Nvidia raises $500 billion with Wall Street to fund its AI clients

The American chipmaker has signed memoranda of understanding with six major asset managers to create financing vehicles for the purchase of its own GPUs. This financial structure eases the balance sheets of hyperscalers but raises concerns among credit investors.

Apollo Global Management, Blackstone, BlackRock, Brookfield Asset Management, Goldman Sachs, and KKR: Nvidia has brought together six of Wall Street's biggest names to create what it calls "computing finance platforms." These structures, fueled by institutional funds, insurers, and private lenders, will allow artificial intelligence companies to borrow money to purchase chips, servers, network equipment, buildings, and power supplies. The total funding exceeds $500 billion.

CEO Jensen Huang says he only approached these six groups, and that none refused. Nvidia can guarantee up to a quarter of each deal, which lowers the interest rate paid by its customers while leaving most of the credit risk with the lenders.

The rationale behind the agreement lies in an accounting reclassification. GPUs have long been treated as rapidly depreciating hardware, replaced with each new generation. Nvidia is now asking lenders to consider them as durable infrastructure, comparable to a toll highway or a power plant. "They are now revenue-generating assets," Huang said, describing them as productive, durable, and transferable from one customer to another.

This construction is a response to increasing financial pressure on hyperscalers. Microsoft, Amazon, Alphabet, Meta, and their peers collectively anticipate between $720 billion and $745 billion in capital expenditures in 2026, roughly 77% more than the previous year. Forecasts for 2027 have more than doubled in a year: the consensus estimate jumped from $480 billion in August 2025 to $1.08 trillion this month, according to Bank of America, representing a 127% increase.

Moody's warned that such amounts erode available cash and push tech companies to take on more debt. Alphabet posted negative free cash flow of $5,9 billion in a quarter during which it spent $44,9 billion on its projects. By carrying debt in separate financing vehicles, hyperscalers protect their credit ratings and maintain traditional borrowing capacity.

For smaller players, the effect is even more noticeable. Companies like CoreWeave or Nebius, lacking "investment grade" ratings and constrained by costly financing, thus gain access to conditions previously reserved for industry giants.

The market reaction was more nuanced than the reported figure suggests. Equity investors welcomed the removal of a bottleneck, but the cost of insuring against default on Nvidia's debt rose after the announcement and has roughly doubled since the end of May. The concerns center precisely on the reclassification of the GPUs. "The chips depreciate quickly and lose value as soon as a new generation arrives," warned Nigel Green of the consulting firm deVere Group, emphasizing that lending against these assets only makes sense if the collateral retains its value.

Critics also point to the circularity of the arrangement: Nvidia is helping to finance the purchase of its own products. Goldman Sachs CEO David Solomon called the deal "a pivotal moment in a historic cycle of AI investment." The true significance of this moment depends on a question no one can yet answer: what will the value of a current GPU be in five years?