The chip leader is launching a formal revenue share program alongside Australian AI infrastructure operator Sharon AI (40,000 GPUs) and Singapore headquartered AI data center builder Firmus Technologies (170,000 GPUs in Indonesia), formalizing a shift from chip seller to…
Nvidia, the AI chip leader, will take a share of revenue from fast-growing AI startups in exchange for access to its scarce GPUs. The swap turns hardware sales into a financing instrument, according to a company blog post announcing the program on July 2, 2026 and CNBC reporting on the deal.
The program lets cloud-based AI firms, model builders, and other enterprises share both product revenue and cloud revenue with Nvidia in exchange for credits redeemable for Nvidia GPU time and cloud capacity now. The mechanism is the deal. Nvidia is not just selling chips. It is taking a future claim on the upside of the companies that depend on its hardware.
The first two named compute partners illustrate the scale. Sharon AI, an Australian operator, is lined up to deploy up to 40,000 Nvidia GPUs. Firmus Technologies, headquartered in Singapore, is building a data center in Batam, Indonesia, that is expected to scale to 360 megawatts and house up to 170,000 Nvidia GPUs. Together, those two commitments alone exceed 200,000 GPUs. That scale hints at how Nvidia is converting silicon supply into equity-like claims on the AI buildout.
CNBC framed the move against a backdrop of GPU scarcity so persistent that some users have begun hedging it like a commodity. Coverage has likened GPUs to oil, and at least one trader has reportedly tied chip access to futures contracts. For startups that cannot pay cash upfront, revenue-sharing is becoming a financing workaround, not just a sales tactic.
The deal fits a wider pattern in which AI firms trade equity or revenue for compute rather than paying cash upfront. As operators on Hacker News noted the same day, Nvidia is now formalizing that pattern on its own balance sheet. The structure looks less like a lease, which locks in fixed payments for hardware access, and more like a venture-style claim: Nvidia's payout scales with the startup's success, and shrinks toward zero if the startup flops.
That parallel matters because Nvidia is also raising large amounts of debt to finance its role in the AI buildout. Revenue participation plus balance-sheet leverage underwrites the chip leader's shift from chip seller to what is, in effect, a compute-power broker.
The mechanism has a clear falsifier. If TSMC or Intel foundry expansion, or hyperscaler in-house silicon such as AWS Trainium, loosens Nvidia GPU supply, the scarcity premium that makes revenue-sharing attractive to startups disappears. The deal structure then becomes a footnote rather than a template.
What to watch next: the first startups to sign up, the rev-share percentage Nvidia is asking for, and whether any hyperscaler follows Nvidia's template by taking equity-like claims on its largest customers.