- Nvidia is launching a revenue-sharing model under its DSX AI factory program that grants token credits to AI developers and startups in exchange for a share of their future sales, connecting them with cloud data center operators and eliminating the need for upfront capital to access large-scale GPU compute resources.
- The model, announced by CFO Colette Kress in a blog post Wednesday, creates a new “usage-linked earnings stream” for Nvidia — meaning the company earns recurring revenue tied to the commercial success of the AI products built on its hardware, rather than just the one-time chip sale to the data center operator.
- For AI startups, the program removes the main bottleneck to compute access: they no longer need to wait through “site selection, power procurement, construction and hardware bring-up” — they can access full-stack accelerated computing through cloud partners like Firmus Technologies immediately, paying Nvidia a portion of revenues as their products generate income.
- The move extends Nvidia’s dominance deeper into the AI value chain: rather than just selling GPUs to hyperscalers and neoclouds, Nvidia is now positioning itself as a financial partner and infrastructure broker for the next wave of AI companies — taking a cut of the AI economy’s future revenue rather than just its current capex spend.
What Happened?
Nvidia CFO Colette Kress announced Wednesday that the company is expanding access to its AI hardware through a revenue-sharing model that gives developers and startups token credits — units of computing capacity — in exchange for a portion of their future revenues. The program operates under the DSX AI factory branding, with Nvidia connecting AI model builders, inference providers, agent platforms, and enterprise AI teams with cloud service providers that host the actual hardware. One early partner is Firmus Technologies. Rather than requiring startups to secure their own data center contracts — a process involving site selection, power agreements, and lengthy construction timelines — the DSX model lets them access GPU compute immediately and pay back through revenue sharing as their products generate income.
Why It Matters?
The revenue-sharing model is a strategic move that extends Nvidia’s financial exposure to AI’s upside well beyond hardware sales. Today, Nvidia earns when a hyperscaler or neocloud buys GPUs — a one-time capex event. Under DSX, Nvidia earns a recurring percentage of revenue from every AI product built on its infrastructure through the program, creating a software-like annuity stream on top of hardware sales. It also expands Nvidia’s total addressable market: the revenue-sharing model lowers the capital barrier to entry for AI startups that couldn’t previously access Nvidia’s most advanced hardware, effectively seeding the next generation of AI companies on Nvidia infrastructure — making them dependent on Nvidia’s ecosystem from inception. This is a direct competitive response to the threat posed by alternative AI chip providers (AMD, Intel, custom silicon from AWS, Google, and Microsoft) that are trying to win early-stage AI companies before they scale.
What’s Next?
The success of DSX will depend on two things: the commercial performance of the AI startups Nvidia seeds through the program (since Nvidia’s revenue share is worth nothing if those startups fail), and the degree to which cloud partners are willing to offer capacity on a deferred-payment basis. The model also raises questions about how Nvidia’s revenue-sharing terms are structured — what percentage, on what revenue base, for how long — details not yet disclosed publicly. For investors, the DSX model represents a potential transformation of Nvidia’s business model from a pure hardware company toward something resembling a platform or infrastructure-as-a-service company with recurring software-like revenue characteristics, a shift that could meaningfully change how the market values the business over the medium term.
Source: Bloomberg
