Sequoia Capital
Alfred Lin & Pat Grady (co-stewards from Nov 2025, succeeding Roelof Botha); registered investment adviser
Generational-company backer via the evergreen Sequoia Capital Fund. In 2026 broke its own taboo by backing competing AI labs (OpenAI + SpaceX (xAI) + Anthropic) — betting the frontier-AI category is too big to pick one winner.
- Not open — Not an open-ended vehicle — analysis only, not an available allocation.
Key positions
Long-held frontier-model stake (~$852B last round)
Tied to Sequoia's broad Musk relationship (X, SpaceX, Boring, Neuralink)
In talks Jan 2026 to join the ~$25B megaround at ~$350B+ valuation — breaking the no-compete taboo
Pre-IPO stake; filed for a record IPO at ~$1.75T in 2026
Marquee fintech position, long-held
Track record of generational outcomes
Recent moves
In Jan 2026 moved to back Anthropic despite already holding OpenAI and SpaceX (xAI) — a deliberate break with its historical one-bet-per-category rule, signaling conviction that frontier AI supports multiple winners.
Loading disclosed positions…
Our take
Backing all three leading labs at once is the defining 2026 Sequoia story — maximal exposure to whoever wins, at the cost of the disciplined no-compete focus that built its reputation. Leadership just changed hands, so watch for strategy continuity.
The multi-lab pivot is usually narrated as conviction.
The sequence suggests governance. Roelof Botha — publicly cautious on concentrated AI exposure — stepped down in November 2025 under partner pressure, and the first Anthropic cheque followed roughly ten weeks later under Lin and Grady. A strategy that changed after the dissenter left is a different object from one that changed because the evidence did. The deeper point is what owning all three does to the return math. Sequoia's excess return historically came from being early and alone. Backing OpenAI, Anthropic and xAI is closer to the opposite trade: it earns the category's return, not the picker's, net of venture-grade fees. Entry price then becomes the whole edge — and the entries diverge sharply. Sequoia led SpaceX's Series J in February 2021 at roughly $74B post, and its OpenAI exposure predates 2026. Anthropic was a first cheque at roughly $350B post in the ~$25B round co-led by GIC and Coatue. One of those is an early bet; one is late-stage participation priced like a crossover fund.
Thesis
Sequoia made two changes at once in 2026, and the market reads them as one. It abandoned one-bet-per-category to back OpenAI, xAI and Anthropic simultaneously, and it runs an evergreen architecture — an open-ended public-equity fund that is the sole LP of closed-end venture sub-funds, with exit proceeds recycling back rather than distributing out. Together they push a concentrated stock-picking franchise toward owning the frontier-AI category at venture terms. The question is no longer whether Sequoia picked right, but whether owning the category still counts as picking.
Two layers since the 2021 restructuring: the open-ended Sequoia Capital Fund holds public positions and is sole LP to closed-end sub-funds doing the venture investing, seed through growth. Sub-fund proceeds recycle back into the evergreen fund rather than out to LPs, who hold accounts with annual redemption rights. Private positions sit in the closed-end vehicles; the evergreen fund is the destination of exits, not the holder of private marks.
Assessment
- Evergreen structure fixes a real venture defect: forced 10-year distributions dump compounders early. Recycling exit proceeds is the right answer.
- Diversifying across labs is defensible if frontier AI is a multi-winner oligopoly rather than winner-take-all — a view the 2026 evidence does not refute.
- Co-steward model (Lin + Grady) restores the Moritz/Leone structure of Sequoia's strongest decades; single-steward concentration was itself a risk.
- RIA registration lets the firm hold listed stock across asset classes without the forced-seller behavior that constrains drawdown-fund peers.
- The strategy change followed the removal of its loudest internal skeptic on concentrated AI. Conviction and consensus-after-dissent-departs look identical from outside and are not the same thing.
- Owning three competitors shifts a claimed alpha franchise toward category beta, still charged at venture economics. That is a fee-to-value mismatch unless entry prices were exceptional.
- Recycling by design means LP liquidity runs through redemption rather than distribution — contractual, but dependent on the fund honouring it in size when marks are stressed.
- The Anthropic entry near $350B is late-stage pricing. It needs Anthropic to compound several times from an already-enormous base just to earn a venture-grade multiple.
- Reported AUM ($56B to ~$88B depending on source) diverges widely and none we found traces to a current Form ADV. The headline number is not independently pinned.
- 2026-05That base compounded about 2.8x on paper in four months, $350B in Jan 2026 to $965B in May. The point inverts: having co-led the $965B round, Sequoia's marginal cheque and blended basis are the late-stage ones.
Record
No fund-level IRR, TVPI or DPI is disclosed, and the evergreen structure makes vintage attribution hard. What exists is marks — and one leg is no longer private. SpaceX absorbed xAI all-stock on 2 Feb 2026 ($1.25T combined), listed 12 June 2026 at $135/share (~$1.77T), and closed day one at $161.11, ~$2.1T. Against Sequoia's Series J entry at ~$74B post in Feb 2021, that is a listed, mark-to-market position far above basis, not a negotiated private mark. Anthropic was struck near $350B. So the record is now half-testable. The listing is the first mega-mark that can convert to cash; whether Sequoia sells post-lockup, and whether redeeming LPs get proceeds without gating, will say more than any marketing document. The private leg remains unproven at the point that counts.
- 2026-08No longer prospective: the first lock-up tranche frees roughly 911.5M shares, about $123B at recent prices, on 6 Aug 2026 — the cited driver of July's -36%. The test lands with SpaceX below its listing price.
- 2026-07SpaceX closed $108.37 on 31 Jul 2026, about $1.43T — below the $135 June listing price, some 52% off the $225.64 high of 16 Jun and -36% in July. The ~$2.1T reading was a high-water print.
- 2026-05Anthropic's Series H closed 28 May 2026 at $65B on a $965B post-money, with Sequoia a named co-lead. The ~$350B figure is its Jan-2026 Series F basis, not the live valuation.
Risks & fit
- Theme concentration: the largest positions all key off the same frontier-AI narrative and the same compute-capex cycle, listed and private alike.
- Mark credibility on the private leg — round prices are negotiated by a small set of repeat participants and are not clearing prices for a whole position.
- Redemption under stress: if AI marks compress and many LPs redeem at once, a structure holding illiquid sub-fund interests faces a gating decision.
- Key-person and cultural turnover after a contested leadership transition; partner departures follow governance fights more often than they precede them.
- Reflexivity — funding rivals of its own portfolio companies creates information-wall and board-conflict problems the one-bet rule prevented by construction.
The 'category beta at venture prices' read is already weakened on the SpaceX leg: Sequoia led the Series J at ~$74B against a ~$2.1T listed mark, so its blended basis there is plainly early, and pre-2026 OpenAI entries point the same way. What remains open is Anthropic. The read breaks if that ~$350B entry proves a small share of frontier-AI dollars deployed, or if Anthropic compounds enough to clear a venture-grade multiple from that base. It is confirmed if the labs converge on similar economics and Sequoia's blended frontier-AI return tracks the category average net of fees.
Useful for understanding how the evergreen-plus-sub-fund architecture changes what 'returns' mean in venture — DPI becomes a redemption-policy question, a governance disclosure rather than a performance one. Also a live case study in whether category-wide AI exposure still produces manager-level excess return, now that one leg is publicly priced. Access is institutional and closed; nothing here suggests the fund is available to or appropriate for any reader.
Fee terms for the Sequoia Capital Fund are not publicly disclosed in sources we verified. Reporting on the 2021 restructuring described a move away from standard 2/20 toward a lower management fee with carry; current terms unconfirmed.