D. E. Shaw
Founded by David E. Shaw (computer scientist, ex-Morgan Stanley quant; the firm seeded the founders of Two Sigma). Blends systematic and discretionary strategies.
A pioneer of computational/quant investing that now blends systematic and human-run discretionary strategies. The 13F long book leads with exactly the AI-compute complex: NVDA, Microsoft, Broadcom (AVGO — custom AI silicon/networking), Alphabet, Apple, with the top-5 (NVDA, MSFT, QQQ, PLTR, MU) ~11% of the equity book. AVGO and MU give it deeper semiconductor breadth than the pure index-cap names. Strong 2025 across its flagship Composite and macro Oculus funds.
- Not open — Not an open-ended vehicle — analysis only, not an available allocation.
- Self-reported — Headline returns are manager- or firm-reported and unaudited.
Top holdings
#1 holding, +$584M added — largest equity position Q1 2026 and increased; clearest AI-trade signal.
Top-3 position. Mag-7; Azure AI compute.
Top-5 position. Custom AI accelerators + AI networking — semis exposure beyond NVDA.
Top-5 position. Mag-7; Gemini/TPU.
Top-5 position. Memory/HBM for AI accelerators.
Recent moves
Q1 2026 13F (period ending Mar 31, 2026): added ~$584M NVDA; top-5 NVDA/MSFT/QQQ/PLTR/MU. Firm grew to ~$85B AUM and paused capital returns. Source: Insider Monkey / Alternative Fund Insight.
Loading disclosed positions…
Our take
D.E. Shaw shows the broadest genuine semiconductor breadth of the cluster — not just NVDA but AVGO (AI silicon/networking) and MU (HBM memory) — and it added to NVDA in Q1, the strongest visible AI-trade tilt here. Because the firm blends systematic and discretionary books, its 13F is a partial step closer to conviction than a pure quant's, but still a netted gross-long snapshot. Honest limit: 13F is long US equity only; Oculus's macro/derivative bets (its 28% driver) are entirely invisible here.
Two things worth separating. First, the 13F is close to meaningless here.
The disclosed gross notional (~$166.3B, period ending Jun 30 2026) exceeds even the firm's Form ADV regulatory AUM (~$120.5B, Apr 2024) and is roughly double net AUM (~$85B) — that alone tells you the filing is an un-netted, gross-long, US-listed-only snapshot of a leveraged book with shorts and derivatives invisible. Reading NVDA/AVGO/MU as an 'AI conviction tilt' misreads a systematic portfolio's long leg as a thesis. Second, attribution. Oculus (+36.1% 2024, +28.2% 2025) is the return story, and it is macro/discretionary — closer to a Brevan Howard than to a stat-arb machine. Two consecutive 28-36% macro years in a volatile-rates regime is either exceptional judgment or a regime that suited the book; that only resolves on the next regime turn. Composite's ~12.9% annualized net since inception is the more honest, longer-baseline number — and it is net of a fee load near the industry ceiling.
Thesis
D. E. Shaw is the rare quant house that solved its own succession — founder David Shaw stepped out of day-to-day management in 2002 and now serves as chief scientist of D. E. Shaw Research, leaving a seven-person Executive Committee the firm says has worked together more than 15 years. The flagship Composite fund has been closed to new capital since mid-2013, a deliberate capacity cap rather than an AUM grab. The tension in 2025-26: headline returns increasingly come from Oculus, a DISCRETIONARY macro book, not the systematic engine the brand is known for.
Multi-strategy across systematic and discretionary books. Composite (~25 years old, closed since 2013) is the multi-strat flagship; Oculus is the global-macro vehicle; Valence a separate multi-strat. Breadth spans equity market-neutral, macro, credit, reinsurance, private capital and commodities, run by 3,000+ staff. Capacity is actively managed — capital has historically been returned to hold funds at size, though that practice was paused entering 2026.
Assessment
- Key-person risk already discharged: 24 years running without its founder in operations, under an Executive Committee the firm states has worked together 15+ years.
- Capacity discipline is real, not marketing: Composite closed to new capital since mid-2013, capital historically returned rather than compounded into decay.
- Genuine breadth — systematic plus discretionary macro, credit, reinsurance, private capital — means no single alpha source has to carry the firm.
- ~12.9% annualized net since inception across a quarter-century spans multiple regimes, a longer baseline than most quant records offer.
- The pause on returning cash entering 2026 breaks the capacity discipline that is the firm's own stated reason for performing — either capacity genuinely expanded, or discipline is loosening after a good year.
- Headline returns are increasingly discretionary-macro (Oculus), not systematic. A different, less-diversifying, more key-person-dependent risk than the quant brand implies.
- Fee load sits near the industry ceiling: performance fees rose 5pts in July 2022 to 35% on Composite (2.5% mgmt) and 40% on Valence (3.5% mgmt), after a prior increase effective Jan 2019.
- Everything material is self-reported. Returns come via press sourcing, not audited public filings; there is no independent verification path for an outside reader.
- Systematic equity alpha is the most crowded, most decay-prone factor space in the industry, and AUM grew from ~$65B to ~$85B in a single year.
Record
Composite returned an estimated ~18.5% in 2025 (vs ~18.2% in 2024), with ~12.9% annualized net since inception. Oculus, the macro fund, returned ~28.2% in 2025 after ~36.1% in 2024. Firm AUM went from ~$65B (2024) to ~$85B (Jan 2, 2026) — growth driven by new products and multi-strat, not by reopening Composite. Attribution honesty: the eye-catching numbers are Oculus's macro book, while the long-baseline number is Composite's 12.9%. A 12.9% net compound over 25 years, after a 2.5-and-35 toll (lower in earlier years), implies a very large gross return — the open question is how much of that gross is skill versus leverage on a modest per-unit edge. All figures are firm-sourced via trade press, not audited disclosures.
Risks & fit
- Regime dependence in Oculus: back-to-back 28-36% macro years may not survive a rate/vol regime the book is not positioned for.
- Alpha decay in crowded systematic equity factors as the firm scales past $85B net AUM.
- Fee compression cuts both ways — a 2.5-and-35 structure demands sustained outperformance, and a flat year is expensive for LPs.
- Concentration of discretionary macro judgment in a small set of decision-makers reintroduces key-person risk the firm otherwise resolved.
- Opacity: no audited public performance, so a bad year may surface late and via press leak rather than disclosure.
Wrong if Oculus's macro gains prove regime-specific rather than skill-based — watch for a materially negative or flat Oculus year on the next rates/vol turn while Composite holds up, confirming the macro book is beta to a favorable regime. The capacity concern is falsified if the firm resumes returning capital after the 2026 pause, or shows the pause funded genuinely new capacity rather than absorbing more capital into existing books. A sustained gap between Composite's net return and its 12.9% inception average would signal scale-driven decay.
Relevant to anyone studying how a quant firm manages capacity, succession, and the systematic-to-discretionary drift that comes with scale. Composite has been closed to new capital since mid-2013 and the firm's vehicles are private funds open only to qualified institutions and very-high-net-worth investors — a case study in institutional asset management, not an accessible vehicle. The 13F is a poor proxy for the book and should not be read as a replicable portfolio.
Bloomberg/Hedgeweek (Oct 2022): performance fees +5pts from July 2022 — Composite 2.5%/35%, Oculus 2%/30%, Valence 3.5%/40%. Prior hike effective Jan 1, 2019, reported by Institutional Investor as Composite at 3-and-30; sources conflict on mgmt fee.