Michael Burry Calls AI Infrastructure "Fraud": Examining the Depreciation Math Behind the Accusation

Michael Burry just accused Big Tech of accounting fraud.

The investor who famously predicted the 2008 housing collapse posted that Meta, Oracle, Microsoft, Amazon, and Google are manipulating depreciation schedules on AI infrastructure to overstate earnings. According to Barchart's analysis, Burry estimates these companies could be overstating profits by 20-26% through 2028 by spreading chip depreciation across longer periods than the assets actually remain useful.

The accusation landed the same week Burry reportedly shut down his hedge fund to transition to a family office. Which either validates his concerns about market conditions or suggests he's tired of explaining why everyone else is wrong while they keep making money.

The Core Accounting Claim: AI Chip Depreciation Schedules

Burry's argument centers on a straightforward math problem. AI chips cost millions of dollars and become obsolete within 2-3 years as newer, more efficient models arrive. But hyperscalers are depreciating these assets over longer periods—sometimes 5-7 years—which reduces annual depreciation expenses and inflates reported earnings.

According to Barchart's breakdown, Burry estimates the accounting gap represents:

• Oracle: +26% earnings overstatement
• Meta: +20% earnings overstatement
• Microsoft, Amazon, Google: ~$176 billion in unrecognized capital decay collectively

If accurate, this isn't minor accounting discretion. It's systematically misrepresenting asset values to smooth earnings and maintain stock prices while actual capital decays faster than balance sheets acknowledge.

The comparison Burry uses? Enron's special-purpose vehicles that hid debt and inflated profits before the company's spectacular collapse.

How AI Infrastructure Financing Actually Works

Barchart's John Rowland explains what Burry calls the "daisy chain" financing structure driving AI buildout:

Companies borrow massive amounts to buy AI chips. Private lenders finance this debt outside traditional banking regulation. The collateral is the chips and servers themselves. Those chips depreciate faster than the loans tied to them. New debt finances additional chip purchases. The cycle accelerates.

CoreWeave exemplifies the pattern. The AI infrastructure provider now has interest expenses exceeding operating income. New debt carries 12.3% rates, up from 9.6% months earlier. The business model requires constant borrowing to fund expansion.

Oracle's debt ballooned to $90 billion with a debt-to-equity ratio above 3—the highest among peers and what Rowland calls "a cautionary tale."

The concern: If these firms pause spending or acknowledge true depreciation rates, the infrastructure boom deflates immediately. Revenue projections collapse. Debt obligations remain. The math stops working.

Why This Matters for AI Market Valuations

Burry and short seller Jim Chanos are highlighting four classic bubble indicators:

1. Insider selling
2. Financial distress
3. Massive leverage
4. "Funky" accounting

All four appear present in AI infrastructure spending. Executives are selling stock. Companies are borrowing aggressively. Debt grows faster than revenue. Depreciation schedules don't match asset lifespans.

The market hasn't cared yet. Meta, Microsoft, Amazon, and Google keep posting strong earnings. Stock prices climb. Investors assume AI spending eventually generates returns justifying current valuations.

But if Burry's depreciation math is correct, reported earnings don't reflect economic reality. The profits funding stock buybacks and dividends might not exist. The balance sheet strength looks solid until you adjust depreciation to match actual chip lifespans.

The Counterargument: Why This Might Not Matter

Let's examine why markets might be rationally ignoring Burry's warnings.

First, diversified tech giants have revenue streams beyond AI infrastructure. Microsoft's Office and Azure businesses generate consistent cash flow. Google's advertising empire remains profitable. Amazon's retail and AWS operations fund AI experiments. These aren't pure-play AI bets vulnerable to single-point failure.

Second, accounting depreciation schedules always involve judgment calls. Companies routinely depreciate assets over longer periods than functional lifespans if they maintain residual value. A 3-year-old AI chip might be obsolete for frontier model training but still useful for inference workloads.

Third, revenue growth might actually validate the spending. If AI services generate enough income to cover debt service and replacement cycles, the depreciation timing becomes academic. The assets pay for themselves before becoming worthless.

Fourth, this could be Burry doing what Burry does: Making dramatic contrarian calls that sound prescient when he's right and get forgotten when he's wrong. His track record includes brilliant housing bubble predictions alongside numerous failed shorts and questionable positions.

What AI Infrastructure Spending Reveals About Market Dynamics

Whether Burry's right about fraud, he's definitely right about one thing: AI infrastructure spending is growing faster than revenue, faster than reported earnings, and faster than most depreciation schedules acknowledge.

CoreWeave's interest expenses exceeding operating income signals genuine financial stress. Oracle's debt-to-equity ratio above 3 represents real leverage risk. The entire AI infrastructure buildout depends on future revenue materializing to justify current spending.

That's not necessarily fraud. It might just be really aggressive growth investing with optimistic accounting assumptions. The difference matters legally but looks similar financially if the bets don't pay off.

The deeper question: Are we building AI infrastructure because demand justifies it, or because competitors are building it so everyone must? The latter creates overcapacity and stranded assets regardless of depreciation schedules.

How Investors Should Respond to Burry's Warning

Burry isn't predicting immediate crash. He's identifying structural weaknesses that could matter when market sentiment shifts. The same patterns preceded previous bubbles—doesn't guarantee this bubble pops the same way.

For investors, the implications are straightforward:

Watch debt trends relative to revenue growth. Companies where debt accelerates faster than income face genuine risk regardless of accounting choices.

Differentiate diversified tech giants from pure-play AI infrastructure firms. Microsoft's risk profile differs dramatically from CoreWeave's.

Understand that high stock prices require high future earnings. If depreciation adjustments cut reported profits 20-26%, valuations become harder to justify.

Pay attention when insiders sell aggressively while publicly promoting growth stories.

Maybe Burry's right and AI infrastructure represents another bubble inflated by optimistic accounting. Maybe he's wrong and the revenue growth validates current spending. Either way, the depreciation math deserves scrutiny investors aren't currently providing.

Need help separating genuine AI opportunities from accounting-inflated growth stories? Winsome Marketing focuses on strategies grounded in unit economics, not balance sheet optimism. Let's talk: winsomemarketing.com