Nvidia teams up with six Wall Street firms for $500 billion AI data center financing
NVIDIA
OpenAI
Nvidia and six major financial groups plan to mobilize over $500 billion in third-party capital for AI infrastructure such as data centers. To secure financing, Nvidia will guarantee up to 25% of the residual value of its own installed chips. Critics warn of a financial bubble due to rapidly aging hardware, but CEO Jensen Huang argues rising rental prices and long processor lifespans justify the move.
Nvidia has signed letters of intent with six major financial groups—Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR—to mobilize over $500 billion in third-party capital for data centers, chip factories, and power plants. According to the Financial Times, which reported the deal in advance, Nvidia's stock fell about 1.4% afterwards, losing more than $70 billion in market value. CEO Jensen Huang described the move as a transition from individual projects to repeatable financing platforms, stating that 'AI factories' should become financeable as productive infrastructure similar to power or transportation networks. Nvidia will support individual projects with a so-called residual value guarantee: if the resale or reuse value of installed hardware falls below expectations at the end of the financing term, Nvidia covers part of the difference, up to 25% of a transaction, assessed project-wise. This means the chipmaker assumes part of the risk of value loss on its own products. Huang noted this share is 'significantly lower' than in other compute financings, and the actual credit assessment remains with the capital providers. The company regularly supports its partners with debt raising, driving its own sales, and is also negotiating a guarantee for a 10-gigawatt data center in Ohio leased to OpenAI. Huang's argument directly responds to criticism from investor Michael Burry, who called hyperscalers' depreciation practices 'one of the most common frauds of the modern era', claiming GPUs become obsolete too quickly for five-to-seven-year useful lives due to Nvidia's two-to-three-year cycle. Huang countered that the A100, introduced in 2020, is still commercially in use six years later, with its economic lifespan approaching a decade, and CUDA continuously improves installed hardware. He cited rising rental prices as market evidence: H100 annual contracts rose from $1.70 per GPU hour in October 2025 to $2.35 in March 2026, with B200 capacity at $5.30 to $7.05. Morgan Stanley expects hyperscaler spending of $3.5 trillion from 2026 to 2028, while Apollo President Jim Zelter speaks of over $8 trillion investment needs. The Bank of England warned in July in its financial stability report that the pace is historically without precedent and a shock among highly indebted AI companies could affect global financing conditions and lead to a credit crunch.
- Abbreviations
- GPU = Graphics Processing Unit — графический процессор
- CUDA = Compute Unified Device Architecture — архитектура унифицированных вычислений на GPU
Source: The Decoder (DE) —
original
