NVIDIA Bets $500 Billion on AI Compute Power: Why Jensen Huang Calls It an 'Investable Asset'

NVIDIA (NVDA-US) has signed a memorandum of understanding with six major asset management firms—Apollo, BlackRock (BLK-US), Blackstone (BX-US), Brookfield, Goldman Sachs (GS-US), and KKR (KKR-US)—to form an independent compute financing platform. The initiative plans to mobilize more than $500 billion in long-term capital for AI infrastructure development. Amid market skepticism over potential 'circular financing,' NVIDIA CEO Jensen Huang emphasized that AI compute power has transformed from a rapidly depreciating tech product into a 'investable asset' with long lifecycle, cross-customer transferability, and sustained revenue generation.

In an interview with CNBC, Huang revealed he approached only these six firms—and none declined. In a statement, he said: 'We started by manufacturing chips, but now we're helping build a new kind of infrastructure that creates productivity and is investable: the AI factory.'

Following NVIDIA's announcement of the $500 billion financing platform, market concerns quickly emerged. NVIDIA's stock fell approximately 2.8% on Monday (the 10th), wiping out over $70 billion in market value. The 5-year credit default swap (CDS) surged nearly 6 basis points in a single day, doubling insurance costs since late May and approaching the record highs seen in late July.

Market skepticism centers on the 'circular financing' model, with fears that Wall Street funding AI companies could lead them to use loans to purchase NVIDIA chips—effectively creating chip demand through external leverage rather than organic industry demand.

Under this scenario, if downstream AI companies fail to generate expected revenues and cannot repay massive loans, it could trigger a chain reaction of debt defaults. Some market observers have compared the current situation to the supplier financing chaos that preceded the dot-com bubble burst in 2000.

Huang pushed back, arguing that AI compute is now a new form of infrastructure. He stated that today's AI computing power has become a foundational infrastructure of the new era—akin to electricity, the internet, and highways—and should no longer be evaluated under the old logic of fast-depreciating consumer electronics.

Huang pointed out that while GPUs were once seen as rapidly updated and quickly depreciating, AI factory compute power now possesses three key characteristics: 'long lifecycle, cross-customer transferability, and sustained revenue generation'—making it a standard investable asset.

He cited the A100 chip launched in 2020 as an example, noting it is still widely used commercially around the world, with equipment lifespans extendable up to 10 years.

Meanwhile, market demand for compute power continues to outstrip supply, driving up rental prices for H100 and Blackwell series chips. Industries such as pharmaceuticals, manufacturing, finance, and retail are also steadily increasing their compute needs.

Huang further explained that traditional semiconductor cycles are driven by consumer electronics upgrade cycles (e.g., smartphones, PCs), resulting in clear economic cycles. In contrast, the current wave of AI infrastructure development stems from global digitalization demands across industries, with nations actively building compute infrastructure.

Currently, shortages exist across chips, HBM, power, and data center land, with capacity struggling to meet demand—a long-term trend. According to Huang, AI compute demand is not artificially inflated by capital speculation, so there is no fundamental basis for a sharp near-term demand collapse.

Regarding concerns about excessive leverage, Huang clarified that the $500 billion is not a one-time injection but the total long-term capital the platform can mobilize. He noted that cloud providers and AI labs often require hundreds of billions in investment to expand compute capacity. Relying solely on internal funds would severely constrain R&D and other business expansions.

By leveraging third-party financing from Wall Street, tech firms can share financial burdens and accelerate AI deployment. Huang compared this to how highways and power grids have long relied on external capital, positioning the AI compute financing platform as a continuation of that model.

He emphasized that the $500 billion is not to be deployed all at once but will be allocated gradually in line with customers' compute build-out timelines, avoiding a short-term spike in industry leverage.

The financial giants backing NVIDIA are endorsing the 'assetization of compute.' Blackstone President Gray likened AI compute to mortgage-backed securities, highlighting its stable, long-term return profile and positioning it as a high-quality financing asset.

Goldman Sachs CEO Solomon stated that institutions recognize NVIDIA's leadership position and view the compute credit market as a historic investment opportunity.

BlackRock CEO Fink believes accelerating capital into AI is key for the U.S. to maintain its global AI leadership, even comparing this collaboration model to the financial innovation of mortgage securitization in the 1970s.

The debate around NVIDIA's $500 billion financing plan ultimately hinges on how markets assess long-term AI demand and the sustainability of high-leverage infrastructure expansion.

On one hand, stock price and CDS movements reflect investor risk awareness. On the other, Huang and Wall Street financial leaders are attempting to establish a new valuation framework: AI compute is no longer just rapidly depreciating chips, but infrastructure assets capable of generating long-term cash flows and supporting external financing.