Michael Burry's New Stock Picks (September 2026) — and How They Are Doing

Michael Burry spent the summer of 2026 betting against the market’s biggest winners. In the middle of August he took some of it off — and in the second half of the month he put much of it back on. On 13 August he covered his Tesla and Applied Materials shorts outright, trimmed Caterpillar by a quarter, and pushed what was left further into AI infrastructure: Oracle, Micron, Nebius, Nvidia and Palantir, alongside a short of the iShares Semiconductor ETF he calls his largest bearish position and a QQQ put position he rolled out to June 2027. Cash went to about 12%.

Most of those have a disclosed entry price, which means most of them have a scoreboard. Below is each position, what he paid, where it stands at the 8 September close, and the part that matters most for how much weight to give any of it: where this information actually comes from. It is not investment advice.

What are Michael Burry’s newest stock picks?

Eight bearish positions and at least a dozen long ones are on the public record, disclosed between 30 June and 1 September 2026. The table carries the ones with a reported entry price. The QQQ puts, the semiconductor-ETF puts and three of the longs — Zoetis, MercadoLibre and Adobe, named only when he trimmed them — do not have one, so they are not scored.

The book turned over in a week, and mostly against him. On 1 September this table had four shorts in profit and Micron flat. Eight sessions later five of the six shorts are under water — Nebius, Oracle, Micron and the Nvidia puts all moved against him, and only Caterpillar and the semiconductor ETF are still working. Palantir flipped the other way, from 2.8% down to 2.7% up.

Nebius is the swing. The page’s previous update flagged it as the position most likely to be misstated, and it was: printed at 5.8% in profit, it is now about 15% against him after four consecutive rising sessions.

⚠️ DraftKings has been removed from this table. He sold the position in full on 5 August 2026 — a month before this page was last updated — so it should not have been listed as live in the 19 August or 2 September versions. That was our error, not a change in his book.

⚠️ Two positions cannot be scored and are shown anyway. No entry price has been published for Lululemon or the CoreWeave short. Leaving Lululemon out entirely, as this page previously did, was the bigger distortion: it is reportedly 17.4% of his portfolio, his largest single holding.

Each entry above is the first disclosed price for that position. Where he has added since — and he has, repeatedly — the average is better or worse than the number in the table, and no average has been published.

And two he closed

No exit price was reported for either, so neither has a final number and this page does not invent one. His stated reason is worth having in his own words: “Quick sizable short sale gains are gift horses in this market.”

What is Michael Burry shorting?

Five names opened on 30 June after what he saw as overextended rallies — then added to, rearranged and in two cases closed over the six weeks since.

The 30 June batch was Caterpillar, Nvidia, the iShares Semiconductor ETF, Applied Materials and Tesla. On Caterpillar he was explicit that this was out of character: “I have never shorted Caterpillar. It has always done great for me on the long side in the past.” Caterpillar is still on, though a quarter smaller than it was: he cut it by 25% on 13 August, having added at $893.49 on 24 July and again at $844 on 11 August.

Tesla was the best of them and is now closed. He shorted at $416.22, the stock fell about 24% through July including a 15% single-day drop after quarterly earnings, and on 13 August he covered. Applied Materials went the same day. Neither exit price has been reported.

What is left is a bet on one industry rather than on the market’s winners generally:

⚠️ One entry price is disputed and the disagreement is left visible. The 24 July reporting put the Micron short at $933.86; Benzinga’s account of the 13 August post says he “first short[ed] the stock near $1,052 in July”. Both can be true if the position opened before 24 July and $933.86 was an addition, but nobody has published that, so the table keeps the price that was reported at the time and this paragraph records the conflict.

His reasoning on Nvidia is the most quotable thing he has said all summer, and it is a demand argument rather than a valuation one: “I believe much of current and future demand is not driven by end customers, end demand. Much and possibly most is financed, off-balance sheet and not lit. Future revenues are majority financed in a circular arrangement, per the 2026 BIS annual report.”

By 13 August that had hardened into a timetable. He argues that debt is a growing share of the capital being injected into AI, which “puts the bubble on a clock”, and that 2028 is his base case for when compute becomes too much — while noting that markets tend to react well before such a turn. On the shorts themselves he was blunter: they were, he said, “a bit like shooting fish in a barrel.”

