Nvidia Offers GPU Credits to AI Startups for Revenue Share

Nvidia launched a program that gives AI startups token credits redeemable for GPU compute in exchange for a share of future product and cloud revenue. The program targets cloud-based AI firms, model developers and other companies that require large-scale compute.

Qualifying startups will receive token credits they can redeem for computing capacity and will share a portion of product and cloud revenue with Nvidia. The company described the arrangement as providing access to full-stack infrastructure powered by its graphics processing units.

Nvidia named two partners to supply compute capacity. Sharon AI committed to deploy up to 40,000 Nvidia GPUs. Firmus Technologies is developing a data centre in Batam, Indonesia, that the company expects to scale to 360 megawatts and accommodate as many as 170,000 Nvidia GPUs. Nvidia added the partnerships are intended to increase available computing infrastructure for startups.

Access to GPUs has become a central issue for AI companies because the chips are in high demand for training and running large models. Fluctuations in pricing and availability have led to arrangements resembling futures contracts for compute capacity.

Several AI companies have entered revenue-sharing or equity-sharing agreements with chipmakers and cloud providers to secure hardware and capital for growth.

Nvidia announced earlier this month plans to raise at least $20 billion of debt, with proceeds earmarked for general corporate purposes and for repaying and refinancing existing debt. The company’s shares have pulled back recently, struggled to hold above the $200 level and have underperformed several semiconductor peers in 2026. The VanEck Semiconductor ETF rose more than 70% in the first half of 2026, and investor attention has shifted toward memory chips and central processing units.

Memory chip makers have gained as supply constraints met rising demand; Micron has been cited as a beneficiary. Companies focused on CPUs, including Advanced Micro Devices and Intel, have drawn attention amid expectations that next-generation AI systems may require a mix of compute architectures beyond GPUs.

The program is one of several industry approaches that link hardware access to future revenue or equity to help startups manage short-term liquidity and capacity constraints.