Nvidia Will Trade Compute Access for Revenue Shares in New Startup Partnership Program

What Nvidia Is Actually Doing

Nvidia announced Thursday that it is entering revenue-sharing agreements with early-stage AI companies. Instead of requiring startups to pay cash upfront for GPU access, Nvidia will extend compute credits and then collect a percentage of the company's future product and cloud revenue in return.

The program targets cloud-based AI firms, model builders, and other enterprises that need high-end compute but don't have the liquidity to buy or lease it outright. Nvidia is positioning itself as the intermediary, connecting startups to infrastructure partners running Nvidia chips and taking a cut of what those startups eventually earn.

The Infrastructure Behind It

Nvidia named two initial Australian companies as the compute providers for the program.

Sharon AI will deploy up to 40,000 Nvidia GPUs. Firmus Technologies is constructing a data center in Batam, Indonesia, expected to scale to 360 megawatts and house up to 170,000 Nvidia GPUs, according to CNBC.

That's a combined ceiling of roughly 210,000 GPUs across two facilities — serious compute capacity, not a pilot program.

Why This Makes Sense for Nvidia

Nvidia is already the dominant supplier of AI chips. This program doesn't change that. What it changes is the financial relationship: instead of a one-time hardware sale, Nvidia gets ongoing revenue participation in the companies its chips help build.

If the startups that go through this program succeed, Nvidia collects revenue it wouldn't have seen from a straight hardware transaction. If they fail, Nvidia's downside is the opportunity cost of compute credits that didn't convert.

The model also addresses a real market problem. GPUs have become a scarce commodity. Developers have compared access to them to access to oil, and according to CNBC, GPU capacity has reportedly been tied to futures contracts as users try to hedge against cost fluctuations and availability crunches. Startups that can't pay for compute simply don't build, or they build on inferior infrastructure. Nvidia's program attempts to close that gap without requiring the startup to raise another round first.

Nvidia's Own Balance Sheet

The announcement comes as Nvidia said earlier in June it was planning to raise debt, with sources telling CNBC the offering could amount to at least $20 billion. The company says proceeds would go toward general corporate purposes, including repayment and refinancing of existing debt.

A company raising $20 billion in debt while simultaneously extending compute credits to startups in exchange for future revenue isn't in financial distress. Nvidia's market position is well established. But it does signal that Nvidia is actively managing its capital structure and expanding the ways it participates in the AI economy beyond straightforward chip sales.

The Broader Trend

Nvidia isn't inventing this model. OpenAI has been doing variations of it for some time. According to CNBC's January reporting, OpenAI has entered deals in which it acquires shares or entertains investments from partners including Amazon and AMD. The compute-for-equity or compute-for-revenue structure has become a standard tool for AI companies navigating the gap between capital requirements and available liquidity.

Nvidia is stepping into that arrangement directly, rather than simply selling chips to infrastructure providers who then rent capacity to startups. Nvidia is now taking a position in the economic upside of the companies its hardware enables.

The Legitimate Concern

There is a reasonable argument that this arrangement concentrates even more leverage with Nvidia. Startups that accept compute credits in exchange for revenue share aren't just customers anymore — they're obligated partners. If Nvidia's chips remain the industry standard, those startups have limited negotiating power at renewal or renegotiation. Critics of Nvidia's market position would argue this program deepens dependency rather than democratizing access.

That concern is worth taking seriously. Nvidia already controls a dominant share of the AI chip market. A program that ties startup revenue streams to Nvidia infrastructure compounds that relationship. Whether regulators eventually view this as an extension of market power is an open question. No investigation has been announced as of July 3, 2026.

What's Still Unknown

Nvidia has not disclosed the specific revenue-share percentages startups would owe under the program, nor has it named any startup participants beyond the two infrastructure providers. The terms that would govern when and how Nvidia collects on these agreements — and what happens to a startup that misses revenue targets — have not been made public.