On November 13, 2026, less than 24 hours after Nvidia co-founder Jensen Huang announced his resignation, The Wall Street Journal published an investigative report. The report said that Nvidia systematically provided financing to AI startups and small cloud service providers unable to afford its GPUs, thereby sustaining the market's perception of "high demand" for chips while helping absorb inventory backlog under its TSMC capacity commitments. According to data cited by The Wall Street Journal, Nvidia's total guarantee commitments in the most recent quarter reached $530 billion, compared with just $184 billion in the previous quarter.
The Logic of Circular Financing
The circular financing mechanism works as follows: Nvidia provides funds or credit guarantees to small cloud service providers or AI startups; these companies use the money they receive to buy Nvidia GPUs; Nvidia's shipment figures then look better; the market forms a judgment that demand for AI compute is strong; and Nvidia's valuation is supported. Money flows out of Nvidia and, after going in a circle, returns to Nvidia in the form of purchase orders—Nvidia is both the seller and the source of funds for the buyer.
This model emerged because of Nvidia's structural dilemma. Over the past two years, hyperscale AI companies—Meta, Google, Microsoft, and Amazon—have been accelerating their shift to custom chips: Google's TPU, Amazon's Trainium, and Microsoft's Maia. These companies' dependence on Nvidia's off-the-shelf GPUs is systematically declining. At the same time, Nvidia has enormous, non-cancelable capacity commitments at TSMC, and silicon wafers keep pouring in. According to reports, Huang himself described this situation internally as an impending "GPU avalanche."
If you cannot sell to major customers, create buyers—that was Huang's solution. Cultivate a group of small cloud service providers and let them use Nvidia's money to buy Nvidia's products.
Intense Internal Conflict
According to The Wall Street Journal, in August this year, Huang held an internal meeting with the senior finance team, including CFO Colette Kress. Kress explicitly proposed that Nvidia should scale back such circular financing arrangements. She also opposed providing up to $250 billion in credit guarantees for OpenAI's Ohio data center project, arguing that the scale was too large and the risk difficult to control.
The report said Huang forcefully pushed back on the spot, calling Kress a "deer in the headlights." The English idiom describes someone who freezes completely under pressure and is unable to make judgments.
The $250 billion OpenAI deal was later renegotiated. According to a filing Nvidia submitted to the U.S. SEC on August 17, 2026, the contingent guarantee obligation was ultimately capped at about $105 billion, covering about 4.25 GW of data center capacity. That figure was less than half of the original proposal.
On November 12, Huang formally resigned as Nvidia CEO. Kress took over as acting CEO.
The True Scale of Risk Behind the Numbers
The $530 billion in guarantee commitments must be placed within Nvidia's overall financial framework to appreciate their weight. In addition to the $530 billion in total guarantees, according to The Wall Street Journal, Nvidia has also locked in $279 billion for memory and wafer procurement, with another $36 billion tied to repurchase agreements for unused cloud capacity; meanwhile, the estimated market value of idle GPU inventory in Nvidia's warehouses has exceeded $200 billion.
Together these numbers form a fragile balance sheet. The deeper hidden danger lies in the commitment chain itself: the small cloud service providers funded by Nvidia simply do not have the financial capacity to keep buying high-end GPUs. The Wall Street Journal explicitly stated that these companies "cannot afford the cost of buying Nvidia GPUs," meaning Nvidia is both the lender and the only potential recipient of payment. Once these companies run into operational trouble, the loop will break—Nvidia will face the dual pressure of bad debt and inventory at the same time.
Distorted Market Signals
Over the past three years, the investment boom in the global AI compute market has depended heavily on a core narrative: GPUs are severely undersupplied, and compute is the hard constraint in the AI race. This narrative supported Nvidia's market capitalization exceeding $3 trillion at its peak and also drove the continuous escalation of the global data center arms race.
But if some of that "demand" comes from buyers manufactured with Nvidia's own money, then investment decisions, government policies, and academic forecasts based on the GPU shortage are all built on a manipulated supply-demand signal.
The essence of the problem is not "which companies Nvidia invested in," but: when a semiconductor company is both the seller and the source of funds for buyers, can the market still accurately distinguish real demand from financing demand? Even Nvidia's competitor AMD must price in this distorted market—assuming demand is real. As a result, part of the premium consumers and enterprises pay for GPUs may stem from a carefully designed illusion of supply and demand.
Regulatory Clouds Have Not Dispersed
Nvidia has not yet formally confirmed The Wall Street Journal's report, nor has any regulator publicly announced an investigation. But the pattern of behavior described in the report has multiple potential legal trigger points.
First is disclosure compliance at the securities level. Nvidia is a public company. If it can be proven that the purpose of the circular financing was to keep GPU prices high and influence public judgment about supply and demand, this could trigger securities regulatory scrutiny—because investors made actual investment decisions based on these amplified demand signals. Second are antitrust concerns. Nvidia holds more than 80% of the AI training GPU market. If it artificially created scarcity to squeeze out competitors such as AMD, both the U.S. Department of Justice and the European Commission would have grounds to intervene. Third is the adequacy of related-party transaction disclosure: whether the GPU purchases by funded startups were sufficiently disclosed financially under relevant rules is itself a compliance question.
The Cleanup Challenge of the Kress Era
In a sense, Kress was forced to take over a mess she had tried hard to avoid three months earlier. Her opposition to circular financing at the August meeting now looks more like a clear-eyed warning than merely a conservative financial preference. According to reports, Nvidia had already suspended some financing arrangements at the end of August and introduced a high-level review mechanism including board members—moves that preceded Huang's resignation and indicated that internal awareness of the risk was quietly shifting.
The core challenge Kress faces after taking over is: how to orderly shrink these unsustainable circular financing arrangements without triggering market panic over a collapse in demand? The $530 billion in guarantee commitments cannot disappear overnight, but continuing to maintain this balance sheet means Nvidia will bear increasing systemic risk.
Independent Judgment
From the logic of industrial evolution, the predicament Nvidia now faces has its inevitability: as large-model training efficiency improves and inference costs continue to fall, the migration of major customers to custom silicon is an almost irreversible trend. Huang's "circular financing" merely used financing tools to delay the arrival of this cycle, at the cost of systematically accumulating risk on Nvidia's balance sheet until it became too big to hide.
More alarming is the cost of signal distortion. The AI industry's capital allocation over the past three years has been partly built on the premise that "GPUs are extremely scarce." If this premise was artificially amplified, then the enterprises that entered at high prices because of it, the startups that received high valuations because of it, and the governments that formulated GPU compute subsidy policies because of it—all paid tuition for a manipulated market without knowing it when they paid.
Whether Nvidia broke the law is for courts or regulators to decide. But what this report reveals is the complete path by which a monopolistic hardware company, on the eve of a market turning point, used financing tools to transform private inventory risk into industry-wide systemic risk. This mechanism itself—regardless of whether it is ultimately found to be in violation—will long serve as a warning about supply chain transparency in the AI industry.
© 2026 Winzheng.com 赢政天下 | 转载请注明来源并附原文链接