two questions, what are your thoughts on michael burry in general and his investment thoughts. when will there be an actual market crash. everyone always talks about it but it never happens?

Major declines happen; recoveries change how we remember them

The feeling that everyone predicts a crash and nothing happens is understandable. It also overlooks substantial recent losses. Using daily closing prices, the S&P 500 fell about 34% during the pandemic, 25% in 2022 and 19% in the February–April 2025 selloff. Those declines are calculated from each episode’s closing peak and trough. [2][3]

The vocabulary can obscure the experience. A correction conventionally begins at a 10% decline and a bear market at 20%; a crash describes speed and disorder as well as depth. The pandemic decline took roughly 23 trading sessions. The dot-com bear market unfolded over more than two years. Both inflicted serious losses, through very different paths. [4][2]

Source: Yahoo Finance historical S&P 500 closing prices. [4][2][3][5]

Method: trough close divided by peak close, minus one; dividends excluded. Episodes use their respective closing peaks and troughs.

Recovery makes a difference to memory. The S&P 500 regained its pre-pandemic closing high in August 2020, and its February 2025 high in June 2025. The losses were real even though the index subsequently recovered. An investor forced to sell near a trough experienced a different outcome from one able to wait. [2][3]

Waiting can also be painfully long. The closing-price highs preceding the dot-com and financial-crisis declines were recovered only in 2007 and 2013, respectively. We would not use the speed of recent recoveries to dismiss the risk of a prolonged bear market. [4][5][6]

That history establishes the risk investors must accommodate. It leaves the timing problem intact—and that is where Burry’s record needs to be read carefully.

Burry earned his reputation through analysis; timing requires a separate case

The original achievement was substantial. Michael Lewis reports that an investor who stayed with Scion from November 1, 2000 through June 30, 2008 gained 489.34% after fees and expenses. Burry’s mortgage trade combined an investigation of weak loan collateral with an instrument that could profit from its deterioration. Our respect rests on that analytical and investment achievement. The return figure comes from Lewis’s account. [7]

His GameStop investment supplies a useful counterweight to the permanent-bear caricature. In correspondence reproduced in an SEC filing, Scion described a holding of approximately two million shares and urged the company to deploy its remaining $237.6 million buyback authorization while the stock traded below $5. Burry pointed to short interest equivalent to roughly 74% of shares outstanding and a coming console cycle. [8]

“We expect GameStop’s business will perk up during 2020 and 2021 as the new console cycle finally gets underway,” the letter said. That is a recognizable security thesis: valuation, capital allocation and a business catalyst. The letter gives us the reasoning; it supplies no complete entry-to-exit return with which to credit him for the subsequent squeeze. [8]

Public market instructions are harder to use. In his own retrospective, Burry recalled tweeting “Sell” on January 31, 2023, followed by “I was wrong to say sell” on March 30. Between those statements, he reassessed the banking disturbance and said he was “not seeing true danger here.” [9]

We regard the willingness to reverse course positively. We are skeptical of treating the original instruction as an investment system: a follower needs a position size, an exit rule and a way back in. A memorable success cannot supply those rules, and a single failed call cannot settle the question of skill.

There is another complication. Even when investors can see a position, they can misunderstand what it cost.

A $912 million filing value became a $9.2 million premium

Scion’s September 30, 2025 holdings table reported a Palantir put position with an underlying-share value of $912.1 million. Burry later described the purchase as 50,000 contracts, each covering 100 shares, bought for $1.84 per share. Multiplying those terms gives an initial premium of $9.2 million. [10][11]

The two measures differ by about 99 times. One describes the shares underlying the options at the reporting date; the other describes his stated purchase cost. The distinction changes how a reader should interpret the size of the trade. Its eventual outcome also depends on terms, timing and the rest of the portfolio. [10][11]

The dates matter as much as the units. The filing was signed on November 3 and described holdings on September 30. It is a historical snapshot. Scion’s adviser registration subsequently terminated in November 2025, while Reuters reported plans to return outside capital. Those developments leave Burry’s personal investing activity a separate question. [12][13][14]

We think these distinctions materially weaken the usefulness of copying a headline. They leave the underlying investment argument available for scrutiny. In Burry’s current case, that argument concerns the economics of AI spending.

The AI concern is credible because the buyer must eventually earn a return

On September 28, 2026, Burry wrote in the accessible preview of his trading post: “Fundamentally, I am moving timelines up.” CNBC separately quoted him seeking more leverage in short positions through put options. That is a meaningful change in his expressed conviction. [15][16]

The strongest economic version of the concern is straightforward. Infrastructure buyers spend today against expectations of future customer revenue. If the resulting capacity earns too little, boards can reduce spending. Suppliers then face weaker orders, earnings expectations fall, and investors may also pay less for each dollar of earnings. Financing pressure can accelerate that adjustment.

Oracle shows why the concern deserves attention. In the quarter ended August 31, 2026, capital expenditure reached $28.5 billion, compared with $8.5 billion a year earlier. Cloud revenue rose to $11.6 billion from $7.2 billion. Calculated from those comparative filing figures, capex grew 235% and cloud revenue 62%. [17]

Source: Oracle fiscal first-quarter 2027 Form 10-Q. [17]

Note: comparable fiscal first quarters; dollars in billions. Capital expenditure purchases assets serving future periods, while cloud revenue covers each quarter.

