NVIDIA announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to establish independent financing platforms aimed at mobilizing over $500 billion in third-party capital for AI infrastructure. The move marks a formal step toward treating AI factory compute as an investable asset class, shifting from project-by-project data center construction to financing productive infrastructure at scale.
The announcement, published on NVIDIA's AI Blog and authored by CEO Jensen Huang, frames the initiative as a response to surging demand from frontier AI labs, AI-native startups, enterprises, cloud providers, and countries. The capital is not NVIDIA revenue, a single fund, or a commitment to a single customer. Instead, the financial institutions will independently assess each opportunity, including customer, demand, utilization, cash flow and residual value.
What the Financing Platforms Will Do
The six partners are global alternative investment managers, infrastructure investors, and investment banks. Each will establish its own independent platform, with NVIDIA providing the AI factory platform and the financial institutions providing long-term capital and financing expertise. The structure is designed to address concerns about circular financing, where money flows between related parties without clear economic substance.
NVIDIA may provide residual-value support for up to 25% of an opportunity, on a project-by-project basis. That support is limited, residual-value based, and designed to complement independent underwriting rather than replace it. The company says this is substantially lower than other compute-financing arrangements, though no comparative figures were provided.
The platforms will build capacity around real customer economics, with independent evaluation of each deal. Financial institutions will not rely on NVIDIA's assessment alone. They will examine the underlying demand, the utilization of the hardware, the cash flow generated, and the expected residual value of the equipment over time.
Huang argues that AI has reached an inflection point, moving from research into production. NVIDIA compute is not just a chip but a complete AI factory platform, capable of running a broad range of AI models and workloads. The NVIDIA DSX platform, as described, is flexible and fungible, serving many customers and workloads simultaneously.
The comparison to historical infrastructure is central to the argument. Electricity, transportation, communications, and computing were all financed externally, with capital markets providing the long-term funding needed to build out productive capacity. AI factories, Huang writes, are the infrastructure of the intelligence era.
Evidence of Durable Compute Economics
The article cites market data to demonstrate that NVIDIA compute retains value over time. The NVIDIA A100, introduced in 2020, remains in active commercial use for AI training, fine-tuning, inference, and high-performance computing. Customers continue to commit capacity for multi-year deployments of A100, extending its economic life toward a decade.
That longevity is supported by CUDA, NVIDIA's parallel computing platform and programming model. CUDA improves performance and total cost of ownership over time, extending the useful economic value of the hardware. The combination of hardware and software makes AI factories more productive over their lifespan, rather than less.
Rental pricing data reinforces the point. One-year H100 rental pricing rose from about $1.70 per GPU-hour in October 2025 to about $2.35 per GPU-hour in March 2026. Cross-provider on-demand median pricing for H100 rose from roughly $2.00 per GPU-hour in October 2025 to $2.70 in June 2026. These increases suggest that demand for AI compute is not only real but intensifying.
Reported B200 cloud rates span approximately $5.30 to $7.05 per GPU-hour. The B200, based on the Blackwell architecture, is used in cloud services and commands premium pricing. The cloud rates are cited as evidence of market demand for the latest generation of AI hardware.
The return on investment is in the usefulness of AI, driving a virtuous cycle. As AI models become more capable and more widely deployed, demand for compute grows, which justifies further investment. The financing initiative is designed to unlock a large pool of independent capital while maintaining disciplined risk exposure.
Addressing the Circular Financing Question
The financing initiative is designed to address concerns about circular financing, a topic that has drawn scrutiny as AI infrastructure spending has grown. In a circular arrangement, a chipmaker might lend money to a customer, who then uses that money to buy chips from the chipmaker, creating the appearance of demand without independent economic validation.
The new structure avoids that by putting independent financial institutions in charge of underwriting. Each opportunity will be assessed on its own merits, with the financial institutions making their own judgments about whether the deal makes sense. NVIDIA's role is to provide the platform and, in some cases, residual-value support, but not to finance the deals itself.
The $500 billion figure represents aggregate third-party capital across the financing platforms over time. It is not a single fund, not a commitment to a single customer, and not NVIDIA revenue. The capital will be deployed as opportunities arise, with each platform making its own investment decisions.
Huang writes that demand for AI infrastructure is real, coming from frontier AI labs, AI-native startups, enterprises, cloud providers, and countries. The rental price increases for H100 and the premium pricing for B200 are presented as evidence that the market is absorbing capacity and willing to pay more for it.
