Nvidia Pledges $125B Backstop to Unlock $500B in AI Financing

Nvidia has signed memoranda of understanding with six of Wall Street's largest asset managers to mobilize more than $500 billion in outside capital for AI infrastructure. Apollo Global Management, BlackRock, Blackstone, Brookfield Asset Management, Goldman Sachs, and KKR are the partners named in the plan.

The money is meant to fund data centers, chip fabrication plants, and the power generation needed to run them. Nvidia CEO Jensen Huang framed the effort as a move away from financing AI buildouts project by project, toward standing platforms that can be repeated at scale.

The mechanism hinges on Nvidia's own balance sheet. The company is guaranteeing up to 25 percent of the residual value of GPUs pledged as collateral, meaning that if the chips underperform their expected resale value, Nvidia covers a share of the gap.

Huang has said Nvidia could ultimately be on the hook for as much as $125 billion, a quarter of the overall $500 billion target. Importantly, that figure is not new revenue or a single fund. It is an aggregate goal spread across years and multiple deals, and it is not tied to any one customer.

Turning GPUs into an asset class

The deals are structured as MOUs rather than binding contracts, leaving room for terms to shift before capital actually moves. Goldman Sachs is separately exploring a market for credit instruments backed directly by Nvidia compute capacity.

BlackRock CEO Larry Fink drew a comparison to the birth of mortgage-backed securities in the 1970s, when banks began bundling loans into tradable assets. Blackstone President Jon Gray echoed that framing, likening the valuation approach to how lenders assess housing collateral.

Huang has said he personally approached the financial firms with the idea, according to Goldman Sachs CEO David Solomon.

The plan leans on data suggesting GPUs hold value longer than skeptics assume. Nvidia points to its 2020-era A100 chip, still in commercial use six years on, with contract rates for the newer H100 rising from $1.70 to $2.35 per GPU-hour between October 2025 and March 2026.

Nvidia has already committed billions to customers such as OpenAI and Anthropic, along with neocloud operators CoreWeave, Nebius, Firmus, and Lambda. Those relationships sit against a backdrop in which Big Tech's collective AI spending is projected to exceed $730 billion this year, according to Reuters.

Markets did not universally cheer the news. Nvidia's stock fell roughly 1.4 percent following the announcement, erasing more than $70 billion in market capitalization, a reminder that guaranteeing chip value also concentrates risk back onto the guarantor.