Burry exits Microsoft and pushes the AI shorts out to 2027
Rolling a put out a further six months is not the same trade as buying one. It says the thesis needs more time, and that the position is sized to survive being early.
Michael Burry disclosed a series of portfolio changes on August 4. The direction of travel is unchanged: still bearish on the most expensive and most capital-intensive parts of the AI and semiconductor complex, while trimming or closing individual positions that had done their job.
The clearest signal is not any single exit. It is the two rolls.
What actually changed
- Fully exited the Microsoft long/call position.
- Closed the Oracle put after taking profits.
- Sold the nearer-term Palantir puts without rolling them, while keeping a stock short in PLTR.
- Rolled part of the Nvidia put position from December 2026 into June 2027, strikes still in the low-$100s.
- Rolled a set of QQQ puts into February 2027 at a small credit, strikes still in the high-$500s.
- Stock shorts remain open in SOXX, Micron, Nvidia, Caterpillar, Palantir, Tesla and Applied Materials.
Long positions were otherwise left unchanged. Earlier in the period Burry had been adding to non-AI names such as DraftKings, Zoetis and lululemon, on the argument that they had de-risked after multi-year declines.
Why the roll dates matter more than the exits
Closing the Microsoft long after a strong run and banking the Oracle put are ordinary risk management. Position-level decisions, not thesis changes.
The rolls are the thesis. Pushing Nvidia puts to June 2027 and QQQ puts to February 2027 extends the bearish view by two to three quarters rather than pressing it into near-term volatility. A trader who thought the break was imminent would hold the December expiry and let it work. Buying more time is what you do when you believe the direction and not the timing, and it costs something: the QQQ roll went out at a small credit, which is close to the cheapest version of that decision available.
Selling the near-dated Palantir puts while keeping the stock short points the same way. It takes the theta bleed off the table and keeps the exposure.
The Caterpillar short is the interesting one
Caterpillar sits in the short book. In the same week, Caterpillar reported record revenue, a record backlog and raised full-year guidance, with management naming data-center construction as a driver. That is the exact real-economy channel the AI build-out runs through, and it went the other way.
That tension is worth watching rather than resolving. A short book built on the view that AI capital spending is unsustainable will, by construction, be uncomfortable while the spending is still landing in someone's order book. The question the position is asking is whether it keeps landing.
What to watch next
Burry has flagged upcoming analysis of hyperscaler earnings quality and accounting. That is the argument underneath the positioning: not that AI demand is fake, but that the reported economics of financing it may not survive close reading. The 2027 expirations give that argument room to be tested.
Position disclosures describe what someone else did. They are not a recommendation, and a disclosed position can be closed the day after it is filed. Educational content, not financial advice.