A market across time
Compute is becoming a financial market. The opportunity reaches from the capacity available this afternoon to the infrastructure that will serve AI years from now. Both need a way to be priced, committed, transferred, and delivered.
Imagine a buyer reserving a year of capacity, using what it needs, and selling the rest. A provider meeting a delivery commitment with capacity sourced from another operator. A developer financing a new facility against demand assembled from many buyers. These are connected parts of the market Tensors is building toward.
Define what the buyer is owed
The foundation is a contract with a clear deliverable: a quantity of compute, a delivery period, a performance specification, and the conditions under which it can be used. Location, memory, interconnect, availability, and service levels belong in that definition, along with payment and transfer terms.
Standardization makes capacity comparable without treating every workload as identical. MLCommons, for example, defines inference benchmarks around particular workloads, quality targets, and scenarios. Compute contracts can take a similarly specific approach: define the service and how it will be measured, then identify which systems qualify to deliver it.
More than one path to delivery
A standardized delivery obligation can give a provider more than one way to fulfill its promise. If a facility opens late, the provider could purchase qualifying capacity from another operator and deliver that service to the holder while construction finishes. The buyer receives the agreed compute; the provider manages how to source it.
That flexibility should be designed into the contract. A claim tied exclusively to a named project is a different product from an entitlement that permits delivery across qualifying providers. Where substitution is allowed, the specification and acceptance process turn equivalence into something the buyer can rely on.
Liquidity is a business to build
A compute market creates work for participants beyond the buyer who runs a workload and the provider who operates a cluster. Specialist trading firms and funds could quote buy and sell prices, carry inventory, and assemble capacity across delivery windows. Their opportunity is to earn a spread for making it easier for others to transact.
There is also room for arbitrage. If equivalent, transferable claims trade at different prices, a participant could buy the cheaper claim and sell into the higher-priced market when the difference covers execution, transfer, and delivery costs. That activity can bring prices closer together. Clear contract standards expand the set of positions that participants can compare and trade.
Let demand respond to price
Prices can also change where and when compute gets used. Consider a US buyer with a latency-tolerant workload. If qualifying capacity in Asia is cheaper during the hours the buyer needs it, and the savings outweigh data-transfer and operating costs, the buyer can choose that capacity instead. An off-hours price difference is one possible opportunity, not a fixed daily pattern.
That is useful even without a resale trade. Buyers with flexibility gain another way to lower costs. Providers gain access to demand that can move toward available supply. Location, timing, and performance become choices with visible prices.
Connect today’s capacity to tomorrow’s supply
A market for current capacity gives buyers and providers a way to adjust positions as their needs change. Extending that market forward gives developers a way to offer future supply and buyers a way to commit before they need delivery.
The financing opportunity follows: commitments that are clearly specified and easier to transfer can attract earlier demand, giving capital a stronger basis for funding new capacity. Tensors is building toward this connection between compute, contracts, and capital. The ambition is a market in which capacity can find its next buyer, commitments can support new infrastructure, and every delivery obligation has a defined path to use.