NVIDIA announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to create independent financing platforms designed to mobilize more than $500 billion in external capital over time to support the construction of AI infrastructure. The company presents the move as a shift from purchasing chips and building data centers on a project-by-project basis to financing “AI factories” as productive infrastructure capable of generating revenue.
According to NVIDIA, AI is moving from the research phase into production, while computing has become a resource directly linked to revenue. The financing platforms target eligible AI companies, laboratories, enterprises and specialized clouds that need extensive computing capacity but do not always have the financing at the scale or cost required for rapid expansion.
A Platform Beyond Chips
NVIDIA defines an “AI factory” as an integrated platform that includes accelerated computing, networking, system software, AI frameworks and the global developer ecosystem. It says NVIDIA DSX factories can run a wide range of models, modalities and algorithms, including language, vision, speech, biology, physical AI and robotics.
The company believes that a single factory can serve multiple customers and workloads, and that its reliance on an architecture used across all major cloud providers, system manufacturers and enterprises worldwide gives it the flexibility to be redeployed to another customer, cloud or operator as needs change. It also says CUDA updates improve the performance and efficiency of installed infrastructure and reduce the total cost of ownership over time.
NVIDIA cites the A100 processor, which it launched in 2020 based on the Ampere architecture, and says it remained in commercial use six years later for model training, fine-tuning, inference and high-performance computing, with customers committed to deploying its capabilities for multiple years—potentially extending its economic life to about a decade.
Pricing Indicators and the Role of Investors
The company says the price of renting an H100 for one year rose from approximately $1.70 per GPU operating hour in October 2025 to approximately $2.35 in March 2026. The median on-demand price average across service providers also rose from nearly $2 per GPU hour in October 2025 to $2.70 in June 2026, while advertised cloud prices for B200s ranged from approximately $5.30 to $7.05 per GPU hour.
NVIDIA explains that the $500 billion represents the total external capital the platforms are designed to mobilize over time, not revenue for the company, a single fund or a commitment to a specific customer. Financial institutions will independently evaluate each opportunity, including the customer, demand, utilization rates, cash flows and residual value, while NVIDIA provides the AI factory platform and these investors’ expertise in long-term financing.
Limited Residual-Value Support
In response to questions about the possibility of circular financing, NVIDIA says the initiative relies on independent, long-term institutional capital and that financing decisions will be based on customers’ actual economics. In some cases, the company may provide residual-value support of up to 25% of the financing opportunity, with each case evaluated individually. NVIDIA describes this support as limited and tied to residual value, complementary to—not a substitute for—independent evaluation.
The company links the return on these investments to the use of AI in software development, drug discovery, product design, customer service, process automation and the creation of new services. In its view, more computing leads to better AI, which then increases usage and revenue, supporting additional demand for computing.