Nvidia Is Becoming the Central Bank of AI

The Balance Sheet Is the Product

Nvidia is no longer just a chip company. It has become the financial architecture underneath the entire AI industry.

At the end of July, Nvidia’s public and private equity holdings reached $95.6 billion, up from less than $100 million in early 2020. CFO Colette Kress put the broader number at roughly $99 billion. That portfolio includes stakes in Intel, CoreWeave, Coherent, Nokia, Synopsys, and Nebius. On the conference call, Jensen Huang didn’t deflect from the scale of it. His only regret, he said, was that he hadn’t invested more and sooner.

This isn’t a side hustle. Nvidia is now in the business of financing the companies that buy its products. And it is starting to show up in the numbers.

Accounts receivable — money customers still owe — jumped to $63.1 billion. The average collection period rose to 60 days from 45. Some large customers can now take as long as a year to pay for data center purchases. Nvidia also disclosed up to $108.5 billion in guarantees, $105 billion of which is tied to the SB Energy data center buildout for OpenAI.

Morgan Stanley has been calling it balance-sheet-as-a-service. Nvidia’s financial strength is being deployed to keep its fastest-growing customers spending, and in return Nvidia collects twice: once on the hardware sale, and again through a share of rental revenue.

How the Central Bank Metaphor Holds Up

A central bank doesn’t just print money — it decides who gets credit, on what terms, and under what conditions. That is exactly what Nvidia is doing now.

Frontier AI labs have extraordinary demand for computing power, Kress said. But they are growing faster than their balance sheets and credit profiles can support. Nvidia is stepping into that gap. The implicit trade-off is straightforward: Nvidia gets preferential positioning in the upside of its customers’ success, and those customers get the capital they cannot easily source elsewhere.

Consider what this means for a company like CoreWeave. It is both a customer buying Nvidia GPUs and a portfolio holding Nvidia owns. If CoreWeave stumbles, Nvidia absorbs part of the hit. If it soars, Nvidia shares in the gains. That relationship didn’t exist a few years ago. Today it is structural.

Nvidia expects AI labs using its balance sheet to account for roughly a quarter of its business next year. That is not a footnote. It is a new revenue engine built on financial engineering rather than silicon alone.

Who Gains and Who Loses

The winners are obvious in the short term. Frontier labs that could not otherwise fund their expansions now have access to capital. Cloud providers like Nebius get guaranteed revenue floors that make bank financing easier to secure. Nvidia’s own moat deepens because switching from Nvidia hardware becomes riskier when your balance sheet is intertwined with the supplier’s.

The risks are less visible but more dangerous.

Nvidia has effectively concentrated a massive amount of counterparty exposure onto its own balance sheet. $108.5 billion in guarantees is a enormous number for any company, even one valued at trillions. If the AI capex cycle slows — and cycles always slow — those guarantees become liabilities. Revenue floors turn into losses. Receivables go unpaid. The very customers Nvidia is propping up become the source of its fragility.

There is also a competitive dimension that competitors are only beginning to grasp. AMD and Intel sell chips. Nvidia now sells chips plus capital allocation. A buyer choosing AMD isn’t just changing hardware; it is opting out of a financing relationship that has become embedded in the economics of running an AI data center. That switching cost is real, and it is growing every quarter.

The Second-Order Consequences

The most important shift may be how this changes the geometry of the entire semiconductor ecosystem. Nvidia is simultaneously a customer, a supplier, an investor, and a lender. In practice, that means it can influence the market in ways no pure chipmaker could.

It can pressure its portfolio companies to favor Nvidia hardware. It can structure guarantees that only apply when customers stick with Nvidia. It can collect receivables from companies whose survival it partly underwrites. This is vertical integration at a financial level, not just a product level.

Regulators in the United States and the European Union have not yet framed this as an antitrust concern. But it fits a pattern that historically triggers scrutiny: when a dominant firm controls not just the market but the capital markets around that market. Nvidia is not charging interest the way a bank does. It is extracting value through equity appreciation and revenue-sharing agreements. The economic effect is similar. The legal exposure is different.

What Comes Next

Nvidia’s quarterly revenue guidance of roughly $108 billion confirms the underlying business remains strong. But the composition of that revenue is shifting. A quarter of next year’s business may come from customers leveraging Nvidia’s balance sheet rather than paying outright. That makes Nvidia’s earnings profile more correlated with the health of its own portfolio companies than it was two years ago.

The central bank metaphor is useful precisely because it captures what is happening: Nvidia is creating liquidity for an ecosystem that would otherwise run out of it. The Fed prints dollars to keep credit flowing. Nvidia prints optionality — through equity stakes, guarantees, and extended payment terms — to keep AI companies building. The difference is the Fed can walk away from bad bets. Nvidia cannot. Its fortune is now woven into the same fabric as the companies it finances.

That is either the smartest hedge Nvidia has ever taken or the most concentrated risk it has ever accumulated. Both descriptions are true at the same time. The only question is which one becomes reality first.