The Nvidia puts are still losing money, and by more than before. The stock closed at $225.73 on 8 September, against the $198.09 where the puts were opened — about 14% against him on the same arithmetic this page uses throughout, up from 10% a week earlier. His exposure there is options, not stock: puts expiring December 2026 and June 2027 with strikes in the low $100s. On 27 August he complicated that picture by buying Nvidia calls, described as a hedge, and reportedly selling roughly half of them within twenty-four hours; by 31 August he was reported adding to the Nvidia and QQQ puts again.

Nebius has swung back against him, hard. He first shorted it at $211.77 and added around $247. It closed at $243.88 on 8 September, after rising in every session since the 1st — 199.54, 204.09, 210.63, 226.39, 243.88 — which puts the short roughly 15% against its first disclosed entry rather than 5.8% in front. The last leg came on 8 September, when Palantir named Nebius its preferred sovereign-AI infrastructure partner and the stock rose 7.7% on the day.

His argument is about what the company’s own pricing implies rather than about its results, which were good — revenue up 454% — and it is precisely the results he reads as a top signal: customers paying extraordinary prices for scarce capacity now, while longer-dated compute is priced far lower. Nothing in the last week speaks to that argument either way. What has changed is the mark.

A Palantir short has run alongside these since well before the summer. CNBC reported in April 2026 that he was betting against it; on 4 August it jumped 29.5% in a single session, from $125.65 to $162.66, after second-quarter results. On 11 August he shorted the stock again, this time at a disclosed $175, and added to puts expiring December 2026 and March 2027 with strikes in the low $100s — reporting of 19 August put the March 2027 strike at $100, bought with the stock at $175.23. It closed at $170.30 on 8 September, down 2.3% on the day, which puts the 11 August short about 2.7% in profit — a sign flip from a week ago, and the only short on the page that moved his way. He was reported adding to it again in the last week of August.

What has he done since the start of August?

He posted trades on almost every trading day through mid-August, and then the cadence collapsed.

⚠️ This page previously said “most trading days”, and that is no longer true. The public archive shows posts on 4, 5, 6, 7, 10, 11, 12 and 13 August, then one covering 18–20 August, then 26, 27 and 28 August, and then a single “Trading Post 9/4/2026”. There is nothing dated after 4 September, so no new trade has been disclosed across the 8 and 9 September sessions — including the week in which his largest position hit an eight-year low and four of his shorts moved against him.

All the posts themselves are subscriber-only. Only a handful have been reported in enough detail to log here with a price, and the gaps matter as much as the entries — many days of trades exist that nobody outside his subscriber list has seen.

4 August — “My Options.” He closed a Microsoft long, an Oracle short and a Palantir position, rolled the Nvidia puts out to June 2027 and a QQQ short to February 2027. The post went out roughly two hours before the close of the session in which Palantir rose 29.5%. Rolling options out by months costs premium and buys time — it is what someone does who thinks they are right about the destination and wrong about the schedule. He also published a market view the same day, that it is “possible we are near a major top, and possible a 1987-type fall”, on the day the S&P 500 hit its first record close since June. We have taken that argument apart separately in Michael Burry’s 2026 market warning .

11 August — the shorts go back on, including Oracle. He shorted Palantir stock at $175, added to Caterpillar at $844, Oracle at $145 and the semiconductor ETF at $533, and bought more Molina Healthcare at $198, saying it was “now more on par with my largest positions again”. The Oracle line is the one worth noticing: he had closed that short a week earlier and put it back on lower.

13 August — the reshuffle. He covered Tesla and Applied Materials outright, cut Caterpillar by 25%, and took cash to about 12%. Against that he added to Nebius around $247, Micron around $924 and Oracle around $152, and rolled his QQQ puts from January 2027 to June 2027, moving the strikes up from the mid-to-high $500s to the mid-to-high $600s and taking the position to about 6% of the portfolio. In his words the goal was “to reduce gross exposure, and to free up some cash, while maintaining the short bias”. The semiconductor-ETF short, at roughly 7%, is what he describes as his largest bearish position.

⚠️ Two reports of that day disagree about the SOXX put options, and neither is followed here. One account says he closed them as a losing trade and swapped the exposure into the larger QQQ position; another says he added to March 2027 puts on the same ETF. Both were published on 14 August by named outlets. What is not in dispute is the short in the shares, which every account keeps as his biggest bearish position, so that is what this page reports.

