Nvidia Recruits Six Wall Street Giants to Mobilize $500 Billion for AI Factories

Nvidia has enlisted six of the world's largest asset managers to bankroll the next phase of the AI buildout, announcing partnerships designed to mobilize more than $500 billion of third-party capital over time. Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR have each signed on to create independent compute financing platforms.

The structure matters as much as the number. Rather than loading the buildout onto its own balance sheet, Nvidia is positioning its hardware as collateral that outside investors can underwrite directly. The agreements were struck as memorandums of understanding, and financing is expected to flow through special-purpose entities capable of raising tens of billions at a time through private offerings and bonds.

From Chips to Asset Class

Chief executive Jensen Huang framed the shift as a change in how the industry acquires compute. Companies once bought chips and built data centers one project at a time; now, he argued, AI factories can be financed as productive infrastructure, backed by repeatable platforms and long-term institutional capital. In his framing, compute is revenue.

Nvidia's pitch to investors rests on residual value. The company points to its Ampere-based A100, introduced in 2020 and still in commercial use six years later for training, fine-tuning and inference, with customers committing to multi-year deployments that stretch its economic life toward a decade. Because the architecture is standard across every major cloud, Nvidia argues, capacity can be redeployed to another customer or operator when demand shifts.

Rental economics support the case. One-year H100 pricing climbed from roughly $1.70 per GPU-hour in October 2025 to about $2.35 by March 2026, while cross-provider on-demand median rates rose from around $2.00 to $2.70 over a similar window. Blackwell capacity commands a premium, with reported B200 cloud rates ranging from approximately $5.30 to $7.05 per GPU-hour.

Where Nvidia Takes Risk

The company was careful to bound its own exposure. The $500 billion figure represents aggregate capital the platforms are designed to mobilize over time, not Nvidia revenue, a single fund, or a commitment to any one customer. Each financial institution independently underwrites its own deals, assessing the customer, demand, utilization, cash flow and residual value.

In some cases Nvidia may offer a residual-value support mechanism covering up to 25% of an opportunity, evaluated project by project. The company describes that backstop as substantially lower than other compute-financing arrangements, and says it is meant to complement independent underwriting rather than replace it.

Huang reportedly approached only these six firms, and none declined. The announcement follows reports that Nvidia had been in talks to guarantee financing for a quarter-trillion-dollar data center project tied to OpenAI, one of its largest customers.

What Comes Next

For AI labs, enterprises and neocloud providers, the practical question is access. Demand for accelerated compute has consistently outrun the ability of smaller buyers to finance it, and Nvidia is betting that standardized underwriting will close that gap the way project finance once did for power plants and fiber networks.

The open question is whether utilization holds up long enough to justify decade-length assumptions about residual value. If it does, GPU capacity becomes a genuinely investable asset class. If it does not, a very large pool of institutional capital will have arrived at exactly the wrong moment.