Third Point · Dan Loeb
Latest 13F holdings with ERM3 risk decomposition — how much of the portfolio's return is market beta versus genuine stock-picking, and where the residual risk concentrates.
- Reported AUM
- $4.0B
- Holdings
- 31
- Top-10 weight
- 67%
- As of
- 2026-06-30
- Filed
- 2026-08-14
Return attribution — beta or stock-picking?
Sum of monthly ERM3 attribution over the trailing 12 months (2025-07-31 → 2026-06-30), modeled subset of holdings. Stock selection (residual return) detracted about 44% of the gross return over this window — the factor legs carried the book.
Gross
+16.3%
Market
+24.8%
Sector
-2.3%
Subsector
+0.7%
Stock selection
-7.2%
Top holdings
31 reported positions; top 15 shown. “Stock-specific” is the share of each position's variance not explained by market, sector, or subsector factors.
| # | Holding | Weight | Value | Stock-specific |
|---|---|---|---|---|
| 1 | WBDWarner Bros. Discovery, Inc. Series A | 13.2% | $533M | 93% |
| 2 | AMZNAmazon.com, Inc. | 10.4% | $417M | 46% |
| 3 | GOOGLmodeled on GOOGAlphabet Inc. | 9.1% | $366M | 60% |
| 4 | TDSTelephone and Data Systems, Inc. | 5.8% | $232M | 96% |
| 5 | LYVLive Nation Entertainment, Inc. | 5.2% | $208M | 93% |
| 6 | KEYSKeysight Technologies Inc | 5.1% | $205M | 64% |
| 7 | SGISomnigroup International Inc. | 5.0% | $202M | 49% |
| 8 | XYZBlock, Inc. Class A | 4.8% | $194M | 66% |
| 9 | NSCNorfolk Southern Corporation | 4.7% | $189M | 79% |
| 10 | FLEXFlex Ltd | 4.1% | $167M | 53% |
| 11 | COFCapital One Financial Corp | 4.1% | $166M | 43% |
| 12 | HUTHut 8 Corp. | 3.8% | $152M | 72% |
| 13 | CRSCarpenter Technology Corporation | 3.3% | $133M | 84% |
| 14 | APGAPi Group Corporation | 3.0% | $121M | 55% |
| 15 | MTZMasTec, Inc. | 3.0% | $121M | 68% |
91% of reported AUM is in the ERM3 model universe; attribution and risk shares cover that subset. Source: SEC Form 13F-HR, filed 2026-08-14.
How concentrated is this book?
Quarter-end concentration across 83 filings from 2005-12-31 to 2026-06-30, with the latest filing set against this manager's own median.
- Effective N
- 15.9
- median 12.1
- Top 5 weight
- —
- no median
- Top 10 weight
- 67.4%
- median 77.3%
- Weight HHI
- —
- no median
Effective N is inverse-Herfindahl breadth, not a holding count: it falls well below the number of positions when one name dominates, which is the case worth seeing. This book held 31 positions at 2026-06-30.
Interrogate this portfolio
Ask the AI risk analyst what changed last quarter, where the concentration risk sits, or how your portfolio overlaps with Dan Loeb's.