What he said about the size of it. Asked in effect how aggressive this is, he put it against his own history: “I am now carrying a similar short load, maybe even more aggressive in some way than 2020.” He also named a stock he would not short despite sharing the structural concern — CoreWeave, which he called “the leading candidate for memesville” — on the grounds that retail momentum is its own risk.

He then shorted it anyway. A CoreWeave short was opened later in August and disclosed in the trading post covering 18–20 August. No entry price has been published, so it cannot be scored; CRWV rose 17.4% across August and closed $99.83 on 8 September, up 11.7% on the day. ⚠️ This rests on press reporting of the paid newsletter rather than any filing.

So the honest state of play: the priced entries and exits on this page run to 13 August, but the trading did not stop there. Between 19 August and 4 September he exited Alibaba into JD.com, opened a CoreWeave short, bought Nvidia calls as a stated hedge — describing the premium as offset by his existing short, and saying he was “not playing for gains here” — added again to the Oracle, Palantir and Nebius shorts and to the Nvidia and QQQ puts, trimmed Veeva after a near-100% gain and added to PayPal. Almost none of that has a published entry price, which is why it is described here rather than scored. The 30 July batch has never been reported, several August posts have not been either, and the July posts are titled by trade count — “4 Shorts, 3 Longs” — so more positions exist than have been named anywhere.

What is Michael Burry buying?

A sports-betting pair, a much bigger Chinese e-commerce position since he dumped Alibaba, a health insurer — and a long book that has turned out to be wider than the summer’s reporting showed.

Lululemon, which this page should have led with

🔴 His largest disclosed position is Lululemon, at a reported 17.4% of the book, and until now it appeared nowhere on this page. That is a significant omission and it is corrected here: on any reasonable reading it is the most important single thing in his portfolio, roughly twice the weight of the semiconductor short that this page has been treating as the headline.

It is also where the damage is. On 4 September 2026 LULU fell 17.4% in one session to close $100.61, touching an intraday low of $97.99 — an eight-year low — after second-quarter revenue of $2.42bn missed, comparable sales fell 9%, and full-year revenue guidance was cut to $10.35–10.50bn from $11.00–11.15bn. It closed $103.19 on 8 September and is down more than 52% so far in 2026.

He addressed it directly, after hours on 3 September:

“Today, lululemon (LULU) is the trickster in my portfolio. This time the trickster is my largest position, and it does seem determined to take me where mermaids fear to tread.”

He added that he continues to hold it, and that he would buy more if it traded under $100 the following morning. It did trade under $100 — and no report confirms he actually bought. That is where the record stops, and this page is not going to fill the gap.

⚠️ No entry price for Lululemon has ever been published, so unlike every other row in the table above, there is no way to say whether the position is up or down overall — only that it has had a very bad week.

On 8 July he disclosed a full-sized position split roughly 60% Flutter Entertainment at about $107 and 40% DraftKings in the low $26s. The thesis is regulatory rather than operational: prediction markets, he argues, currently sit in a loophole beside a heavily taxed industry, and that will not last. “I believe that the political climate will not tolerate this. Prediction markets exist in a loophole adjacent to a heavily regulated and taxed industry. In time, prediction markets will be subsumed into regulation and taxation.”

On the businesses themselves he described DraftKings as “inflecting as an operating business” — approaching the point where profit compounds faster than revenue — and Flutter as “a fundamentally very good operating business with terrific scale” held back by past capital misallocation now washing through.

🔴 The pair is no longer a pair, and this page was a month late in saying so. On 5 August 2026, ahead of DraftKings’ second-quarter report, he sold the DraftKings position in full and more than doubled Flutter, taking its average cost down “into the $98s”. He called Flutter a “fat pitch” after the Q2 selloff. Flutter is now his only sports-betting holding.

That trade was reported on 6 August and restated in a 31 August review, so it predates both the 19 August and 2 September versions of this page, which continued to list DraftKings as live. It is corrected here.

Flutter closed at $100.70 on 8 September. Against the original ~$107 entry that is about 6% down; against the “into the $98s” average he has actually paid, the position is marginally in front. The table above uses the first disclosed entry, as it does for every row, so it shows the worse of the two numbers.