Our concern is the future earnings and financing burden created by that acceleration. Oracle added roughly $20 billion of quarterly capex while quarterly cloud revenue increased by $4.4 billion. The investment needs time to mature, so these figures cannot settle its eventual return. They make the burden visible. [17]

The supplier’s current economics look very different. Nvidia reported $96.2 billion of fiscal second-quarter 2027 revenue, up 106% year over year, and a 75% GAAP gross margin. Data Center revenue reached $89 billion, up 117%. Money is reaching the infrastructure supplier in considerable volume. [18]

Palantir also reported vigorous operating results: second-quarter 2026 revenue of $1.94 billion, up 93%, and GAAP operating income of $912 million, equivalent to a 47% margin. Those results sit uneasily with any claim that the operating businesses have already collapsed. [19]

The evidence is therefore split in an economically coherent way. Buyers are committing capital at an exceptional pace; suppliers and software businesses are reporting strong sales and profits. We take the investment-return concern seriously while remaining skeptical of the bearish timing. Strong supplier profits can coexist with poor eventual returns for buyers, just as strong businesses can carry excessive stock valuations.

The decisive transition would be from concern about future returns to actual spending cuts, cancellations, financing impairments and weaker guidance. For a broad-market crash assessment, we also need to look beyond the AI companies.

The current baseline falls short of an imminent-crash case

Broad-market valuation offers less support for an immediate crash call than the AI debate might suggest. FactSet’s September 25 report put the S&P 500 at 19.2 times forward earnings, against a five-year average of 19.8 and a ten-year average of 19.0. On that measure, the index was close to its recent historical range. [1]

The vulnerability sits partly in the denominator. FactSet expected third-quarter earnings growth of 29.1% and approximately 32% growth for calendar 2026. Those are substantial expectations. A market priced reasonably against strong forecasts can become expensive quickly when those forecasts fall. [1]

Credit provides an independent check. On September 30, the high-yield option-adjusted spread was 3.12 percentage points, and the investment-grade corporate spread was 0.84 points. We read those observations as inconsistent with an active broad financing seizure, while recognizing that credit conditions can change abruptly. [20][21]

The labor picture is mixed. September payrolls rose by only 29,000, with combined downward revisions of 60,000 to the preceding two months; unemployment was 4.2%. Yet initial unemployment claims were 197,000 in the week ended September 26. Weak hiring and a wave of layoffs are different stages of deterioration. The evidence points more clearly to the former. [22][23]

Our judgment is therefore vigilance rather than an imminent-crash forecast. Burry’s increased conviction is insufficient reason, by itself, to exit broad equities. That judgment carries considerably more confidence than any prediction we could make about the market’s next twelve months.

The arithmetic shows how the view could break. An index price reflects expected earnings multiplied by the valuation investors assign them. A recession or financing shock can damage both at once.

Lower earnings and a lower valuation can jointly produce a large decline without either adjustment being extreme.

Source: our calculations using FactSet’s September 25 starting multiple of 19.2×. [1]

Method: ending P/E = 19.2 × (1 − price decline) ÷ (1 − earnings reduction). Scenarios measure changes from a fixed starting valuation and forward-EPS estimate.

A 10% reduction in expected earnings takes the multiple required for a 20% decline to about 17.1 times. A 30% decline under the same earnings assumption requires about 14.9 times. The table makes a substantial fall easy to envisage. It cannot tell us when expectations will change. [1]

Being right about the eventual fall can still make waiting expensive

A warning has to survive the path between today and the eventual decline. Companies can grow their earnings, policy can change, and prices can rise far enough that a later crash leaves them above the level at which the warning began.

Alan Greenspan’s December 1996 question about “irrational exuberance” captures the problem. From the S&P 500’s 744.38 close on the speech date, the index rose to 1,527.46 at its March 2000 peak—a gain of about 105%. Even the subsequent October 2002 trough of 776.76 stood about 4.3% above that starting price. These endpoint calculations exclude dividends and interest earned on cash. [24][25][4]

The simple arithmetic is enough to show why direction and timing differ. In a hypothetical path, a market rising from 100 to 150 and then falling 30% finishes at 105. A rise from 100 to 125 followed by the same decline finishes at 87.5. A 42.9% advance exactly offsets a subsequent 30% fall on price alone. An actionable waiting strategy needs a view on both legs, plus a re-entry rule.

Policy also changes the outcome that a warning extrapolates. The April 2025 tariff pause arrived immediately after the market’s closing trough. In March 2020, the Federal Reserve announced purchases of Treasury and agency mortgage-backed securities “in the amounts needed” to support market functioning. Those interventions show how the response to stress can alter its course. [26][27][2][3]

There is no unconditional rescue. In June 2022, with inflation too high, the Fed increased its policy rate by 75 basis points during an equity bear market. Investors must allow for periods when the policy objective conflicts with supporting asset prices. [28]

Our practical conclusion is to make exposure survivable: keep money needed soon outside equities, and address excessive concentration or leverage without requiring a forecast date. The right allocation depends on the investor’s obligations and ability to endure losses. A long horizon helps only if the investor can actually remain invested.