Nvidia Launches Revenue-Share Program Offering AI Startups GPU Access Without Upfront Payment

The model rests on three components. First, Nvidia provides token credit advances to capital-limited AI developers, allowing them to begin training and deploying models immediately without waiting for lengthy infrastructure procurement cycles ^[3].

Second, Nvidia earns a revenue-share royalty — standard upfront hardware revenue from selling GPUs to cloud partners, plus an ongoing percentage of the cloud revenue generated on that supported capacity. The specific revenue-share percentage has not been disclosed ^[1][3].

Third, Nvidia offers financial repurchase guarantees to its cloud partners: if a participating provider cannot fill its GPU compute slots, Nvidia will buy back unsold capacity at predetermined prices, reducing the risk for smaller cloud operators entering the market ^[3].

Sharon AI, the Australian cloud provider, will deploy up to 40,000 Nvidia Grace Blackwell GB300 GPUs under the arrangement. The company focuses on sovereign AI infrastructure, providing compute within specific national jurisdictions ^[2].

Firmus Technologies, co-headquartered in Singapore, is building a large-scale DSX AI Factory campus in Batam, Indonesia. The facility is designed to reach 360 megawatts of power capacity and house up to 170,000 Nvidia GPUs, making it one of the largest planned GPU deployments in Southeast Asia ^[2].

The program addresses a structural tension in the AI industry: Nvidia's most advanced GPUs cost tens of thousands of dollars each, pricing out startups and smaller research labs that lack the capital reserves of hyperscalers like Microsoft, Amazon, and Google. By accepting a slice of future revenue rather than demanding payment in full, Nvidia opens a new customer segment while locking in long-term economic exposure to AI workloads ^[1][4].

The move also diversifies Nvidia's revenue concentration. The company has drawn investor scrutiny over its heavy reliance on a small number of hyperscale cloud customers for the bulk of its data center revenue. By cultivating an ecosystem of independent AI cloud providers and startup customers, Nvidia reduces that concentration risk while expanding its total addressable market ^[4].

For the broader AI startup ecosystem, the program functions as a form of non-dilutive financing. Startups that would otherwise need to raise venture capital specifically to pay for GPU compute can instead preserve equity and pay Nvidia from future revenue — a model more common in music and entertainment than enterprise technology ^[1].

Nvidia shares fell 1.4% to $195 on July 2, in line with a broader tech selloff, giving the company a market capitalization of approximately $4.72 trillion. The stock is up roughly 4.5% year-to-date and 24% over the past 12 months, though it remains about 18% below its 52-week high of $236.54 ^[5].

Analysts have not yet issued formal commentary on the revenue-share program's financial impact. The key question for investors will be whether the recurring revenue stream from usage-linked royalties compensates for the working-capital cost of buyback guarantees and deferred payments — and at what scale the program operates relative to Nvidia's $130 billion-plus annual data center revenue run rate ^[5].