Nvidia pauses AI cloud revenue-share deals

Nvidia has paused parts of the AI Compute Partnership it launched in July 2026, the program through which it helps smaller AI cloud providers finance large GPU purchases, according to a Wall Street Journal report carried by Reuters and Quartz. Read this as a pause of some deals, not a confirmed scrap. Nvidia has since denied freezing its AI cloud commitments outright, so the scope is unsettled.

How the AI Compute Partnership is structured

This is not a loan, and it is not a residual-value guaranty on its own. The AI Compute Partnership bundles three things: credit support so a smaller cloud can finance chips, a backstop to buy back unused GPU capacity, and a revenue share layered on top. Nvidia and each provider set a base hourly rate covering the provider's costs, and Nvidia takes 50% of any revenue earned above that line, per the reporting. In its quarterly filing this week the company disclosed about $36 billion of commitments under the program, on agreements that typically run six years, a figure Nvidia expects to shrink as providers find other customers. That is the mechanism most one-line takes skip: a cost-plus split, not a rental fee.

Where the dispute actually is

The pause is not about the split. Reporting ties it to internal concern about antitrust exposure, specifically how far Nvidia can dictate a customer's business. In the program's early weeks Nvidia told providers they had to vet renters through an approval process, and made clear it wanted capacity spread across many smaller AI firms rather than concentrated with one large buyer, terms that irritated some prospective partners. Tom's Hardware reported Nvidia then denied pausing the initiative after that backlash, while the same report says the leasing-to-approved-customers restriction was part of what partners resisted. Both are being said at once. Whether this is a full stop, a revamp, or a fold into a differently structured program is the open question, and the sources do not close it.

The forward case is still on the record. On Wednesday's earnings call, finance chief Colette Kress told investors the revenue share could bring in billions of dollars over the medium to long term. That is a program Nvidia is reworking under pressure, not one it has abandoned.

The takeaway

If you are tracking how Nvidia manufactures its own demand, do not book the $36 billion as revenue. Book it as a commitment the company is actively renegotiating around a constraint it flagged itself, with the approved-customer clause, not the money, as the part that broke. Watch the next quarterly filing for whether that figure holds, grows, or migrates into a new structure. Until then, treat any deal quoting the old 50%-above-base split as provisional, and price it as terms Nvidia may republish rather than terms it will keep.

For related context, see our coverage of Nvidia's $96.2B quarter and $108B guide, its $105B OpenAI Ohio backstop, and the reported $12.9B move for Hugging Face.