Coatue Management
Philippe Laffont (founder, CIO)
Tiger-cub tech specialist that has rotated down the semiconductor supply chain — from the most-owned AI name into the equipment, foundry, and power layers that make AI compute possible.
- Not open — Not an open-ended vehicle — analysis only, not an available allocation.
- Partial 13F — The disclosed book is longs-only and ~45 days lagged — a slice, not the strategy.
Top holdings
Top position; foundry for every leading AI chip; added ~7% in Q1 2026
Data-center power / grid buildout bet
Semi-cap equipment
Semi-cap equipment; increased ~79% in Q1 2026
Custom AI silicon / networking
AWS hyperscaler
Recent moves
Initiated a ~$655M ASML stake, increased Applied Materials ~79% and TSMC ~7%, while trimming Nvidia for the 11th time in 12 quarters — a clear rotation from the headline GPU name into semi-cap equipment, foundry, and power. Listed among investors in Anthropic's $65B Series H (June 2026).
Loading disclosed positions…
Our take
Coatue's positioning signals a conviction that the durable AI margin sits in the picks-and-shovels (litho, semi-cap, foundry, grid power) rather than in NVDA itself — they're underwriting the capex super-cycle, not the single stock.
Two things the marketing deck won't lead with.
First, the 13F tells you far less than it appears: the reported book fell from $39.96B (Q4 2025) to $29.05B (Q1 2026), about 27%, in a stretch when the flagship compounded strongly. Reading that as de-risking is the amateur error — a 13F is longs-only, US-listed, ~45 days stale, and excludes shorts, non-US lines, derivatives and the entire private book. It is a partial view of one sleeve. Second, the semi-cap tilt is a factor bet dressed as diversification. TSM, LRCX, AMAT, ASML and GEV are not independent of the GPU complex; they are the same hyperscaler-capex variable with more operating leverage and, historically, deeper cyclical drawdowns — and Broadcom's top-six slot means chip-design exposure never actually left. Continued Nvidia trimming (Q1 2026 coverage counts an eleventh such trim) cut single-name risk and arguably raised cycle risk. If AI capex guidance is cut, equipment order books — booked years ahead — reprice harder than a chip designer's earnings do.
Thesis
Coatue is a Tiger-cub tech long/short whose disclosed book now sits largely one layer down the AI stack — foundry, semi-cap and grid power (TSM 10.8%, GEV 7.7%, LRCX 7.4%, AMAT 6.2%, Eaton 5.8%, plus a new ~$655M ASML stake, per the Q1 2026 13F filed 15 May 2026) — though Broadcom at 5.9% shows the rotation is partial, not a clean exit from chip design. The bet is that durable AI margin accrues to the capex chain. Our critique: that diversifies names, not the underlying risk factor — nearly all of it keys off one variable, hyperscaler capex.
Concentrated, thematic technology long/short built around secular platform shifts, with a crossover private/venture arm (growth, structured capital) that lets one theme be expressed pre- and post-IPO. Founded 1999 by Philippe Laffont, ex-Tiger Management. Sizing is high-conviction: the top six disclosed longs are roughly 44% of the reported $29.05B equity book (Q1 2026 13F).
Assessment
- Genuine domain depth in semis and infrastructure; the move down the stack was made early and sized meaningfully, not as a token hedge.
- Crossover structure underwrites one thesis across private and public stages — real reach, e.g. co-leading Anthropic's May 2026 round.
- Willingness to keep cutting a winning consensus position (an eleventh Nvidia trim by Q1 2026) is rarer and harder than adding to one.
- Post-2022 discipline: 2023 +21.5% and 2024 +18.7% (Institutional Investor) rebuilt above the high-water mark rather than reaching for risk.
- Names diversified, factor did not. Foundry, semi-cap, power and chip design all discount the same hyperscaler capex curve; correlation converges in a capex cut.
- The 'one layer down the stack' framing is only half true — Broadcom, a fabless AI chip designer, is a top-six position at 5.9% of the disclosed book.
- 2023 is the attribution problem: +21.5% in a year the Nasdaq 100 rose far more. A tech-concentrated book underperforming its own beta in a tech melt-up is hard to square with stock-picking skill.
- May 2026's +14.2% — reported as the best month in 25 years — is beta-shaped, arriving inside a ~20% first-half Nasdaq 100 run. Magnitude that large usually implies gross exposure, not selection.
- Key-person concentration on Laffont, founder since 1999, with no publicly named successor. Tiger-cub firms have a poor record of surviving founder transitions.
- 2026-07That correlation risk was realised in July 2026 rather than prospective: AMAT -29.8%, LRCX -32.3%, TSM -15.4% and GEV -15.7% in a single month while the S&P 500 was -0.1%.
- 2026-07In July 2026 that trimming sat on the wrong side: NVIDIA +0.3% and Broadcom roughly flat were the haven, while the semi-cap sleeve the proceeds funded fell about 30%.
- 2026-07The residual chip-design exposure earned its keep in July 2026: Broadcom at 5.9% was roughly flat while LRCX and AMAT fell about 30%. It read as a buffer on the rotation, not a logical flaw in it.
Record
Disclosed record, vintage-stamped: 2022 −19%; 2023 +21.5%; 2024 +18.7% (Institutional Investor's Rich List, which notes this put the main long/short fund above its high-water mark); 2025 approximately +13% (press-reported, not fund-published); 2026 YTD +24.5% through 3 July 2026, including +4.7% in June, +14.2% in May and a near-5% March drawdown (Hedgeweek attributes that month to Middle East conflict; other coverage of the same Nasdaq 100 selloff points to US tariff threats — treat the cause as contested, the magnitude as reported). Attribution read: the 2026 first half broadly tracks the Nasdaq 100's ~20% 1H move, so the headline is consistent with a levered long tech book capturing beta rather than separable alpha. The −19%/+21.5% pair is more telling — it takes about 23.5% to recover a 19% loss, so 2023 roughly restored capital rather than adding; 2024 is what cleared the high-water mark.
Risks & fit
- Hyperscaler capex digestion: semi-cap order books are the most cyclical link in the chain and reprice ahead of revenue.
- Concentration — top six disclosed longs ~44% of the reported equity book; a single thesis break is portfolio-level.
- Geopolitical: TSM at 10.8% carries Taiwan-strait risk that diversification within the chain does not remove.
- Leverage and gross exposure are undisclosed; drawdowns of the observed magnitude imply it matters and cannot be sized from public data.
- Private-book marks are round-price, not liquid — illiquidity and mark credibility bite if the AI funding cycle cools before exits arrive.
The read that this is capex-cycle beta rather than skill fails if, in a period where AI infrastructure names fall meaningfully, the flagship holds up materially better than a semi-cap/foundry basket — that would evidence real short-book and hedging contribution invisible in the 13F. It also fails if a full drawdown-and-recovery cycle shows the fund beating the Nasdaq 100 on a risk-adjusted basis in both directions, not just up-tape years. Conversely, the concern is confirmed if a hyperscaler capex guide-down produces a fund drawdown at or worse than the AI-infrastructure complex itself.
Readers analysing how a large, well-resourced tech specialist is positioning for the AI capex cycle, and anyone using 13F data who needs to understand its limits. Coatue's flagship is a private fund not generally open to retail; this is an analytical read on the strategy's logic and record, not a route to access it and not a suggestion to mirror any position.
Coatue's specific fee terms are not publicly disclosed, so we state no number. What is observable is the high-water-mark mechanic: the 2022 −19% meant no incentive fee could accrue on that capital until recovery, which reporting indicates took until 2024.