Much has been said about the shaky financing of the current AI wave: the pool of money is huge, the bets are risky, and U.S. economic growth hinges on the market’s continued expansion. This past month has brought these issues into stark relief, with CoreWeave’s stock plummeting amid its mounting debt burden and fears of an AI bubble.

But the corporation escaping scrutiny in all this is Nvidia, whose monopoly over chip design and vast amounts of capital have allowed it to shape the AI capital market to its own ends, furthering its dominance of AI infrastructure and destabilizing the U.S. economy as it proceeds unchecked.

Nvidia was founded in 1993 as a small startup optimizing semiconductor chips for gamers who needed more computing power to render graphics quickly. But things started to change in the early 2010s with the rapid surge of interest in developing artificial intelligence. AI models required a large amount of specialized computational power, and Nvidia’s graphics processing units were ideally suited to AI use cases.

This marked a turning point for Nvidia. In 2016, co-founder and president Jensen Huang personally gifted an Nvidia supercomputer to OpenAI, recently founded by Elon Musk, Sam Altman and others, virtually guaranteeing that OpenAI’s software would primarily use Nvidia chips. Nvidia also began courting scientists and AI researchers by donating its processing units and supercomputers. By 2023, Nvidia’s stock price had begun a steep upward climb that culminated, by mid-2025, in the company being named the most valuable public company in the world.

Once the AI world started running on Nvidia graphic processing units, Nvidia’s tech stack created lock-in that made it nearly impossible to switch.

The centerpiece of this strategy is the company’s Compute Unified Device Architecture, a proprietary software language optimized to run on the chips and achieve their best performance. Although the chips can be used without this specific software, the performance drop-off is steep, meaning that most AI applications require it.

Nvidia’s few competitors have tried to breach this barrier by designing software that more easily translates this language, but Nvidia maintains a powerful moat. In 2024, Nvidia updated its licensing agreements to disallow using the language on rival chips. The company also helps hyperscalers optimize their Nvidia stack, hiring engineers specifically for this purpose. As Nvidia executives noted during their most recent earnings call, “Customers can still mix and match. However, most companies just don’t have the skills to do that or a desire to do that.”

Nvidia’s strategy has borne fruit; the company enjoys a nearly 90 percent monopoly over the design market for graphics processing units. What’s more, Nvidia manages a 70-75 percent gross margin on the chips that it sells.

The power that Nvidia has amassed through this lock-in and its enormous stock price growth has allowed the corporation to begin functioning as, to use The New York Times’s phrase, “Silicon Valley’s central banker.”It has begun offering complex financing arrangements to partners, investing heavily in infrastructure that supports the use of its chips, such as neoclouds, and evaluating outside companies to acquire. Although Nvidia has long invested in AI startups, there has been a notable uptick in its deal activity in recent months.

By operating as an unregulated banker, Nvidia is shaping the U.S. economy outside the regulatory bounds intended to keep risky bets from collapsing the financial system — and the company has a vested interest in fueling the AI mania.

The more AI models that need to be trained faster, and the more data centers there are, the more Nvidia’s technology stack is being purchased and integrated into the American AI industry. But Nvidia’s hardware is expensive, so the company has begun offering complex financing arrangements to incentivize rapid adoption and compute purchases. These arrangements, of course, come with requirements to purchase Nvidia products and, often, to upgrade to the newest chips at regular intervals.

This is not the only deal structure Nvidia has piloted. The company has increasingly pursued a strategy of directly investing in neocloud infrastructure — data centers that rent cloud space to AI companies — and in turn requiring the neocloud companies to purchase a certain amount of Nvidia hardware and to upgrade it at a regular cadence. One of Nvidia’s biggest investments is in CoreWeave, of which it owns nearly 7 percent.

In late 2024, Nvidia completed an acquisition of Run:ai; in its most recent earnings call, the company announced commitments of $18 billion in equity investments for the remainder of the fiscal year, with its private company holdings almost doubling from 2025 to 2026. And in late August 2026, The Information reported Nvidia’s intention to acquire Hugging Face, a platform for open-source AI models, for $12.9 billion. Huang has noted that AI agents use “15-100x” the compute that a human using an AI model uses; it may be that Nvidia-owned Hugging Face begins preferencing agentic models, driving AI toward compute- and energy-heavy use cases (that benefit Nvidia).

In early August 2026, Nvidia announced a partnership with six private equity and investment bankers including Apollo, BlackRock, Goldman Sachs and KKR to “turn Nvidia compute and full-stack AI infrastructure into an investable asset class for global capital,” according to the press release. The partnership will pool capital and offer “attractive rates” for enterprises to invest in Nvidia infrastructure.

In other words, the AI bubble is now an asset for some of the biggest players on Wall Street, while Nvidia grows its bottom line.

The explosion of the AI industry and the vast amounts of capital sloshing through seem to have taken many political leaders by surprise. But in one sense, this is an old story: a powerful corporation with a near-monopoly on a valuable product has amassed enough capital to begin acting as an unregulated banker, putting the U.S. economy at risk. If AI demand fails to keep pace, Nvidia’s financial entanglements across the space become a massive liability.

It is time for Congress to investigate these financing practices, demand more transparent financial disclosures, and pass a ban on Nvidia’s (and others’) ability to move vertically through the tech stack through acquisition or other investment. Nvidia has become a corporate giant, and these steps alone may not curtail its march through the AI industry — but Silicon Valley’s unregulated bank must be brought to heel.

Olivia Kosloff is a senior fellow at the American Economic Liberties Project, where Colin Dunkley is a legal analyst.