Nvidia guarantees chip resale value to unlock $500B AI infrastructure financing

Nvidia has signed letters of intent with six major financial institutions—Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR—to mobilize over $500 billion in third-party capital for AI infrastructure projects including data centers, chip factories, and power plants. The announcement triggered a 1.4 percent drop in Nvidia's stock, erasing more than $70 billion in market cap.

CEO Jensen Huang framed the partnership as a shift from one-off project financing to repeatable financing platforms, arguing that AI infrastructure should be fundable as productive infrastructure similar to power grids or transportation networks. Huang emphasized that the $500 billion is an aggregate target spread over years, not Nvidia revenue, not a single fund, and not a commitment to any one customer. The company did not disclose terms, individual commitments from each partner, or a timeline for deployment.

To unlock the capital, Nvidia is guaranteeing a portion of the residual value—the resale or reuse value—of its own chips. The company will cover part of the gap if that value falls below expectations at the end of a financing term, up to 25 percent of a given transaction, with each project reviewed separately. Huang described this as "significantly lower" than residual-value guarantees in other compute financing arrangements and stressed that the actual credit assessment—evaluation of customer credit quality, demand, utilization, cash flow, and residual value—remains with the capital providers. Nvidia is also negotiating a guarantee for a 10-gigawatt data center in Ohio leased to OpenAI.

Huang's framing directly addresses criticism from investor Michael Burry, who called hyperscaler depreciation practices "one of the more common frauds of the modern era." Burry argued that GPUs become obsolete too fast for the five-to-seven-year useful lives claimed in depreciation schedules, because Nvidia releases new chip generations every two to three years, and that depreciation would be understated by roughly $176 billion between 2026 and 2028 alone. Huang countered that the A100, launched in 2020, is still in commercial use six years later and that economic lifespan can stretch toward a decade, supported by improvements to CUDA (Nvidia's software layer) that enhance installed hardware over time. As market evidence, he cited rising rental prices: H100 annual contracts went from $1.70 per GPU-hour in October 2025 to $2.35 in March 2026, while B200 capacity runs between $5.30 and $7.05 per GPU-hour.

The scale of the underlying financing wave is enormous. Morgan Stanley expects hyperscaler spending of $3.5 trillion between 2026 and 2028, while Apollo president Jim Zelter estimates total investment need at over $8 trillion. The Bank of England flagged the risk in its July Financial Stability Report, warning that the pace is historically unprecedented and that a shock hitting highly leveraged AI companies could ripple through global financing conditions and trigger a credit crunch. Banks and private credit firms, the report noted, have limited visibility into their indirect exposure to AI infrastructure debt.