
- Lenders - people lending to the compute projects.
Combinatorial auctions, the core of what OneChronos does, are seen as key to unlocking value across these three sides.
Why Compute Is Not Oil
Crude oil's forward curve was powerful for lending markets. As an asset, oil and compute could hardly be more different, so market structure matters a lot. Do not force compute into a commodity mold. Compute is different, so price single assets. Build a world of very specific claims, then financial products and securitization around things like individual data centers, rather than pretending there is one "barrel of crude" equivalent for compute.
Existing pricing often comes down to the cost of a GPU hour. That is the wrong level for understanding the economic value of compute. You cannot make GPUs fungible at the hardware level right now. What Jensen Huang gets right is that compute can be made fungible as an investable, durable asset. That supports securitization over the crude oil model. The value sits in the claims on cash coming off GPUs. One cluster of GPUs in the Pacific Northwest is not equivalent to one in Virginia.
Spot Market and Forward Market
OneChronos works two markets: a spot market for immediate training and inference, and a financial market for longer-dated compute. Commodity markets only work well when forward-dated prices converge into the spot market.
To reach that, OneChronos starts with physically delivered compute and capital formation around those projects. Auctions bring together those who want to finance and lend to compute projects, those who want to build them, and those who want to consume the compute. Once that spot market exists, you can start defining the forward curve, which then allows different types of underwriting and securitization. You have to start with the physical market.
Hedging Compute Costs
Companies could use futures and forward contracts to lock in compute costs, the same way businesses hedge other input costs. Commodity markets set the precedent. A farmer can plan what to plant and how large a crop to grow because they can take some risk, price how much they want their crop price to move, and pre-sell some capacity. This behavior already exists in compute, where large anchor clients back the buildout of a big new cluster. It needs to get much bigger and more efficient at the scale of capital markets. That is the next layer of financialization for GPU compute.
What It Means for Nvidia
Nvidia (NVDA) now acts as a proxy for compute demand. A real compute market would give investors a clearer read on actual demand and bring transparency. For Nvidia and everyone else, it makes a much larger pie and a better story around capital formation for future projects. The past year proved that both the demand and the value are real for what people build on AI compute. What comes next is scaling it out as much as possible. That is why OneChronos starts with lending products built specifically for compute: it begins with bringing capacity online, and capital markets are a strong way to broaden access to capital.
