Nvidia officially launches the global “Compute for Revenue Sharing” program! Startups can skip buying GPUs and exchange future profits for computing power.

Nvidia officially announced a new revenue-sharing business model, allowing AI startups to exchange future revenue for GPU computing power, eliminating the need for massive upfront hardware investments. This model is led by Nvidia globally, with Sharon AI and Firmus as the first partners, marking Nvidia's strategic shift from hardware sales to "Compute as a Service."

(Previous report: Nvidia to shift to cloud subscription? First base deploys 170k GPUs with revenue-sharing and credit support) (Background: Nvidia's in-depth analysis: AI is a "five-layer cake"! Trillions of dollars in infrastructure just beginning)

Table of Contents

Toggle

• "Compute as a Service": From selling shovels to sharing gold
• Sharon AI and Firmus lead the way: From Indonesia base to global expansion
• Computing power solutions for AI-native startups
• Strategic significance of global expansion

Nvidia is reshaping the resource allocation rules of the AI industry. On July 6, the GPU giant officially announced a new revenue-sharing business model, allowing AI startups to trade future revenue for GPU computing power without bearing massive upfront hardware costs. This move marks Nvidia's formal transformation from a pure chip seller into a deep participant in AI infrastructure operations.

"Compute as a Service": From selling shovels to sharing gold

In the past, AI startups faced two options to obtain sufficient GPU computing power: either foot the bill themselves, spending millions or even billions of dollars to purchase Nvidia GPU hardware, or sign long-term leases with cloud service providers. For most startups, however, both routes impose heavy financial pressure, especially as demand for the Blackwell architecture's GB300 GPU continues to rise.

Nvidia's new model is essentially a hybrid structure of "Compute as a Service (CaaS)": Nvidia's cloud partners provide GPU computing services to AI startups, while Nvidia not only earns standard hardware sales revenue but also takes a cut of the partners' cloud operating revenue.

This means Nvidia is no longer done after selling chips; instead, it directly ties its interests to the long-term operational performance of the ecosystem. For AI startups in high-growth phases with tight cash flow and urgent computing needs, this model opens a path that doesn't require "raising money first before building."

Sharon AI and Firmus lead the way: From Indonesia base to global expansion

The first partners to adopt Nvidia's business model are Sharon AI and Firmus. Sharon AI is deploying up to 40k Nvidia Grace Blackwell GB300 GPUs, while Firmus is building a DSX AI factory park on Indonesia's Batam Island, planned to expand to 360 megawatts of power capacity and accommodate up to 170k Nvidia GPUs.

Notably, on July 2, the Block reported exclusively on a pilot collaboration between Firmus and Nvidia in Batam, Indonesia, which was a "single-base trial" of the revenue-sharing model. Now, Nvidia has officially established it as a global business strategy, upgrading it from a case-by-case project to an institutionalized product—the clearest signal yet of Nvidia's transformation from a hardware company to an infrastructure ecosystem operator.

Sharon AI co-founder and CEO James Manning said: "The strategic collaboration with Nvidia is a key moment for Sharon AI to realize its sovereign-scale AI computing power vision." Tim Rosenfield, co-CEO of Firmus Technologies, noted: "AI-native companies need scalable, energy-efficient, and cost-effective computing infrastructure to compete globally."

Computing power solutions for AI-native startups

Baseten, Fireworks AI, Together AI, and other AI-native companies are potential beneficiaries of this model. These startups need immediate access to AI cloud computing power for model training, post-training fine-tuning, and high-concurrency inference services, but their products are often in the transition phase from pilot to production, with business models not yet fully formed.

Under traditional hardware procurement or long-term lease frameworks, these startups are either crushed by capital expenditures or miss market windows due to insufficient computing power. Nvidia's revenue-sharing model effectively lowers the barrier for startups to access advanced computing power, making resource allocation dependent on product potential rather than existing balance sheets.

In its official blog, Nvidia stated: "AI infrastructure needs are shifting from model development to production inference. AI factories need to run continuously, generating tokens at scale. The new model allows AI clouds to integrate Nvidia platforms faster, accelerating growth in the AI-native space."

Strategic significance of global expansion

From Batam, Indonesia, to locations worldwide, Nvidia is pushing this revenue-sharing model to all regional AI participants, including startups, model developers, enterprises, research institutions, and regional AI operators. Behind this is Nvidia's deep understanding of the AI computing market structure: a few major cloud providers (AWS, Azure, GCP) control most GPU supply, but a large long-tail demand comes from startups and regional players, who cannot access computing power under traditional financing structures.

Through a combination of revenue-sharing and credit support, Nvidia not only creates a more stable recurring revenue stream but also extends its ecosystem from top-tier major cloud providers down to the long-tail market—a classic evolution of a platform business model.