The article argues that NVIDIA compute has characteristics of an investable infrastructure asset: revenue production, broad market, performance improvement over time, and redeployability. These characteristics make AI factories suitable for long-term financing, similar to how power plants, toll roads, and data centers have been financed historically.
The Market for AI Compute Is Real
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
The initiative is framed as the beginning of an open capital market for AI infrastructure. As the platforms mature and demonstrate their viability, more capital may flow into the sector, further expanding the pool of funding available for AI compute buildout.
The partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR bring together some of the largest pools of long-term capital in the world. These firms have deep experience in infrastructure investing, and their involvement lends credibility to the idea that AI factories can be financed like other productive infrastructure.
The shift from project-by-project data center building to financing AI factories as productive infrastructure is significant. It suggests that the AI compute market is maturing, with capital markets beginning to treat AI hardware as a long-lived, income-generating asset rather than a short-term technology bet.
The residual-value support of up to 25% of an opportunity is designed to complement independent underwriting, not replace it. NVIDIA is taking on some risk, but the bulk of the risk sits with the financial institutions, who will make their own judgments about each deal.
The article is dated August 11, 2026, and the pricing data spans from October 2025 through June 2026. The trend is clear: prices for AI compute are rising, and the market is willing to pay more for access to NVIDIA's hardware.
The financing initiative is a bet that this trend will continue. If AI compute remains in high demand and retains its value over time, the platforms will generate strong returns for their investors. If not, the residual-value support will limit NVIDIA's exposure, but the financial institutions will bear the brunt of any downturn.
Huang's article is both a market update and a pitch. It presents the data on rental prices and cloud rates as evidence that AI compute is a sound investment, and it positions the financing platforms as the mechanism to capture that value. Whether the market agrees will become clearer as the platforms begin deploying capital.
The $500 billion target is ambitious, but the partners involved have the scale and expertise to pursue it. Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR collectively manage trillions of dollars in assets. Their willingness to participate signals confidence in the long-term economics of AI infrastructure.
What Comes Next
NVIDIA GTC Berlin is scheduled for October 20-22, likely in 2026, where the company will likely provide more details on the financing initiative and its AI factory platform. The article also links to a related news item, "Why Scaling AI Compute Performance Requires a New Power Architecture," also dated August 11, 2026.
The A100 example is telling. Introduced in 2020, the GPU is still in active commercial use six years later. That kind of longevity is rare in technology, and it is a key reason why NVIDIA compute can be financed like infrastructure. The hardware does not become obsolete quickly; it remains productive for years.
CUDA is central to that longevity. By improving performance and total cost of ownership over time, CUDA extends the useful economic value of the hardware. The software layer makes the hardware more valuable over its lifespan, rather than less.
The rental price data supports the argument. One-year H100 rental pricing rose from $1.70 per GPU-hour in October 2025 to $2.35 per GPU-hour in March 2026, a 38% increase in five months. Cross-provider on-demand median pricing rose from $2.00 to $2.70 per GPU-hour over a longer period, from October 2025 to June 2026.
The B200 cloud rates of $5.30 to $7.05 per GPU-hour show that the latest generation of hardware commands a significant premium. That premium is evidence of market demand for the most advanced AI compute available.
The financing initiative is a response to that demand. By creating independent platforms with access to over $500 billion in third-party capital, NVIDIA and its partners are positioning themselves to fund the next wave of AI infrastructure buildout.
The article frames this as the beginning of an open capital market for AI infrastructure. If successful, it could transform how AI compute is funded, moving from corporate balance sheets to institutional capital markets.
The partnerships are announced, the capital target is set, and the market data is presented. The next step is execution. The platforms will need to find deals that meet their underwriting standards, deploy capital efficiently, and generate returns for their investors.
Huang's article is confident, but it is also careful. It acknowledges the concerns about circular financing and addresses them directly. It presents the data on pricing and longevity as evidence of durability. It positions the residual-value support as limited and complementary.
The article is dated August 11, 2026, and it points forward to NVIDIA GTC Berlin on October 20-22. The conference will be an opportunity for NVIDIA to provide more details on the financing initiative and its broader AI strategy.
For now, the announcement stands as a significant milestone. Six of the world's largest financial institutions are partnering with NVIDIA to finance AI infrastructure at scale. The target is over $500 billion in third-party capital. The goal is to make AI factories a permanent part of the global infrastructure landscape.