Nvidia began fronting startups compute in exchange for a share of the revenue that compute generates — a “usage-linked earnings stream,” in their CFO’s words. Read that again: the largest chipmaker on earth just became a percentage business.
It is not an isolated move. BlackRock’s chairman calls compute futures an emerging asset class. OpenAI traded tokens for equity across an entire accelerator batch. Infrastructure everywhere is sliding from invoices toward outcomes — from selling capacity to holding a position in what the capacity produces.
Why pricing is architecture
A vendor can only price on outcomes it can see. Per-seat pricing is the tell of the old architecture: a platform that watches humans log in meters the only thing visible to it — humans. The moment work is executed by agents instead of people, seat pricing meters precisely the thing that is disappearing.
Infrastructure that can see outcomes can price like a partner. Infrastructure that cannot, meters seats.
An operations platform that runs the transaction stream — orders, payments, inventory — sees outcomes natively. It can align its revenue with the operator’s revenue, growing when the business grows, at zero marginal metering cost. That is not a billing preference; it is a structural advantage of owning the operating layer, and it is unavailable to any vendor who merely watches dashboards get opened.
The tell
Ask what a platform meters and you learn where it actually sits. Seats mean it watches people. Tokens mean it resells models. A share of outcomes means it runs the operation. Choose accordingly.