Is the “big short” in a hurry? Bury criticized Nvidia's $500 billion financing as a “2008 financial project,” yet the market used real money to bolster AI computing power demand

The Zhitong Finance App learned that during the opening session of the US stock market on Wednesday, Michael Burry (Michael Burry), who has the title of “big short”, fiercely criticized the 500 billion US dollar artificial intelligence infrastructure financing arrangement led by the “AI chip superhero” NVDA.US (NVDA.US). He compared the complex financing network composed of private credit and asset-backed securities to the systemic financial risks accumulated on the eve of the 2008 financial crisis, and called the arrangement “a similar 2008 trick unique to Wall Street.”

Michael Burry, who has the title of “Big Short”, has been posting pessimistic remarks about “the end is coming” on his Substack subscription platform, and is shorting popular AI technology stocks at a time when global capital continues to flood into the topic of AI computing power infrastructure. The prototype character of the movie “The Big Short” recently focused more on crowded transactions related to AI computing power infrastructure and AI semiconductor capital expenditure cycles. For example, Burry expressed continued bearish positions on Nvidia, Tesla, Micron, Applied Materials, Palantir Technologies, and iShares Semiconductor ETFs through put options or bears.

In fact, it has upgraded its short portfolio into a systematic bet on “AI capital expenditure plus AI valuation.” However, the recent strong counterattack on AI semiconductors and the broader AI computing power theme, especially the rise in technology stocks driven by moderate US CPI growth data on Wednesday, is indeed putting significant pressure on this “shorting AI” trading mechanism. In particular, the current CoreWeave capacity sold out, the A100 extended to 2029, Nebius's price increase and $40 billion customer commitment, and Nvidia's falling credit spreads all significantly biased the balance in favor of Wong In-hun/Morgan Stanley's “AI Factory Financeable Asset” framework.

Some investors also commented that Big Bear's latest fierce attack on Nvidia's leading financing activities — approving the $500 billion financing project as a Wall Street financial project similar to 2008 seems to be “out of a hurry.” At the same time, CoreWeave raised its capital expenditure in 2026 to 35 billion US dollars to 39 billion US dollars. Nebius Q2 revenue surged year-on-year and indicated that it could even sell out the 2027 planned capacity under current conditions. Actual transactions and the industry chain level are clearly strengthening the multi-faceted narrative that “demand for AI computing power is still real and scarce.” By contrast, Bury, who is shorting Nebius, may now be under pressure to make up.

Scion Asset Management, where Bury was previously at the helm, deregistered its registration in November 2025, so what we are seeing now is more personal transactions that Bury actively disclosed through the paid subscription platform “Cassandra Unchained” of the Substack platform, rather than a snapshot of traditional 13F positions. According to recent public information, he still maintains a bearish outlook on Nvidia, Palantir, SOXX Semiconductor ETF (Philadelphia Semiconductor ETF), and QQQ (Nasdaq 100 ETF), and extended some Nvidia/QQQ put options until 2027. At the same time, on August 7, he also revealed that he directly shorted Nebius for about $212, and claimed that the position size was “relatively larger”. However, after Neocloud (Neocloud) CoreWeave and Nebius announced strong financial reports on Wednesday, the two major financial reports The stock price of cloud companies is large Up nearly 20%.

Previously, Palantir surged by about 29.5% in a single day after the financial report exploded on August 4, and the entire Palantir bear camp lost a record amount of about $3 billion on the same day; however, Bury recently re-established Palantir off-price put options expiring in 2027 before the AI app leader announced strong financial reports, so its bear portfolio apparently was under severe pressure from a new mark-to-market — that is, when market prices rise, shorters (bears) sold assets at higher prices than when they were bought due to higher asset prices than when they were bought To achieve losses, brokers will require them to continuously supplement their security deposit.

The $500 billion AI financing was sniped by a “big short”! Bury questions Nvidia-led circular leverage

Nvidia, one of America's top seven tech giants, has partnered with Apollo Global Management (APO.US), Blackstone Group (BX.US), Blackstone (BLK.US), Brookfield (BN.US), Goldman Sachs (GS.US), and KKR & Co. (KKR.US) signed a high-profile memorandum of understanding to establish the first batch of such AI computing power infrastructure financing platforms worldwide. The initial funding scale has reached 500 billion US dollars.

As a market platform for matching transactions, Nvidia plans to provide up to 25% guarantee for individual projects through a “residual value mechanism” to support these financing transactions. If a project is in trouble, the chip giant will provide financial support by selling its proprietary AI chip products, which are in high demand, on a large scale, or finding new companies to rent related artificial intelligence computing power capacity, so as to minimize losses.

However, Burry alleges on the subscription platform that there is a highly leveraged and revolving funding chain behind these transactions. According to an infographic he produced, annuity premiums paid by American retirees to Athene, a subsidiary of Apollo (APO.US), will be transferred through an offshore reinsurance company based in Bermuda, and further leveraged by Apollo and converted into asset-backed debt. Allegedly, Apollo issued $3.5 billion in debt to a special purpose entity called Valor, while Nvidia directly injected $1.9 billion in equity capital. VCI then purchased Nvidia GB200 GPUs worth $5.4 billion and leased these GPUs to xAI for use in its Grok supercomputer cluster.

For those who doubt the AI boom, such as the “big bears,” this complex financing structure artificially boosts income data while passing on the risk of depreciation to unaware retirees.

“I probably know what this will look like in the end,” Barry wrote in a post on the X platform. Bury recently further increased direct shorting bets on Nvidia and other leaders in the AI computing power industry chain through put options. “Meet the new boss. It's actually no different from those old bosses in 2008.” Barry said sarcastically.

