The agreements, structured as memoranda of understanding, were announced following NVIDIA's GTC Taipei conference on May 31, 2026, alongside the company's unveiling of its Data Center Exchange (DSX) platform. The six partners are Apollo Global Management, BlackRock, Blackstone, Brookfield Asset Management, Goldman Sachs, and KKR.
"Compute Is Revenue"
NVIDIA chief executive Jensen Huang has built the financial case around a single idea: AI compute generates measurable, recurring revenue, which means it can be financed the same way any other revenue-producing infrastructure can.
"Compute is revenue," Huang has said in describing the thesis. In NVIDIA's framing, an AI factory is not a cost center. It is an operating asset that produces inference output, charges tenants, and generates cash flows capable of servicing debt.
That argument has been gaining traction among institutional investors searching for yield. The $500 billion figure is a mobilization target, not a committed capital pool, and the MOUs are frameworks rather than binding agreements. But the scale of the institutions involved signals that this financing model is being taken seriously at the highest levels of global asset management.
The Six Partners and Their Roles
Each firm brings a different capability to the structure.
Apollo Global Management, which oversees roughly $650 billion in assets, is expected to focus on private credit and direct lending to AI infrastructure projects. Apollo has been one of the more aggressive players in asset-backed lending and sees GPU-secured financing as a natural extension of that business.
BlackRock, the world's largest asset manager with more than $10 trillion under management, is participating through its AI Infrastructure Partnership, a vehicle it launched to channel institutional capital into data centers and related assets. BlackRock's involvement opens the arrangement to sovereign wealth funds and pension allocators that typically require the credibility of a large, regulated manager before committing capital.
Brookfield Asset Management manages around $900 billion and ranks among the largest owners of renewable energy and infrastructure in the world. Its role is expected to center on powering AI facilities, a critical bottleneck given the energy demands of large GPU clusters.
Goldman Sachs is approaching the arrangement through its investment banking and asset management divisions. Goldman Sachs chief executive David Solomon has described the opportunity as the creation of "a market for credit backed by NVIDIA compute," suggesting the bank sees a structured finance ecosystem forming around GPU collateral, similar to how mortgage-backed securities once created a liquid market around real estate cash flows.
KKR is participating through its Helix Digital Infrastructure platform, a dedicated vehicle for data center and digital infrastructure investment. KKR manages more than $550 billion in assets and has been building out its infrastructure capabilities for several years.
Why GPUs Can Be Financed Like Infrastructure
Traditional infrastructure finance works because assets produce predictable cash flows, have long useful lives, and can be repossessed and redeployed if a borrower defaults. GPUs historically failed most of those tests. They depreciated quickly, were tied to specific software stacks, and were difficult to redeploy across different tenants.
NVIDIA's argument is that the latest generation of AI accelerators is different. Software updates extend the productive life of hardware. Standardized interconnects and operating environments make clusters more fungible across workloads and customers. And demand for inference compute is growing fast enough that utilization rates on well-located AI factories remain high.
The DSX platform is central to that argument. Introduced at GTC Taipei, DSX standardizes how AI factories are built, monitored, and reported on, giving financial underwriters a common framework for assessing GPU cluster performance and value. Without that standardization, each financing deal would require bespoke due diligence. With it, the asset class becomes legible to capital markets in a way it previously was not.
Debt-Funded AI Buildout
This announcement follows a broader pattern of debt financing entering the AI infrastructure sector. Several large cloud and colocation operators have already tapped bond markets and private credit to fund GPU acquisitions, and borrowing costs have fallen as lenders have grown more comfortable with the asset class.
NVIDIA's move formalizes and accelerates that trend by bringing the chip designer itself into the financing ecosystem. By helping create the conditions under which its hardware can be financed, NVIDIA expands the pool of potential buyers and can smooth demand cycles over time.
If the MOUs convert to final agreements at scale, the practical effect would be a meaningful reduction in the cost of capital for AI infrastructure projects. That matters most for smaller operators and emerging-market deployments that currently face higher borrowing costs than the largest hyperscale cloud providers.
What Comes Next
The $500 billion figure will take years to materialize, and the MOUs carry no binding commitment. The next milestones are the conversion of these frameworks into definitive agreements, the structuring of specific financing vehicles, and the deployment of capital into actual projects.
