Where Has Nvidia, the “Central Bank of AI,” Put Its Money?
Through publicly traded equity holdings, investments in frontier-model companies, venture capital, and data-center commitments, Nvidia has built a portfolio worth more than $100 billion while securing demand for purchases of its own GPUs. As of September 30, its shares in eight publicly listed companies had a market value exceeding $48 billion. This year, the company participated in 129 venture-capital and startup-accelerator deals and disclosed commitments totaling $518 billion, including long-term supply agreements and data-center leases. Following Jensen Huang’s departure, the market has begun to question whether these investments and guarantees can be sustained and fulfilled.
Why read this
AI score 78
The article analyzes Nvidia’s circular financing structure using three categories of data—equity holdings, venture capital, and off-balance-sheet commitments—revealing the relationship between chip demand and capital investment.
Topics
Body · AI translation
Nvidia, the “Central Bank of the AI World”: Where Has All the Money Gone? Download the Cailian Press app for an exclusive first look. November 20, 2026, 11:43:03, from Beijing. Updated November 20, 2026, 13:47:34
In recent years, Nvidia has been hailed by industry insiders as the “central bank” of the AI circular-financing economy. Under former CEO Jensen Huang, Nvidia not only built the core foundation underpinning the global AI boom—GPU chips—but also used its formidable financial resources to amass a vast “capital arsenal.”
However, following the abrupt departure of the “man in the leather jacket,” Wall Street and Silicon Valley are deeply concerned: Will this seemingly inexhaustible flow of funding continue? If the tap is suddenly turned off, what chain reactions and collateral damage will follow?
According to industry estimates, Nvidia’s investment portfolio exceeds $100 billion, ranging from equity investments in frontier laboratories to bets on power producers and cloud-computing companies. This series of capital maneuvers has placed Nvidia at the very center of the AI financial landscape—and while injecting capital, it has also secured a steady stream of purchase demand for its own computing chips.
Nvidia interim CEO Colette Kress said during Tuesday’s earnings call that Nvidia was proud of the role it had played in supporting the industry’s development and would continue to rigorously evaluate future deals. She added that the company “still has a great deal of work to do to determine whether these are the right types of investments for us to make.”
A Vast Public-Market Footprint
Regulatory filings show that as of September 30, Nvidia’s holdings in eight publicly listed companies were worth more than $48 billion, with its stakes concentrated mainly in partners that purchase Nvidia products such as GPUs.
This practice has indeed drawn criticism from some investors and analysts, who call it “circular financing” and worry that weakness at one company could trigger ripple effects across the entire market.
Take cloud-computing unicorn CoreWeave, for example: Nvidia now owns a 10% stake in the company. CoreWeave buys Nvidia chips in bulk and then rents or sells the resulting computing capacity to its data-center customers.
Nvidia’s capital reach also extends to fellow cloud-computing platform Nebius Group, semiconductor stalwart Intel, and xAI before its merger into SpaceX.
Notably, amid Nvidia’s own transition, investors have recently begun rushing for the exits in related stocks.
Since Nvidia’s share price peaked at a record high on October 6, the eight listed companies in which it holds stakes have lost a combined $820 billion in market value during the sell-off, with SpaceX and Intel accounting for most of the decline. Meanwhile, sharply rising borrowing costs have also left many AI companies struggling.
Betting on the Frontier-Model Camp
During Jensen Huang’s tenure, Nvidia also placed enormous bets on leading frontier-model companies such as Anthropic and OpenAI.
Although these companies are engines of innovation for the broader economy, they do not have investment-grade ratings. To train models and support applications built on top of them, frontier laboratories have a seemingly bottomless appetite for computing power—and that has translated directly into a flood of lucrative orders for Nvidia’s high-end chips.
Take Anthropic, for example. As part of Anthropic’s initial public offering on Thursday, Nvidia committed to purchasing $10 billion worth of Anthropic shares. In response, media reports say, Anthropic signed a $35 billion agreement with Lambda, another Nvidia-backed company, to lease data-center computing capacity powered by Nvidia chips.
This year, Anthropic also reached an agreement with Nscale, another Nvidia-backed cloud-service provider, to obtain additional computing capacity.
A Venture-Capital Spree
Beyond making major bets on Wall Street’s AI darlings, Nvidia is also one of the world’s most active investors in startups racing to drive the next wave of growth.
According to PitchBook data, as of November 13, the company had participated in 129 venture-capital and startup-accelerator deals this year. In this year alone, venture investments involving Nvidia and other investors totaled $220 billion, accounting for 27% of all such investment worldwide.
PitchBook estimates that Nvidia has completed at least 293 investment and acquisition deals since the AI boom shifted into high gear in 2023.
Ballooning Future Commitments
In addition, Nvidia’s total off-balance-sheet financial commitments for the coming years are even larger.
Nvidia reported Tuesday that its commitments involving long-term supply agreements, data-center leases and other arrangements totaled $518 billion, up sharply from approximately $85 billion a year earlier.
One such guarantee: Nvidia will provide up to $105 billion in guarantees for data centers that SoftBank-owned SB Energy plans to build in Ohio, helping it issue bonds to cover construction costs, while also making a $1.5 billion equity investment.
OpenAI has agreed to lease computing capacity from the data centers once they are completed, while Nvidia holds exclusive rights to supply chips for the first half of the project.
Of course, there is now a major question mark over whether many of these commitments and guarantees will actually materialize.
For example, in August this year, Jensen Huang spearheaded efforts to secure commitments from Apollo, Blackstone, BlackRock, Brookfield, Goldman Sachs, KKR and other firms to raise a staggering $500 billion, with Nvidia providing backstop arrangements and guarantees. The enormous fund was originally intended to help Nvidia’s startup customers finance purchases of the chips they needed. However, people familiar with the matter said that no capital has actually been deployed to date.
Now, following Jensen Huang’s ouster, investors are increasingly doubtful that the plan can achieve its ambitious goals.
“Special Notice: The content above (including any videos, images or audio) was uploaded and published by users of Dafeng Hao, a self-media platform operated by Phoenix New Media. This platform merely provides information-storage services.
Notice: The content above (including any videos, pictures or audio, if any) was uploaded and posted by a user of Dafeng Hao, a social-media platform that merely provides information-storage services.”
AI Trends