Is $500 billion a “financial trick” or financing the AI industrial revolution? Credit spreads have declined, Neocloud is booming, and the capital chose to trust Hwang In-hoon for the time being

After Nebius announced its financial outlook on Wednesday, the US stock market once soared by nearly 20% at the beginning of the session, which happened to be once again approaching the cost of shorting Pariot's $212. Some Wall Street analysts even thought that the short recovery might be one of the amplifiers of the company's strong rise in stock prices. Needless to say, Bury is experiencing severe bearish price pressure, but there is currently no reliable evidence that Bury himself has been forced to close its position or must make up for it.

The recent explosion performance and strong performance outlook of the two most important upstream AI computing power industry chains, Asmack and TSMC, as well as the strong performance and outlook recently announced by CoreWeave, Hon Hai, and Nebius, are actually sending an important signal to the global stock market: the AI computing power industry chain has gradually moved from the “AI capital expenditure supercycle of training AI models” to a new stage of “exponential expansion of AI inference computing power demand driven by large-scale applications of smart devices”. These latest signals can be described as shocking recent signs that have caused AI computing power in particular The pessimistic argument that the AI semiconductor sector has plummeted “excess computing power.”

Coreweave, which is closest to the demand for AI computing power terminals, has just raised its 2026 capital expenditure forecast to an astonishing 35 billion to 39 billion US dollars. Q2 revenue significantly exceeded expectations, and said it will continue to sign cloud computing power contracts on favorable terms while supply is still limited. At the same time, SK Hynix also reported record quarterly profits and revealed that it has signed long-term supply agreements with about 10 core customers, reflecting the shift in demand for high-performance storage such as HBM from short-term orders to multi-year lock-ups. It can be said that these various links at the industrial chain level are sending a positive signal that demand for AI computing power is still expanding at an accelerated pace.

Does the market believe in Bury at the moment, or does it believe in Wong In-hun+Morgan Stanley? The answer is pretty clear: marginal funding is now clearly on the latter side, but Burry's long-term risk argument has not been declared invalid.

The most important evidence is not even stocks, but the credit market: when the $500 billion plan was first announced, the market feared that Nvidia would bear huge off-balance sheet risks; Wong In-hoon later made it clear that this 500 billion US dollars was third party capital to be gradually mobilized, not Nvidia revenue, nor funds already promised to a certain customer. Apollo, Blackstone, Brookfield, Goldman Sachs, and KKR will independently underwrite each project. Nvidia's maximum support is only about 25% of the collateral value of a single project (residual-value) Backstop).

The interest spread of Nvidia 2056 bonds with US bonds then narrowed by about 2 bp to 113 bps, and the 5-year CDS narrowed at most by about 5 bp to 72.11 bp — this is a very clear bond market language: investors lowered Nvidia's credit risk pricing after Huang Jen-hoon clarified the structure.

The judgment of Morgan Stanley and Bank of America is also broadly consistent: third-party professional capital dominates due diligence decisions, clearly weakening the purest circular AI prosperity logic of “Nvidia pays — customers buy Nvidia GPUs — Nvidia inflates revenue”; at the same time, actual revenue linked to usage may also allow Nvidia to further share the AI factory cash flow from selling GPUs at once. In other words, the market currently does not think that the $500 billion is “without credit risk,” but rather that the risk is more like underwriting and priceable infrastructure financing risk, rather than the distorted demand for assets like before 2008.

Bury is betting that the 2008 Wall Street financial project will be repeated, but Coreweave and Nebius handed over the demand explosion questionnaire

Burry is concerned about end-of-line credit risk that may eventually result from leverage, GPU depreciation, and asset securitization; long capital is currently trading in short supply of computing power, long-term contracts, price increases, and sustainable cash flow. Nvidia's current platform itself is also mainly designed to mobilize capital from third parties. Nvidia can provide up to 25% residual value support for each project; at the same time, CoreWeave disclosed a backlog of more than 100 billion US dollars of orders, and Nebius disclosed customer promises of more than 40 billion US dollars.

What really poses the strongest challenge to Bury's “2008 financial engineering” argument is the physical demand, price, and residual value evidence given simultaneously by two Neocloud (Neocloud) leaders — CoreWeave and Nebius today. CoreWeave Q2 revenue reached about $2.6 billion, +112% year over year, and the backlog of orders even rose sharply to $104.2 billion, and did not include the promise of more than $25 billion signed at the beginning of the quarter; the company clearly stated that near-end capacity was “actually sold out,” so it was signing new contracts at increasingly favorable prices. The J.P. Morgan analyst team even attributed this price increase to the demand environment and the accelerated improvement in ROI for cloud computing and AI application customers.

More importantly, CoreWeave directly gave Bury an extremely lethal counterexample to the “GPU only has an economic life of 2 to 3 years” argument: the company just signed a new contract for the A100 released in 2020, and the price is still attractive; management also made it clear that the old GPU fleet is still basically sold out, and that re-leasing after the initial contract ended was an incremental benefit over the existing return on investment, and observed a longer usage cycle and higher price.

Nebius' evidence is even more aggressive: with Q2 revenue of $582.3 million and over the agreed estimate of $5727.5 million, AI Cloud (AI-related cloud computing business) revenue was nearly six times the same year over year, signing 4 contracts averaging more than $1 billion in a single quarter. The value of new customer contracts increased more than 9 billion times month-on-month, with over 40 billion US dollars in customer commitments + expected to exceed 9 billion US dollars in customer advance payments this year; the company even said that under current conditions, it can already sell out the full planned capacity for 2027. This latest set of operating data strongly supports Hwang In-hoon's statement that “demand is real, and AI computing power in production is a scarce productive asset,” at least as of August 2026, rather than supporting “financing has created false GPU demand.”