Lone Pine Capital
Stephen Mandel Jr. (founder); Kelly Granat & David Craver (co-CIOs)
Classic Tiger-cub high-conviction growth fund that rotated out of mega-cap tech into more direct AI-infrastructure, power, and energy-transition plays.
- Not open — Not an open-ended vehicle — analysis only, not an available allocation.
Top holdings
Top position; AI-data-center power
EUV litho monopoly
Specialty alloys; aerospace/industrial
Financials diversifier
AI-driven ad-tech
Recent moves
Exited Microsoft, Amazon, and Broadcom in favor of AI-infrastructure / power / energy-transition names; top new Q1 2026 buys included Teradyne, Corning, and MasTec, with Vistra (data-center power) the largest position.
Loading disclosed positions…
Our take
Lone Pine's exit from the obvious mega-cap AI names into power (Vistra), litho (ASML), and infrastructure buildout (MasTec, Corning) is a deliberate move down the value chain — a bet that the bottleneck (and the alpha) has shifted from chips to the energy and physical buildout feeding them.
Two things a marketing deck won't foreground.
First, the record on display is not Mandel's — he stepped back from day-to-day investing in January 2019 — and the successor structure has itself just changed. At the start of 2026 three senior investors (Samuel Harland, Zachary Gleser, Kevin Salimian) left to launch their own firms, Rahul Anne was promoted to PM (the first in over a decade), and portfolio decisions now require a majority of Anne, Craver and Granat. The brand trades on Tiger lineage; the decision-making body is months old. Second, the 'AI-infrastructure pivot' framing is looser than it looks. Of the disclosed top five, LPL Financial (4.9%, brokerage) and AppLovin (4.7%, adtech) are not AI infrastructure at all, and Carpenter Technology (5.7%) is primarily an aerospace/medical alloys cyclical. The coherent AI-buildout sleeve is Vistra and ASML. A tidy story has been read backwards out of an eclectic book — and the parts that fit are a levered bet on hyperscaler capex continuing.
Thesis
Lone Pine's disclosed Q1 2026 book reads as a clean 'we moved down the AI value chain from chips to power' story — Vistra 7.4%, ASML 6.9%. Our read: that narrative is partly imposed on the filing after the fact, and more importantly it is not diversification away from the AI-capex factor. It is the same factor expressed one layer down, in businesses with more cyclicality, more regulatory exposure and less pricing power than the mega-caps it sits alongside.
Classic Tiger-cub long/short equity: concentrated, deep-fundamental, high-conviction growth, run at the flagship Lone Cypress vehicle. Founded 1997 by Stephen Mandel Jr., a Julian Robertson alumnus. The disclosed Q1 2026 13F holds 36 positions worth $12.54B with the top five at 29.6% of the book — bottom-up single-name selection, not factor or systematic construction.
Assessment
- Where the bottleneck logic is actually expressed it is specific and sized: Vistra 7.4% plus ASML 6.9% is a real bet that power and lithography constrain the buildout.
- Genuine concentration — 36 names, top five 29.6% — is intellectually honest for a fundamental shop. The research either works or it doesn't; there is nowhere to hide.
- Absorbed a 36% drawdown and roughly $3B of redemptions without style-drifting into a diluted, index-hugging book.
- Promoting from within and formalising a majority-vote decision rule at least makes co-CIO accountability explicit rather than diffuse.
- The 'rotation' is a factor re-expression, not a hedge. Vistra, MasTec, Corning and Carpenter all still depend on hyperscaler capex — same driver, worse business quality, higher cyclical beta.
- Power names carry regulatory and political risk the mega-caps didn't: PJM capacity auction outcomes, state fights over data-center tariffs, and PPA repricing can hit the thesis without the AI thesis being wrong.
- The 13F shows the long book only. Lone Cypress is long/SHORT — the hedges are invisible, so any inferred 'view' from position changes may just be a gross-exposure decision.
- $12.54B disclosed against roughly $19B publicly reported firm AUM means the filing covers on the order of two-thirds of the book — and nothing of leverage or the short side.
- Attribution ambiguity: a reported 16% in 1H 2025 came largely from Meta, Intuit and Microsoft — mega-cap tech that required no edge that half-year. That figure is a paywalled press estimate we could not independently verify.
Record
The record is a barbell, not a compound. Institutional Investor reports Lone Cypress lost 36% in 2022, gained 19% in 2023, and was up about 23% through Q3 2024. Do the arithmetic the tearsheet skips: 0.64 x 1.19 x 1.23 = 0.94 — still roughly 6% below the 2022 high-water mark almost three years later, and that is gross, before fees. A 36% loss needs a 56% gain to get back. If the press-reported 16% for 1H 2025 is right, the mark was cleared sometime in 2025 — three-plus years to even. The loss and the gains share one explanation: concentrated, levered, long-growth exposure that moves violently both ways. Real and repeatable, but closer to a high-beta expression of the growth factor than to a market-neutral skill engine. No clean firm-wide 2025 full-year or 2026 YTD net figure is public.
Risks & fit
- Governance is months old: a new majority-vote structure (Anne, Craver, Granat) formed after three senior investors departed in early 2026, untested through a drawdown.
- AI-capex deceleration hits the current book harder than a mega-cap book — cyclicals de-rate without a secular multiple to fall back on.
- Concentration at 29.6% in five names means two bad calls define a year.
- Redemption reflexivity: the 2022-23 episode showed roughly $3B of outflows following a drawdown, forcing selling into weakness.
- Power-sector policy shifts (interconnect queues, ratepayer politics) can impair Vistra independent of AI demand.
The cleanest test: whether the power/buildout sleeve produces return that Vistra and ASML don't. If coming 13Fs show these names held while reported returns track them roughly one-for-one, the result is beta to an AI-capex basket, not selection skill. Conversely, if the complex de-rates on a hyperscaler capex guidance cut and returns hold, the edge is sizing and timing rather than the thesis. Second test, on the narrative itself: if forward power curves and PJM capacity pricing weaken while data-center announcements continue, the 'power is the bottleneck' premise is wrong on its own terms.
A study in how a concentrated fundamental shop re-expresses one macro thesis across a value chain, and a caution about reading strategy from 13F data. The filing is informative about which large US-listed longs a serious research team underwrote; it is near-useless for inferring net exposure, hedges, or actual risk, and the 45-day lag means the book has likely already moved. Read it as evidence about a process, not as a portfolio.
Fee terms are not publicly disclosed. Tiger-cub peers typically use high-water marks, which matters here: after the 36% 2022 drawdown, incentive fees would be suppressed until the mark recovered — changing net-of-fee experience versus gross rebound headlines.