The ‘Big Short’ Investor Betting $1 Billion Against the AI Bubble Says Meta and Oracle’s Accounting is Hiding the Brutal Truth

Dr. Michael Burry, the enigmatic investor who cemented his place in financial history by foreseeing the 2008 subprime mortgage crisis, has escalated his war against the artificial intelligence boom. After recently disclosing a massive, billion-dollar short position against AI darlings NVIDIA and Palantir, Burry is now aiming his sights at the industry's biggest spenders: Meta Platforms (META) and Oracle (ORCL).

In a series of new cryptic statements, the head of Scion Asset Management is alleging that the financial justifications for the AI gold rush are built on a foundation of "creative accounting."

Burry's new thesis is that the true, staggering costs of the AI arms race are being effectively hidden within the financial statements of tech giants. He claims that by using aggressive depreciation schedules for AI hardware, these companies are obscuring the "brutal truth" of the industry: the return on investment (ROI) for AI is nowhere near high enough to justify the colossal expenditures.

This new claim marks a significant broadening of his attack. Burry is no longer just betting against the "shovel sellers" like NVIDIA; he is now arguing that the "gold miners" themselves (Meta, Oracle, and by extension, their peers) are not finding any gold, and are simply using accounting rules to pretend they are.

A Recap: The Billion-Dollar AI Short

To understand this new attack, one must first look at the bet that put the AI industry on notice. Earlier this month, regulatory 13F filings revealed that Michael Burry’s fund, Scion Asset Management, had taken on massive bearish positions with a notional value of over $1.1 billion.

The two targets were the epicenter of the AI boom:

• NVIDIA (NVDA): A $187 million (notional) put option bet against the chipmaker that provides the essential GPUs for all large language models (LLMs).

• Palantir (PLTR): A much larger $912 million (notional) put option bet against the data analytics company that has rebranded itself as a premier AI software platform.

This move was a classic Burry play: a bold, contrarian bet against a market narrative that he views as detached from reality. He is betting that NVIDIA's $5 trillion valuation and Palantir's sky-high sales multiples are a speculative bubble, not a rational reflection of their future earnings.

The New Claim: Hiding the "Brutal Truth"

Burry's latest salvo moves beyond just over-valuation. He is now attacking the fundamental economics of the entire AI ecosystem. His new focus is on the "buyers"—the hyperscale companies like Meta and Oracle that are collectively spending hundreds of billions of dollars to build out their AI infrastructure.

The core of his argument is that these companies are using an accounting tactic to make their massive spending appear less damaging to their bottom line.

The Accounting Mechanism: Capital Expenditure and Depreciation

When a company like Meta or Oracle buys $10 billion worth of NVIDIA GPUs, it's not recorded as a $10 billion expense in that quarter. Instead, it's listed as a Capital Expenditure (CapEx), an asset on the balance sheet.

This asset's cost is then expensed gradually over its "useful life" through a process called depreciation. Herein lies the "trick" Burry is pointing to.

Traditionally, a server's "useful life" might be pegged at 4, 5, or even 6 years. This means the $10 billion cost is spread out, becoming a $2 billion or $2.5 billion annual hit to profits.

Burry's argument is that this is a dangerous fiction in the AI era. The "brutal truth" is that the pace of innovation is so ferocious that a top-of-the-line, $40,000 GPU purchased today may be functionally obsolete in 18-24 months, not 5-6 years.

If Meta is depreciating its AI hardware over five years when the actual useful life is only two, it is, in effect, overstating its current profits by billions of dollars every single quarter. The "brutal truth" is that the true cost of staying on the AI frontier is far higher than their earnings reports suggest.

Why Meta and Oracle are in the Crosshairs

Burry's choice of Meta and Oracle as examples is deliberate. They represent two key pillars of the AI spending spree.

Meta: The All-In AGI Gamble

Meta CEO Mark Zuckerberg has publicly stated that the company is on a path to building Artificial General Intelligence (AGI) and is willing to spend "billions" to get there. The company is aggressively building out its infrastructure to power its Llama family of models and its consumer-facing AI assistants.

This spending is colossal. Burry's claim implies that this AGI pivot is a potential black hole for cash, and that Meta is using a long depreciation window to mask the fact that this gamble is not yet generating a proportional return. The AI features in glasses and social apps, while interesting, are not yet the multi-billion dollar revenue streams needed to justify the GPU bonfire.

Oracle: The Cloud Challenger

Oracle, a legacy database giant, is in a desperate and expensive war to catch up to Amazon (AWS), Microsoft (Azure), and Google (GCP) in the cloud computing market. Its strategy has been to position its Oracle Cloud Infrastructure (OCI) as the best platform for AI startups and enterprises.

This requires building data centers at a breakneck pace, a process that involves acquiring billions of dollars in NVIDIA hardware. Burry is likely pointing to Oracle as another key player whose profit margins may be artificially inflated by accounting practices that don't reflect the high-speed, high-cost reality of the AI cloud race.

The "Circular Economy" and Systemic Risk

By targeting the buyers (Meta/Oracle) as well as the sellers (NVIDIA), Burry is highlighting a potential systemic risk in the market.

For over a year, analysts have pointed to a "circular economy" in AI. A company like Microsoft invests $10 billion in OpenAI. OpenAI then uses that $10 billion to buy GPUs from NVIDIA. NVIDIA reports a record quarter, and its stock soars. This, in turn, boosts the value of Microsoft's stock (and its AI narrative), which lifts the entire S&P 500, creating the appearance of a booming economy.

Burry's thesis cuts this narrative at the knees. He is saying the "demand" from companies like Meta and Oracle is not based on a profitable business case, but on a "fear of missing out" (FOMO), with the true costs being deferred via an accounting lever.

If he is right, it means the entire AI sector is built on unsustainable spending, and the "brutal truth" is that the unprecedented CapEx boom is not translating into real, durable profit.

Future Outlook: The Prophet vs. The Permabear

As with all of Michael Burry's predictions, the immediate question is one of timing. Critics will point out that he has been famously early on many of his bearish calls, including his short against Tesla, and has sat out (or bet against) one of the greatest bull markets in history. In this view, he is a "permabear" who is simply misinterpreting a genuine technological revolution as just another financial bubble.

However, his supporters would argue that his fundamental analysis is sound. The dot-com bubble of 2000 was also built on a genuine technological revolution—the internet. But that didn't stop companies like Cisco, which sold the "shovels" (routers) for that boom, from seeing its stock crash by 90%.

Burry is betting that the same dynamic is at play today. He is arguing that the AI revolution can be real, while the valuations attached to it can be a complete fantasy. By questioning the accounting of Meta and Oracle, he is suggesting that the rot is even deeper than anyone realizes, and that the market is ignoring the most fundamental rule of business: at some point, the numbers have to add up.