⚠️ The DraftKings sale and the average-cost figure both come from financial-press reporting of a paid newsletter. No filing exists to check either against — the same second-hand chain described at the end of this page.

In the same post he added to JD.com at $27.58, describing it as one of his top three core holdings. That one is still in front, but only just: JD closed at $27.68 on 8 September, about 0.4% higher, having been up 19.5% in early August. It has also got much bigger. Around 23 August he sold his entire Alibaba position — bought only in April — after Alibaba announced a share sale of roughly $10 billion, and moved the proceeds into JD.com. His stated objection was governance rather than price: management had pointed to a large net cash pile days before issuing equity, and he said he “cannot bless” share issuances.

The newest long is Molina Healthcare, added at $198 on 11 August. He described it as a full position that had shrunk in relative terms as he moved capital around, and said it was “now more on par with my largest positions again” — a long-term belief in the company, with the political season as the near-term reason to size it up. It closed at $196.30 on 8 September, down 2.2% on the day, which puts it about 0.9% under his entry rather than in profit. Later reporting put the position at roughly 9% of his book.

On 13 August he went the other way on the rest of the long book, trimming Zoetis, MercadoLibre, JD.com and Adobe to help take cash to about 12%. No trim prices have been reported, which is why the JD.com entry above is still the 8 July one.

Where does this information come from?

His paid newsletter, reported second-hand by the financial press — not from any regulatory filing.

This is the part that separates a Burry roundup from a Cathie Wood one, where ARK publishes its trades every day . There is no equivalent document here:

• No 13F. Scion Asset Management’s last SEC filing of any kind was on 3 November 2025, and its registration was terminated the following week. There has been nothing in 2026 under any entity.
• The trades appear in a subscription newsletter. Its July trading posts are titled by count — “4 Shorts, 3 Longs” on 24 July, “3 Stock Buys” on 23 July — and August brought eight more, most of them unreported, so more positions exist than have been named.
• What reaches the public does so through journalism. CNBC, Benzinga, Yahoo Finance and others report the disclosures, which is why specific entry prices are quotable at all.
• His free posts are public. He also writes short public notes, which is where the reasoning often shows up before the trade does.

So every number in this article is a report of a disclosure, not a disclosure. That is a meaningful step removed, and it cuts both ways: the reporting is from outlets that name their source, but nobody outside his subscriber list can audit the full position list.

The catch: how much weight should this carry?

Three things to hold in mind before treating any of it as a signal.

The “$1.1 billion bet” you have read about is from November 2025. It is the most-repeated Burry number in circulation and it is worth knowing exactly what it is. His last 13F, for the quarter ended 30 September 2025, showed put options on roughly 5 million Palantir shares and 1 million Nvidia shares, with notional values of $912 million and $187 million — the $1.1 billion. As the reporting at the time was careful to note, a 13F does not disclose an option’s strike price, its expiry, or what was paid for it. The notional is the value of the underlying shares, not money at risk. Anyone quoting that figure today is quoting a filing that is now ten months old and covers a quarter that ended almost a year ago, from a fund that has since stopped filing.

A short and a put are not the same bet. Burry’s Nvidia exposure is reported as put options, while Tesla, Caterpillar, Micron and the rest are short sales. A short sale has theoretically unlimited downside if the stock keeps rising; a put costs a premium and that premium is the most you can lose. Write-ups that call them all “shorts” are describing a risk profile he may not have.

The list you can see is incomplete by construction. Titles like “4 Shorts, 3 Longs” tell you seven trades happened on 24 July. Public reporting names a handful. Since August it is worse: his archive lists trading posts on most trading days of the last five weeks, and only a handful of them carry a reported entry price. Anyone publishing a complete current portfolio is filling the gap with guesswork.

His timing has historically been early. The housing bet that made his name was placed well before it paid, and several of his later calls moved against him for long stretches first. A position that is 25% in profit five weeks in is not the same as a thesis being proved.

This is a roundup and an explainer, not investment advice — Drawpie isn’t a financial adviser, and nothing here is a recommendation to buy, sell, hold or short anything. Short selling carries unlimited theoretical risk, and options can expire worthless. Every position described is a report of a past disclosure and may already have been closed. Do your own research and consider a licensed professional before acting on anything here.