Building data centres at the scale that AI training and inference now demand is, at its base, a capital problem. Land, power interconnection, cooling infrastructure, and compute hardware all require committed spending that runs for years before a facility goes live. Apollo, Blackstone, and Goldman Sachs are among the Wall Street groups working with Nvidia to raise up to $500 billion aimed at closing that gap.
The chipmaker's position in the deal
Nvidia is the chipmaker at the centre of this effort. GPU clusters are the primary cost driver inside a modern AI data centre, making hardware procurement inseparable from the financing decision. Operators who cannot secure long-duration capital delay their hardware orders. A financing structure assembled alongside major alternative asset managers and investment banks addresses that constraint before it feeds back into order flow.
Who is building the coalition
Apollo and Blackstone anchor the effort from the private credit and alternative asset management side, instruments suited to the long payback horizons that data centre construction requires. Goldman Sachs brings traditional investment banking depth. The source describes all three as "among" the groups involved, which means the coalition is still forming. How Nvidia fits into the deal structure and how capital would be deployed is not specified.
Reading the $500bn figure
Five hundred billion dollars places this initiative in unusual territory for infrastructure finance. The figure appears to encompass the full data centre build cycle: real estate, power contracts, cooling systems, and the compute racks that carry the highest per-unit costs. Private credit, with its tolerance for illiquidity and multi-year lockups, suits that duration profile better than short-term revolving facilities. How close the effort gets to that ceiling depends on how many additional institutions ultimately join Apollo, Blackstone, and Goldman Sachs.