Nvidia is starting to take a share of cloud providers' revenue
- Core Viewpoint: Nvidia is transitioning from a chip seller to the "central bank" of the AI computing ecosystem. By providing financial guarantees for new cloud service providers' GPU capacity leasebacks, it secures a share of their revenue, thereby lowering customer financing barriers and expanding market control.
- Key Elements:
- Nvidia launched the "AI Computing Partnership Program," committing to lease back unsold GPU capacity from new cloud providers at agreed-upon prices, using this as a form of credit endorsement.
- In exchange, Nvidia receives a gradually decreasing revenue share from these cloud providers, creating recurring income tied to usage volume.
- The first participants include cloud service provider Sharon AI (planning to deploy 40,000 GB300 GPUs) and Firmus (building a 360-megawatt AI facility in Indonesia).
- This model aims to solve the core problem of emerging AI companies struggling to secure GPU financing due to low credit ratings, accelerating their infrastructure development.
- This move is a strategic continuation of Nvidia's effort to reduce dependence on major clients like Amazon and Microsoft, who are developing competitive AI chips in-house.
- Nvidia has already invested tens of billions of dollars in companies like CoreWeave through equity investments and capacity leasebacks. For example, it once committed $6.3 billion to purchase CoreWeave's unsold capacity.
- Nvidia added another $3.5 billion to guarantee client data center leases in exchange for stock purchase rights, building a multi-layered interest-binding mechanism.
Original author: Dong Jing
Original source: Wall Street Insights
NVIDIA is leveraging its strong balance sheet as a market tool, providing financial guarantees to emerging cloud service providers in exchange for revenue sharing, quietly evolving from a chip seller into the "central bank" of the AI computing ecosystem.
On July 1, according to tech media outlet The Information, NVIDIA is offering financial backstop commitments to young cloud service providers that lease its GPUs — if these companies fail to find enough AI developers to rent computing power, NVIDIA will buy back their unsold GPU capacity at an agreed-upon price.
In exchange, NVIDIA will take a percentage cut of these cloud service providers' revenue, with the share gradually decreasing over the contract term. GPU cloud service providers Firmus and Sharon AI have already joined the program, and three other executives with business dealings with NVIDIA have confirmed the arrangements.
On July 1, NVIDIA announced on its official website a new business model combining revenue sharing with credit support, allowing AI cloud providers to purchase NVIDIA infrastructure without fully bearing the upfront capital expenditure, and to provide computing services to downstream AI-native companies, model developers, and enterprise clients.
According to the report, this program is internally referred to by some at NVIDIA as the "AI Compute Partnership." An NVIDIA spokesperson also confirmed the existence of the initiative. This marks a significant strategic shift for NVIDIA:
On one hand, it expands the customer base by lowering the financing threshold for emerging cloud service providers; on the other hand, it directly participates in the profit distribution of the downstream computing market through revenue sharing, extending its control further down the AI industry chain.
Model Shift: From Selling Chips to Sharing Cloud Revenue
According to NVIDIA's official press release, beyond standard product revenue, NVIDIA will additionally share in cloud service revenue, creating a recurring income stream tied to usage volume. The core intent of this model is to break down the financing barriers that have historically restricted startup AI companies from accessing large-scale computing power.
NVIDIA positions this framework as the "DSX AI Factory" model, targeting AI service scenarios requiring cross-regional continuous operation, high utilization rates, and multi-tenant accelerated computing.
Sharon AI and Firmus are the first cloud providers to participate in this model. Sharon AI plans to deploy up to 40,000 NVIDIA Grace Blackwell GB300 GPUs; Firmus is building a DSX AI Factory campus on Batam Island, Indonesia, expected to scale to 360 megawatts and house up to 170,000 NVIDIA GPUs. These two deployments directly demonstrate NVIDIA's latest progress in translating computing demand into fundable, deployable infrastructure.
NVIDIA notes that emerging AI companies have historically faced severe limitations in accessing capital-intensive infrastructure — even signing long-term commitment contracts is often insufficient to secure financing for computing purchases. This means a large number of AI-native companies, model developers, and inference service providers must endure long waits when scaling their computing capabilities: site selection, power procurement, construction, hardware commissioning — every step can take months or even longer.
The promise of the new model is: by realigning economic structures, these groups can gain faster access to full-stack accelerated computing power without waiting for traditional infrastructure construction cycles to complete.
The Backstop Logic: Solving the Core Challenge of GPU Financing
According to reports, GPUs are typically the most costly component in AI data centers. For chip buyers with lower credit ratings, securing sufficient loans is a major hurdle in itself.
One data center executive commented that such NVIDIA deals "kill two birds with one stone." He explained that if NVIDIA only guarantees the lease of the data center facility, "you still face the problem of 'how to finance the GPUs.'" But if NVIDIA commits to paying for unsold computing capacity within the facility, "the GPU financing problem is solved, and the data center's financing problem is also solved."
In other words, NVIDIA's backstop commitment effectively serves as a credit enhancement tool, enabling emerging cloud service providers that previously struggled to obtain bank loans to leverage larger amounts of capital and accelerate data center construction.
Strategic Intent: Breaking the Monopoly of Major Clients
NVIDIA's introduction of these initiatives has a clear strategic context. Currently, a few large cloud service providers like Amazon, Microsoft, SpaceX, Oracle, Meta, and Google purchase the majority of NVIDIA's chip capacity. However, several of these companies are developing their own competing AI chips, posing a potential threat to NVIDIA.
To reduce reliance on these giant clients, NVIDIA has spent years supporting a group of emerging GPU cloud service providers, led by CoreWeave. The "AI Compute Partnership" program represents a continuation and deepening of this strategy.
According to a previous report from The Information, NVIDIA has also recently been in negotiations to provide financial guarantees for OpenAI to lease a large data center in Ohio. If fully built out at current chip, labor, power, and other material costs, the data center could cost up to $500 billion.
Capital Investment: From Equity Investment to Capacity Guarantees
NVIDIA's financial commitment in this direction is already substantial.
To date, NVIDIA has invested billions of dollars in several emerging cloud service providers in exchange for equity, and in some cases, has agreed to lease back chips from these companies, involving firms like CoreWeave and Lambda. The total value of related transactions has reached tens of billions of dollars. According to a previous report from The Information, NVIDIA's own researchers have also used GPU servers leased back from Lambda.
Regarding capacity guarantees, NVIDIA began pushing forward with related transactions last fall. In September 2024, NVIDIA committed to buying all unsold capacity from CoreWeave through 2032 if the company couldn't find tenants, with the contract then valued at $6.3 billion. This move effectively alleviated investor concerns about CoreWeave's high-leverage business model, driving its stock price up nearly 30% within the following week.
According to a regulatory filing by NVIDIA in May (covering the quarter through April), NVIDIA subsequently added another $3.5 billion to guarantee client data center leases in exchange for the right to purchase their stock.
Overall, NVIDIA is building a multi-layered interest-binding mechanism: equity investment, capacity leasebacks, lease guarantees, and now revenue sharing. Each layer deepens the financial ties between NVIDIA and downstream cloud service providers, allowing NVIDIA to directly share in the incremental revenue from AI computing commercialization, beyond just chip sales.
