Being busted at a market top is not a defensive trait. It is a diagnosis that can point in two opposite directions. Michael Burry's latest Foundations note on Cassandra Unchained sorts S&P 500 members at the October 6, 2026 record high into three buckets: Leaders within 10% of their 10-year highs, the Broken at 20% to 40% off their tops, and the Busted at more than 40% off. He then compares that snapshot with the index at earlier record highs that preceded bear markets and near-bear markets over four decades. His earlier "Bear Market Heroes" work suggested that if this is 2000 again, the broken names could serve as a haven. Splitting out the Busted asks a sharper question. Does the deepest discount offer the deepest shelter, or does it mark the stocks where the next bear market is already under way?
The mechanism that decides it is simple to state. A stock busted at an index record has already been repriced for something. Through the next bear market, its fate turns on whether that something was a multiple that had run too far or an earnings base about to break. In the first case the stock has already taken its bear market. In the second case it is early in its own. The buckets cannot tell the difference. Only the cause of the coming decline can.
The subscriber-facing results tables sit behind the paywall and are not reproduced here. What follows is a reading of the framework and of the historical episodes it necessarily draws on, not a restatement of Burry's numbers.
Why the Y2K Top Flatters the Busted
The 2000 analogue is the strongest case for the Busted bucket, and it is strong for a specific reason. The Nasdaq Composite closed at 5048.62 on March 10, 2000. The S&P 500 set its high on March 24. By October 2002 the Nasdaq had lost roughly 78% and the S&P 500 roughly half. Yet small-cap value posted gains in both 2000 and 2001, and the equal-weighted S&P 500 held up far better than the cap-weighted version.
What made that possible was the location of the excess. The 2000 bear market was a compression of the multiples paid for technology, media and telecom leaders. Many of the stocks that were down 40% or more at the March top were old-economy companies: insurers, industrials, retailers, REITs, tobacco. Investors had abandoned them because they were boring, not because their cash flows were failing. Capital that left the leaders had nowhere to go but toward the cheap stocks. Those stocks had little multiple left to lose and earnings that a mild 2001 recession dented but did not destroy.
In that setting, being busted meant that the de-rating had already happened. The bear market was a phenomenon happening to the Leaders, and the Busted stood outside the blast radius. The deeper the prior discount, the larger the cushion when relative flows reversed.
When Busted Means First Casualty
The October 2007 top inverts the picture. At that S&P 500 high, the stocks deepest under water were disproportionately homebuilders, mortgage lenders, subprime-exposed financials and consumer credit names. They were not orphans. They were the leading edge. Their prices had already fallen because the market had correctly begun discounting a housing and credit loss that had barely started to hit reported earnings. Over the next 17 months many of them fell much further, and several were acquired under duress, nationalized or wiped out.
The same logic, in a milder form, fits the January 2022 top. The busted cohort then was long-duration speculative growth: unprofitable software, recent IPOs, the names that had peaked in early 2021. The 2022 decline was a discount-rate shock, and those stocks were the purest expression of the asset being repriced. Being 40% down in January 2022 meant being halfway through a move that had further to run.
So the Busted bucket behaves differently in the two kinds of episodes:
- Bear market caused by the Leaders' valuations (2000): the Busted already carry the de-rating, and relative flows favor them.
- Bear market caused by a fundamental break inside the Busted group itself (2007, and in a different way 2022): the Busted are the signal, not the shelter.
A historical table that averages these episodes together will understate how bimodal the outcome is. The median Busted stock across four decades of tops may look respectable. That average says little about which regime the October 2026 top belongs to.
The Index Membership Problem
There is a quieter mechanical issue in any study of this kind, and it pushes in a flattering direction. A stock down 40% or more at an index record is a candidate for deletion. If the subsequent performance of the Busted is measured using the membership list on the day of the top, and every name is held to the end, the study captures the full loss on the ones that fail. If instead the bucket quietly loses names as they are deleted, or if delisted stocks drop out of the price series, the survivors flatter the result.
The construction of Burry's tables is not visible in the free portion of the note, so this is a caveat about the method class rather than a criticism of his execution. It matters most in a 2007-style regime, where the deletions are concentrated in exactly the bucket being measured. In a 2000-style regime the issue largely fades, because the busted old-economy names mostly survived.
Size interacts with this. A stock that has lost 40% of its value within the S&P 500 has usually slid toward the bottom of the index by market capitalization. The Busted bucket is therefore partly a small-cap and mid-cap bucket in disguise. This is the bridge to the "time for small caps?" question in Burry's title. A Busted-heavy long book is a size tilt as well as a value tilt. In 2000 to 2002 both tilts paid. In 2007 to 2009 small caps did not provide shelter.
The chart above frames the setup in one line. The gap between equal-weight and cap-weight performance measures how much of the October 6 record rests on index heavyweights rather than on the median member. A wide gap is a necessary condition for the 2000 template: excess concentrated in the Leaders, with value stranded elsewhere. It is not sufficient. The 2007 top also arrived with plenty of laggards, and those laggards were the problem.
What Would Break the Haven Case Now
The case for the Busted as bear-market heroes in 2026 rests on Burry's framing of the current cycle. His prior Short Thoughts piece described an "AI Cloud Oligopoly" and a flashing bear-market indicator. If the coming decline is a de-rating of a handful of AI infrastructure and cloud leaders whose multiples assume durable capital spending returns, the 2000 template applies cleanly. The Busted would be the stocks the AI trade starved of capital, and they would be positioned to receive it back.
The countercase is that some of today's busted stocks may be busted because of the AI cycle rather than in spite of it. This link is unverified, since the constituent lists are not visible, so treat it as an inference to test rather than a finding. Companies whose business models AI is expected to disrupt, such as seat-based software, outsourced services, staffing, and certain media and education names, would show up as deep losers at an index high. If AI actually erodes their revenue, they are 2007-style casualties whose discount reflects real earnings risk. If the AI trade deflates because the disruption arrives more slowly than priced, they could rebound hard. The bucket holds both possibilities at once.
A second channel would also weaken the haven case: credit. A busted stock with a heavy maturity schedule is a different instrument from a busted stock with net cash. In 2000 the leverage sat largely in telecom carriers and their vendors, inside the Leaders' ecosystem. If 2026's leverage sits in the cheap end of the index, which is where years of buybacks funded by debt and private-equity-style capital structures tend to accumulate, the bear market would find the Busted first.
The Variable That Sorts the Bucket
The useful output of Burry's split is not a verdict that busted stocks outperform. It is a sorting question applied name by name. For each stock down more than 40% at the October 6 high, the test is whether the drawdown came from multiple compression on stable cash flow or from a forecast decline in the cash flow itself. Stocks in the first group carry the 2000 cushion. Stocks in the second group carry 2007 risk, and a bucket average conceals the difference.
The market will reveal which regime applies through earnings, not prices. If the Busted cohort's forward estimates hold steady while the Leaders' multiples compress, the Y2K analogue is working and the discount is a cushion. If the Busted cohort's estimates start falling before the Leaders' do, the bear market has begun where the prices already said it had, and the deepest discounts will keep getting deeper.
