Nvidia said it is partnering with six major financial institutions — Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR — to mobilize more than $500 billion in third-party capital for AI data-center construction, structuring the financing around Nvidia's own graphics processors as the underlying collateral.
The arrangement, announced by Nvidia, creates independent compute-financing platforms, each run by one of the six partner institutions, that will lend capital to hyperscalers, AI cloud operators and enterprise customers to buy Nvidia hardware and build out data centers. Nvidia has reportedly agreed to help guarantee the residual value of the chips used as collateral, addressing a core concern for lenders: unlike real estate or energy infrastructure, GPUs depreciate quickly as newer chip generations arrive.
In AI, compute is revenue. NVIDIA compute is uniquely suited for this role. It is broadly adopted, flexible across models and workloads, fungible and transferable across customers and operators, and continuously improved through CUDA software — extending its useful life and improving its economics over time.
Jensen Huang, Nvidia chief executive
Part of a Wider Capital Build-Out
The financing push arrives as the largest AI infrastructure spenders — Alphabet, Microsoft, Amazon, Nvidia, Oracle and Meta — have amassed close to $1.5 trillion in combined purchase commitments tied to chips, cloud capacity and power contracts, up sharply from roughly $1 trillion just three months earlier, according to an analysis of hyperscaler filings. Those commitments are typically recorded off-balance-sheet until spending actually occurs, a structure that has drawn scrutiny from analysts questioning how much of the AI capital build-out represents durable demand versus vendor-financed circularity.
Suppliers elsewhere in the chip-manufacturing chain are making similar bets. TSMC, which fabricates Nvidia's most advanced chips, raised its 2026 capital-spending guidance to as much as $64 billion, citing sustained AI-driven demand. Intel, separately, raised roughly $15 billion through a stock offering this month to help fund manufacturing expansion as it lifted its own 2026 capital-spending forecast above $20 billion — a reminder that the AI buildout is straining balance sheets well beyond the hyperscalers writing the checks for compute.
Nvidia has cast the new financing platforms as a way to broaden access to scarce compute rather than concentrate it further, arguing that treating GPU capacity as an investable asset class — akin to toll roads or power plants — will draw in capital that would not otherwise flow to AI infrastructure. Skeptics have countered that a financing structure in which the chip supplier also helps guarantee the value of the collateral backing loans for its own hardware introduces risks that traditional infrastructure lending does not carry. Terms for individual deals under the platforms have not been disclosed, and Nvidia said specific transactions are expected to reach market over the coming months.
