The dust had barely settled after the forced resignation of co-founder and longtime CEO Jensen Huang when new details emerged over the aggressive financial strategy that propelled NVIDIA Corp.’s historic rise — and drove graphics processing unit (GPU) prices to record highs.
NVIDIA’s leadership systematically used circular financing to manufacture demand for its own hardware, leaving interim CEO Colette Kress to navigate more than $530 billion in financial and contractual obligations, according to an analysis by The Wall Street Journal.
For more than a year, industry observers accused NVIDIA of artificially prolonging the artificial intelligence (AI) boom by bankrolling the very startups and cloud providers buying its chips. Under Huang’s direction, NVIDIA frequently acted simultaneously as an investor, loan guarantor, hardware supplier, and buyer of excess computing capacity.
According to The Wall Street Journal’s profile of Kress, the strategy served a dual purpose: absorbing a massive oversupply of semiconductor wafers ordered from Taiwan Semiconductor Manufacturing Co. (TSMC) and keeping GPU prices elevated by simulating overwhelming market demand.
At a conference in Berlin, Huang illustrated this dynamic to clients by recounting a story about a Japanese soup vendor who refused him a second serving due to long lines, signaling the value of perceived scarcity in maintaining high prices.
The ripple effects have hit the consumer market hard.
As NVIDIA diverted DRAM, NAND, and semiconductor manufacturing capacity toward lucrative data center hardware, retail GPU prices surged past 160% of their suggested retail prices, rendering flagship cards like the GeForce RTX 5090 virtually unobtainable.
Mounting Risks and Internal Friction
The scale of NVIDIA’s commitments — many routinely refer to it as the central bank of AI — eventually sparked internal friction.
By mid-2026, facing what insiders described as an “avalanche” of incoming TSMC shipments, Kress — then serving as chief financial officer — reportedly raised alarms over the size of the company’s backstops.
Huang allegedly berated her during this period, calling her a “deer in headlights,” though NVIDIA subsequently began scaling back several massive commitments, including a proposed $250 billion backstop for an OpenAI data center.
Still, the company remains heavily exposed. NVIDIA holds $279 billion in contractual purchase obligations to suppliers, and its total financial exposure — including contingent loan and lease guarantees— exceeds $530 billion.
In August, NVIDIA agreed to guarantee OpenAI’s Ohio data center lease up to $105 billion over 20 years. It also recently contributed roughly $10 billion to a $26 billion pre-IPO funding round for Anthropic, which carries $895 billion in long-term computing commitments.
An International Monetary Fund (IMF) review of 14 major AI deals since 2023 underscored the broader systemic risk, finding that Big Tech firms committed roughly $300 billion in equity and $650 billion in debt to weaker AI firms in exchange for $1.4 trillion in future chip and cloud orders.
In a recent Federal Reserve survey, 87% of respondents cited “complex AI financing” as the top risk to the financial system.
“Follow the dollars. What is financing the growth reinvested profit from a healthy P&L or debt supported by leases (and how sound are those leases); funds flowing back from your own investment; and/or cash from a start-up’s venture financing?” asked John Poeta, president at Stealth Mode. “Is it manufactured growth or true growth due to the demands of a paying, profitable customer?”
A Precarious Path Forward
Huang’s abrupt departure — triggered by revelations of extramarital affairs, including one with an executive whose firm is under U.S. intelligence investigation for allegedly smuggling NVIDIA chips to China — removes the primary architect and public guarantor of these deals.
With Kress now at the helm, analysts expect a sharp pivot toward balance-sheet caution, including stricter reviews of existing guarantees and fewer new investments.
However, the transition leaves NVIDIA in a bind as major tech rivals increasingly switch to in-house custom chips. Without Huang’s personal backing, and with 10-year U.S. Treasury yields above 5% ahead of an expected December interest rate hike, AI startups face higher borrowing costs and tighter credit.
While existing backlogs will shield immediate earnings, reduced access to NVIDIA-backed financing is expected to curb chip orders next year.
Continuing the strategy risks turning contingent liabilities into crushing debt, whereas abruptly cutting off funding could trigger widespread startup bankruptcies, depressing chip demand and threatening a broader downturn across the U.S. technology sector.

