Jensen Huang assembles $500B AI compute financing consortium with Wall Street's biggest names

Jensen Huang Assembles $500B AI Compute Financing Consortium with Wall Street's Biggest Names

Jensen Huang has brought together six of Wall Street's largest capital allocators to finance AI infrastructure at unprecedented scale. David Solomon (Goldman Sachs), John Gray (Blackstone), Jim Zelter (Apollo), Bruce Flatt (Brookfield), and Larry Fink (BlackRock) joined Huang on CNBC to discuss a financing consortium designed to unlock capital for the buildout.

The scale is staggering. Jim Zelter stated that the AI infrastructure buildout is "unprecedented," with more than $8 trillion of capital expected to be invested globally. The consortium aims to tap private capital markets to finance a material portion of that spend alongside public funding.

The profitability angle

Huang emphasized that AI labs and startups generating tokens from their compute are "incredibly profitable." He cited $500 billion in venture capital deployed into AI startups over the last six months—the largest monthly venture investing period in recent history. Those companies need compute, and Huang positioned the consortium as the vehicle to supply it at scale.

The numbers suggest why speed matters. At roughly $50-60 billion per gigawatt of compute capacity, a 10-gigawatt data center cluster costs $500-600 billion. Huang framed this as approximately next year's compute demand: Meta alone has announced a 10-gigawatt plan, and the major labs are collectively running roughly three gigawatts today while the industry continues scaling at 3x annually.

The financing structure

Prakash Manda's analysis surfaces the mechanical insight. Banks have historically avoided GPUs as collateral due to unpredictable depreciation—a new generation can obsolete prior hardware. Huang is solving this by offering depreciation insurance to the banks, covering up to 25% of losses.

The deeper move: Huang is also advising banks on standardized reference designs for data centers that make them fungible and comparable. A "Blackstone data center" or "Blackwell configuration" becomes underwritable as a class rather than a unique project. This fungibility allows the debt to be repackaged into asset-backed securities, CLOs, and CDOs—the same securitization structures used in real estate markets. That enables tranching, investment-grade ratings, and eventual sale to pension funds and insurance firms.

The outcome shifts data center financing from venture equity into debt markets. Banks trade idiosyncratic project-specific credit risk for sector-wide credit risk, bringing AI infrastructure financing closer to the cost of capital available for real estate rather than startups.

The visibility gap

Huang's emphasis on profitability reflects a structural problem: the two companies driving AI demand—OpenAI and Anthropic—are private. Their financials leak sporadically, but Wall Street lacks the transparency it typically demands before deploying capital at this scale. The consortium effectively bridges that gap by anchoring financing to Nvidia's visibility into customer workloads and roadmaps, and to the bankers' ability to model sector-wide token demand rather than individual company risk.