Goldman Sachs, BlackRock, Blackstone, KKR, Apollo and Brookfield have signed memorandums of understanding to lend $500 billion for AI infrastructure construction, according to an announcement made Monday. Nvidia CEO Jensen Huang outlined the concept in a CNBC interview, framing it as a coordinated push to build the datacenter capacity required for artificial intelligence deployment at scale.
The $500 billion commitment represents one of the largest coordinated financing efforts in technology infrastructure to date. The six firms control combined assets exceeding $10 trillion, giving them capacity to scale the initial tranche substantially if construction milestones are met. The memorandums are non-binding at this stage, meaning the actual capital deployment depends on detailed project agreements that have not yet been finalized.
Huang told CNBC that the concept emerged from conversations with Wall Street leadership about the infrastructure gap constraining AI development. Current datacenter capacity for training and inference workloads lags demand by an estimated margin that major cloud providers and chip makers say requires 18 to 36 months of sustained construction across multiple geographies. The $500 billion figure targets the initial build phase, with parties indicating additional tranches are possible.
The financing structure uses project-level lending rather than direct equity stakes in Nvidia or its customers. Each of the six firms brings separate expertise: Blackstone and Brookfield have existing real estate and infrastructure operations; Goldman and Apollo manage large credit portfolios; BlackRock and KKR operate in both debt and equity. The arrangement allows them to deploy capital across multiple projects rather than betting on a single operator or geography.
Regulatory approval for datacenter siting remains a variable. Several U.S. states and European jurisdictions have begun expediting permits for AI infrastructure, while others have imposed new restrictions on power consumption and water usage. The six firms said they intend to structure deals to comply with these requirements, though no specific project locations have been named.
The announcement arrives as Nvidia's valuation has approached $3 trillion, making it the world's most valuable company by market cap. The company's gross margins exceed 70 percent on its flagship H100 and newer Blackwell processors, which remain supply-constrained. Huang's strategy appears to shift some capital risk onto Wall Street while securing a pipeline of orders for Nvidia hardware.
The six firms are betting that AI infrastructure will sustain double-digit returns for the next 10 years. If the initial $500 billion finances projects that reach cash flow positive status within five to seven years, the group is positioned to recycle capital into further rounds. The rollout of competing chips from AMD, Intel and others has not dented demand for Nvidia's products, and capacity constraints persist.
The document to watch is the first detailed project agreement naming specific datacenter locations, power sources and hardware requirements, which will determine whether these memorandums convert to binding capital calls within the next 12 months.
