NVIDIA is moving beyond its traditional role as a supplier of AI chips by introducing a new financing and revenue sharing model for cloud providers. Under the arrangement, NVIDIA can support partners with credit for major GPU purchases while also receiving a share of the cloud revenue generated by the hardware after deployment.
The approach could help smaller AI cloud companies build large data center capacity without carrying the full financial burden alone. Instead of simply selling GPUs and stepping back, NVIDIA would remain connected to the performance of the AI infrastructure over time.
That gives NVIDIA a new recurring revenue opportunity while helping partners secure the expensive hardware needed to run AI training, fine tuning, and inference workloads.
NVIDIA is tying GPU sales to long term cloud revenue
In the traditional model, a cloud provider buys GPUs, builds data centers, rents the computing power to customers, and keeps the revenue from that service. NVIDIA earns money from the initial sale and may continue selling more hardware later.
The new structure adds another layer. NVIDIA can help finance the deployment and collect a share of the revenue generated by AI cloud services running on its GPUs.
| Model area | Traditional GPU sales | New NVIDIA model |
|---|---|---|
| Hardware purchase | Customer pays upfront | NVIDIA can provide credit support |
| NVIDIA revenue | Mainly from GPU sale | GPU sale plus ongoing cloud revenue share |
| Cloud provider risk | Carries most infrastructure cost | Shares more financial risk with NVIDIA |
| NVIDIA involvement | Hardware supplier | Hardware, financing, and infrastructure partner |
| Long term incentive | Sell more GPUs | Keep partner AI clouds expanding and profitable |
This could be especially useful for companies trying to compete in the AI cloud market without the same financial resources as Microsoft, Amazon, Google, or Meta. Building AI data centers requires huge upfront spending on GPUs, networking, cooling, storage, power systems, and facilities.
AI cloud firms are preparing major GPU deployments
Several companies are already linked to the new model. Sharon AI is reportedly planning deployments involving up to 40,000 Grace Blackwell GB300 systems, while Firmus Technologies is working on an AI factory campus in Batam, Indonesia.
The Firmus project is expected to scale to around 360 megawatts and potentially support as many as 170,000 NVIDIA GPUs. Other AI cloud companies named in connection with the strategy include Baseten, Fireworks AI, and Together AI.
These deployments show why financing has become such an important part of the AI industry. Demand is not limited to training giant foundation models anymore. AI companies also need growing amounts of inference capacity to serve chatbots, automated agents, enterprise tools, image generators, coding systems, and other always available services.
NVIDIA becomes more deeply involved in AI infrastructure
The new model makes NVIDIA more than a company that sells graphics processors. It positions the company as a financial and infrastructure partner that can help customers bring AI capacity online faster.
That could make NVIDIA even harder to replace in the AI ecosystem. Cloud providers may become more dependent on its hardware, financing, software stack, and long term business support.
For NVIDIA, the risk is that cloud providers may struggle to generate enough revenue from AI services. If demand slows or AI infrastructure becomes overbuilt, revenue sharing agreements could become less attractive than straightforward hardware sales.
Still, the strategy reflects how quickly the AI market is changing. NVIDIA is not waiting for cloud providers to raise capital and place orders on their own. It is helping shape the entire system that buys, operates, and monetises AI computing power.
Discussion (0)
Be the first to comment.