Dan Loeb Deleted Semiconductors From His Portfolio — And Put $533M Into a Contested Merger

Most portfolio changes in a 13F are noise — a 4% trim here, a rebalance there. Every so often a filing shows something categorical, and those are the ones worth reading closely. Third Point's Q2 2026 filing, submitted August 14, is one of them.

Dan Loeb didn't trim semiconductors. He deleted them.

The full sweep

Third Point reported roughly $4.68 billion across 43 disclosed positions as of June 30. Among nine complete exits, five were the same trade expressed five ways:

• NVIDIA — approximately $33.1 million, gone
• KLA — approximately $16.2 million, gone
• Lam Research — approximately $16.0 million, gone
• Broadcom — approximately $15.5 million, gone
• VanEck Semiconductor ETF — approximately $15.3 million, gone

Also exited: Meta Platforms (~$51.5 million), SPDR Gold (~$40.9 million), and SharkNinja (~$13.8 million).

The distinction matters. A manager rotating within semis sells NVIDIA and buys Micron, or swaps foundry exposure for equipment. Loeb sold the chips, the equipment makers, and the index fund that would have given him passive exposure to whatever he missed. There is no residual position left to express a view. That is a thesis change, not position management.

It's also worth noting what these positions were not: enormous. Third Point's chip complex totaled under $100 million out of $4.68 billion — roughly 2% of the book. The signal here isn't the dollar weight, it's the completeness. You can track how widely this pattern repeats across managers on the flow page, where sector-level buying and selling gets aggregated across every tracked filer rather than judged one portfolio at a time.

Where the money went

Sixteen new positions replaced them. The largest by a wide margin: Warner Bros. Discovery, 20 million shares, marked at roughly $533 million on June 30 — about 11.4% of the entire disclosed portfolio and Loeb's biggest new disclosed stake.

That works out to an implied quarter-end mark near $26.65 a share. WBD is the middle of a contested corporate breakup: a separation of its Discovery Global networks business targeted for the third quarter of 2026, a pending acquisition of the studio and streaming assets valued around $27.75 per share subject to a collar, and an antitrust challenge from a coalition of state attorneys general with a trial date set for March 2027. Total enterprise value on the transaction runs near $83 billion.

Read the arithmetic and the position looks less like a media bet and more like event-driven arbitrage — a modest spread to a stated deal price, with the return coming from deal completion rather than from anyone watching more HBO. That is Loeb's original discipline. Third Point built its reputation on event-driven and activist situations long before it owned a single hyperscaler.

The rest of the redeployment leaned the same direction — idiosyncratic rather than thematic:

• Alphabet — increased roughly 486% to about 850,000 shares (~$304 million)
• SpaceX — a new private position
• Flex — roughly $167 million
• TTM Technologies — roughly $94.4 million
• Block and Riot Platforms — new stakes

The Alphabet add is the interesting counterweight. Loeb exited the picks-and-shovels layer of AI while dramatically increasing exposure to one of the companies actually monetizing it. Whether that distinction holds up is a separate question, but it's coherent: sell the capex, buy the cash flow. The overlap tool shows how many other Q2 filers landed on the same side of that trade.

The caveat that never goes away

These holdings are a snapshot of June 30, disclosed on August 14 — a 45-day lag, and it's now late August. Loeb has had roughly eight weeks to change his mind, and the semis have moved considerably since the quarter closed. A 13F tells you where a manager stood, never where they stand.

The other structural gap: 13Fs disclose long equity positions only. No shorts, no options overlay in most cases, no credit. A "complete exit" from semiconductors on the long side is consistent with several different views, including a bearish one expressed elsewhere in ways the filing will never show.

What survives those caveats is the shape of the decision. Whole-category exits are rare, they're deliberate, and they're the kind of thing worth checking against the broader institutional picture rather than treating as a standalone call. The macro consensus view aggregates that positioning across managers, and the full roster of tracked filers lives on the investors page.

One manager clearing his chip book is an opinion. Several dozen doing it in the same quarter is a regime change — and only the aggregate can tell you which one this is.

---

Data sourced from public SEC 13F filings. Educational research only — not investment advice.