What Is “Circular Financing”? Nvidia Funds the Buyers
A Wall Street Journal investigation published November 13 has pulled back the curtain on Nvidia’s financial engineering during the GPU boom: the company allegedly used its own balance sheet to invest in AI startups and small cloud providers — customers who would then turn around and buy Nvidia GPUs and build data centers. The arrangement, known as circular financing, reportedly served a second purpose: absorbing the excess chips and memory Nvidia had over-ordered from TSMC, preventing an oversupply from crashing GPU prices and shattering the shortage narrative.
According to sources cited by the Journal, Jensen Huang privately acknowledged that hyperscalers were pivoting to in-house custom silicon and cutting reliance on off-the-shelf parts — while the prepaid TSMC wafers and memory commitments came at him “like an avalanche.” Rather than let the surplus flood the market, the report says, Nvidia paid to have someone else soak it up.
The August Showdown: Kress Halved the OpenAI Guarantee — and Got Mocked
At the center of the report is an August 2026 confrontation. CFO Colette Kress — now interim CEO — pushed to rein in circular-financing commitments to small cloud operators and major AI customers including OpenAI and Anthropic. Early discussions reportedly floated underwriting roughly 8 gigawatts of OpenAI compute worth up to $250 billion; under Kress’s insistence, Nvidia’s signed guarantee cap was cut to about 4.25 gigawatts, or up to $125 billion — sliced in half.
The price of dissent? Huang reportedly belittled Kress for questioning the deals, calling her “a deer in the headlights.” Three months later he abruptly resigned, Kress took the wheel, and the report says she is now unwinding the riskiest commitments — even contacting TSMC to wind down future chip orders.
$200 Billion of Idle GPUs: Shortage or Illusion?
Industry analyst Ed Zitron has estimated that at least $200 billion worth of Nvidia AI GPUs sit idle in cloud providers’ warehouses — bought, but unpowered for lack of grid capacity or end-user demand. If the Journal’s allegations hold, the contrast with the market’s “sold-out everywhere” narrative is stark: at least part of the severe shortage that propped up memory and GPU prices for years may have been a financial artifact.
The Gamer’s Bill: RTX 5090 at $9,400
For consumers, this is the sharpest edge. TechSpot’s late-2026 pricing survey put the GeForce RTX 5090 at roughly $9,400 in the U.S. against a $1,999 launch MSRP, with the RTX 5080 near $2,200 versus $999. If Nvidia was locking up capacity with demand it funded itself, gamers were effectively bidding against Nvidia’s own portfolio companies for the same silicon.
Caveats: What Is Confirmed vs. Inference
The boundaries matter. Nvidia’s fiscal 2026 annual report does confirm $17.5 billion invested in private companies and infrastructure funds, and a $105 billion cap on guarantee exposure; gaming revenue grew 41% year over year, hitting $3.7 billion in Q4 — gamers are still buying. But the investee list has never been disclosed, the August meeting lacks independent confirmation, and the causal chain “financing to sustain prices” remains an allegation, not a finding by any court or regulator. With Nvidia having shed $490 billion in market value in a single day during the Huang resignation shock, definitive verdicts are premature.
Conclusion
The report’s real punch is not that it proves fraud — it doesn’t — but that it forces the market to re-audit how much of AI chip demand is organic. For investors, Kress’s de-risking and the TSMC order adjustments are the hard indicators to watch over the coming quarters. For gamers, if the “avalanche” eventually lands as inventory clearance, inflated card prices could finally crack. Until the facts settle: read the data, discount the story.
