This week, attention turned once again to Michael Burry, who announced that he is closing his investment fund, Scion Capital. Burry became a prominent figure nearly two decades ago after his early warnings about the fragility of the U.S. mortgage market were first dismissed, then vindicated spectacularly during the subprime crisis. His story later inspired both the book and the film The Big Short, cementing his reputation as an investor willing to challenge consensus. Yet since then, his track record has been uneven. His pessimistic forecasts have periodically captured headlines, such as in the summer of 2022 when he predicted a market crash – only to watch equity indices surge instead.
Fast-forward to 2025, and Burry’s latest decision reflects a belief that equity markets have lost touch with economic reality. On X, he explained that he sees a persistent disconnect between stock valuations and the underlying fundamentals of companies, with a particular focus on technology players. Firms building AI infrastructure – such as Oracle and Meta – are, in his view, overstating profitability by extending the accounting life of their chips. According to Burry, these companies are allocating investment costs over five years despite operating in a sector where components become obsolete far more quickly. The result, he argues, is an artificially inflated picture of performance.
To understand the crux of the disagreement, it helps to revisit the accounting concepts of depreciation and amortization. When a company invests in an asset that will generate revenue over multiple years – for example, a new production machine – it makes little sense to expense the full cost in the year of purchase. Instead, the cost is spread across the asset’s useful life. This is the essence of depreciation. Yet estimating the useful life of a brand-new investment is far from straightforward. The challenge becomes even more pronounced with cutting-edge components like Nvidia chips, which Silicon Valley companies have been ordering in massive quantities. Accountants at Meta and Oracle, like their peers across the industry, have opted to depreciate these assets over a five-year period. Burry, however, believes the real useful life is closer to two or three years at most, arguing that rapid obsolescence should lead to faster cost recognition and therefore lower profits.
This difference in interpretation is anything but trivial. In Burry’s estimation, the profitability gap across the sector could reach $176 billion between 2026 and 2028 – a figure amplified by the extraordinary capital expenditures deployed by technology giants on AI infrastructure over the past three years. Even reading such a number forces us to confront an uncomfortable reality: while accounting is often perceived as precise and objective, it is intrinsically built on assumptions that can meaningfully shift how performance is portrayed.
The purpose of accounting is to provide a quantitative view of an organization’s health and performance. To do this, it relies on three key financial statements. The first, the statement of cash flows, records actual inflows and outflows of cash. Because it reflects hard currency movements, it is largely uncontestable. However, by its nature, it cannot smooth the timing issues associated with multi-year investments or delayed payments, making it an incomplete gauge of performance under typical operating conditions.
The income statement adjusts for these limitations by introducing assumptions that attempt to reflect the business activity of a single accounting period. Depreciation schedules, collectability of receivables, anticipated tax liabilities, and unresolved legal matters all rely on estimates that require judgment. Meanwhile, the balance sheet captures assets and liabilities at a given moment, again requiring assumptions – for example, valuing inventory or ongoing production. In these ambiguous contexts, absolute truth is seldom attainable. Instead, accounting becomes a discipline of constructing plausible scenarios. It is no surprise that some thinkers have gone so far as to question whether accounting might be closer to an art than a science – and there is more than a hint of validity in that perspective.
Difficulties arise when these interpretative layers become so opaque that the true picture is obscured. One of the most infamous cases, Enron, demonstrated the danger of aggressive assumptions paired with complex corporate structures. In the early 2000s, the energy giant collapsed after inflating performance through elaborate accounting maneuvers involving shell companies and questionable valuation choices. The outcome was devastating: roughly $60 billion in losses for creditors, shareholders, and employees.
Against this backdrop, it is challenging to take a definitive stance on Burry’s latest warning. Only time will reveal whether the massive investments in AI will yield the returns expected – and whether those returns will materialize quickly enough to justify current valuations. At present, business models in the AI sector remain early-stage, and the path to sustained profitability is still uncertain. Should these investments fall short of expectations, the adjustment could be significant.
For now, Burry’s withdrawal serves as a reminder that markets often price in optimism long before cash flows catch up, particularly during periods of technological transformation. Whether his caution proves prescient or premature, the debate underscores a deeper truth: numbers may tell a story, but the assumptions behind them shape the narrative.