Burry's $1.1B AI Puts Are Scion's Last 13F Ever — and He's Already Unwound the December Legs

The reason the timing is the trade is that the three signals came in reverse of how the market read them. On October 27, 2025, Burry wrote to investors saying he would \"liquidate the funds and return capital — but for a small audit/tax holdback — by year's end.\" Only on November 3 did the 13F appear in EDGAR, disclosing 50,000 Palantir put contracts and 10,000 Nvidia put contracts with combined notional value of about $1.1 billion. A week after that, on November 10, the SEC's IAPD database flipped Scion's status to \"terminated.\" Investors who saw the put headline without the closure letter got a fresh bearish bet; investors who saw the closure letter first got a legacy position being wound down.

The Q3 2025 information table, filed by Scion on November 3, 2025 (period of report September 30, 2025), reports total holdings value of $1.381 billion across eight lines. The bearish book is dominated by Palantir puts — a notional of $912.1 million on 5,000,000 underlying shares — with a smaller Nvidia put of $186.58 million on 1,000,000 shares. The bullish book is two large call positions, plus a tight equity sleeve.

On September 9, 2026, Burry posted a \"Trading Post\" on Cassandra Unchained and confirmed via subsequent Yahoo Finance and Seeking Alpha reporting that he had exited every December 2026 Nvidia put and every December 2026 Palantir put. Stated reason: \"rapid time-based decay.\" He added that he is \"happy to sit on\" cash while watching how markets develop through the fall. The only AI-related bearish exposure Burry has publicly retained is the long-dated January 2027 Palantir puts — 50,000 contracts at a $50 strike, for which he paid about $9.2 million in premium — plus 2027 puts on the Invesco QQQ Trust and outright short positions in Oracle, Nebius, Nvidia, Palantir, and the iShares Semiconductor ETF.

In dollar terms the closed December 2026 book was almost the entire original disclosure — roughly $1.1 billion of notional exit between them — while the surviving January 2027 Palantir puts represent only about $9.2 million of premium at risk. The asymmetry is the point: Burry has converted a high-conviction headline position into a low-cost long-tail hedge.

Even after the Burry trim, the AI capex story he was betting against is the dominant fundamental in the data-center supply chain. Nvidia guided Q3 FY2027 revenue to roughly $108 billion on the same August 26, 2026 call. Taiwan Semiconductor reported HPC at 55% of Q4 2025 revenue, up from 40% in 2022, and guided AI-chip revenue to compound at a mid-to-high-fifties percentage rate through 2029 — the same thesis Burry was bearish on has, in Nvidia's own numbers, been pulling forward by at least twelve months.

The bear case has not disappeared — Palantir still trades on a trailing EV/Sales of about 65x and a forward P/E near 148x, while Nvidia trades closer to 23x EV/EBITDA on TTM. But the puts Burry originally wrote were sized for the deal collapsing inside six months. With that window now closed and the December 2026 legs unwound, the only remaining public expression of his view is a small, deep-out-of-the-money Palantir January 2027 position.