$500 Billion for AI Infra: Nvidia Brings Wall Street On Board, Pushes Back Against Circular Financing

Set Trending Topics as a preferred source on Google.

t is one of the largest financing efforts Wall Street has ever assembled for a single technology wave: on Monday, Nvidia signed memorandums of understanding with six of the world’s biggest asset managers and private capital firms. Together with Apollo Global, BlackRock, Blackstone, Brookfield Asset Management, Goldman Sachs and KKR, the chipmaker plans to build independent financing platforms designed to mobilise more than $500 billion in third-party capital for the buildout of AI infrastructure over time.

The agreement is still subject to final terms. According to the Financial Times, which first reported on the talks, the six firms are to create dedicated pools of capital to finance data centre construction “at attractive rates for Nvidia customers”. The intended recipients, Nvidia says, are “leading frontier AI labs, enterprises and AI clouds”.

Markets were not uniformly enthusiastic: Nvidia’s stock initially fell about 1.1 per cent after the FT report – already wiping out nearly $60 billion in market capitalisation – and ended the trading day 2.9 per cent lower. Nvidia is currently worth roughly $5.26 trillion.

What the platforms are meant to do

The basic principle: Nvidia supplies the technology, the financial investors supply the capital and the risk assessment. “The financial institutions will independently assess each opportunity — the customer, demand, utilization, cash flow and residual value,” Nvidia CEO Jensen Huang writes in an accompanying blog post. “NVIDIA provides the AI factory platform. The financial institutions provide long-term capital and financing expertise.”

Huang explicitly stresses that the $500 billion figure is not Nvidia revenue, not a single fund and not a commitment to a single customer, but the aggregate third-party capital the platforms are designed to mobilise over time.

“We began by building chips; today, we are helping create a new class of productive, investable infrastructure: AI factories,” Huang said in the release. Jon Gray, President and COO of Blackstone, commented: “We continue to be enormous investors globally across the Nvidia ecosystem, and this announcement further underscores our confidence in their platform and the future of AI infrastructure.”

Huang’s argument: compute as an infrastructure asset

The core of Huang’s blog post is an attempt to position computing power as a classic infrastructure asset class – comparable to power grids, railways or telecommunications networks, all of which were built with external financing. His lines of argument:

• Fungibility: An “AI factory” can serve many customers and workloads – from language and vision to biology and robotics. If needs change, it can be used by another customer, another cloud or another operator. That protects residual value.
• Software instead of decay: Through CUDA and successive software generations, the performance of already-installed hardware keeps improving, lowering total cost of ownership and extending the asset’s useful economic life.
• Longevity: As evidence, Huang points to the A100, introduced in 2020 and still in commercial use six years later – for training, fine-tuning, inference and HPC.
• Pricing: One-year H100 rental pricing rose from about $1.70 per GPU-hour in October 2025 to roughly $2.35 in March 2026, he writes; the cross-provider on-demand median went from around $2.00 to $2.70 by June 2026. Blackwell capacity commands a premium, with reported B200 cloud rates spanning approximately $5.30 to $7.05 per GPU-hour.

The charge: “circular financing”

What stands out is how directly Huang’s blog post takes on a criticism that has followed Nvidia for months. Under the subheading “Is this circular financing?” he writes: “This initiative is designed to address that concern.”

The background: over the past years Nvidia has repeatedly invested in – or backed the financing of – companies that in turn buy Nvidia chips. Most recently that included a reported $5 billion investment in Ilya Sutskever’s Safe Superintelligence, according to Bloomberg, and before that stakes in OpenAI and xAI, among others. Critics see a loop in which Nvidia helps finance the purchase of its own products, tying revenue and sector concentration risk together. The FT itself accompanied the news with a Lex column titled “Nvidia’s chips may be novel, but its ‘circular financing’ isn’t”.

Huang’s counterargument: demand is real and comes from frontier labs, AI-native startups, enterprises, cloud providers and countries; the capital providers underwrite each project independently. This, he argues, is “the beginning of an open capital market for AI infrastructure”.

Nvidia does not remove itself from the risk entirely, however. According to Huang, the company may in some cases provide a residual-value support mechanism for up to 25 per cent of an opportunity, assessed on a project-by-project basis. That is “substantially lower than other compute-financing arrangements”, he says, and is meant to complement rather than replace independent underwriting. In parallel, the FT reports that Nvidia is in talks to provide a large guarantee for a 10-gigawatt data centre project in Ohio leased to OpenAI.

The capital requirement is enormous

The deal also shows how closely Nvidia is now intertwined with the private capital industry. Firms such as Apollo and Blackstone have recently structured AI infrastructure deals for companies including Anthropic, and pioneered multibillion-dollar, investment-grade-rated financings for companies such as Meta and Intel – structures that sit off the corporate balance sheet.

The need is correspondingly large: Morgan Stanley projects that hyperscalers will spend around $3.5 trillion between 2026 and 2028. Apollo President Jim Zelter spoke of more than $8 trillion in expected investment on an earnings call earlier this month: “The sheer size of the AI infrastructure build-out is unprecedented. We see an enormous opportunity for private capital to finance a portion of this along with public capital.”

Whether this genuinely creates an independent capital market for AI infrastructure, as Huang argues, or merely spreads the concentration risk around a single chipmaker across more balance sheets, will only become clear once the first projects are actually financed and running at capacity.