Nvidia is teaming up with six of the world's largest asset managers to mobilize more than $500 billion in financing for the data centers, chips and buildings that underpin the AI boom, a deal that treats AI infrastructure as a new asset class alongside stocks, bonds and real estate.
Under the arrangement, Apollo Global Management, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR have agreed to help Nvidia's customers borrow against future revenue from AI infrastructure projects, rather than requiring cloud companies and AI labs to fund construction entirely off their own balance sheets. The financing is meant to underwrite the data-center campuses, server deployments and chip-manufacturing capacity needed to keep pace with demand for Nvidia's graphics processors, according to SiliconANGLE and other outlets that reviewed the announcement.
"Compute is revenue"
Nvidia framed the deal as a shift in how Wall Street views computing hardware. "In AI, compute is revenue," Chief Executive Jensen Huang said in announcing the initiative, arguing that data centers should be treated as productive, long-lived infrastructure "like electricity, like the internet." KKR's co-chief executives described the challenge ahead in blunter terms, saying that in digital infrastructure, "delivery, not ambition, is the hard part."
The arrangement gives major AI buyers — Google, Microsoft, Meta, Amazon, OpenAI and Anthropic among them — another route to finance the buildout without straining their own balance sheets, at a moment when JPMorgan Chase projects technology-related bond sales will exceed half a trillion dollars this year alone, with hyperscaler capital spending on AI on pace to approach $700 billion.
The scale of the commitment has revived concerns about circularity in AI dealmaking, in which chipmakers, cloud providers and their investors are increasingly financing one another's growth in overlapping deals. Critics have warned that such arrangements can obscure how much underlying demand for AI computing is real versus manufactured by the financing itself, and that a slowdown in any one link of the chain could ripple through the others.
Nvidia has not disclosed which specific projects will draw on the new financing pool first, and the six partner firms have described the agreement as preliminary. The deal is likely to face questions from investors about underwriting standards and asset valuations as the first loans are structured in the coming months, particularly given how quickly demand assumptions for AI computing have shifted over the past year.
