Atreides Management / Atreides Foundation Fund: Hedge Fund Diligence (First Pass)

Prepared 2026-10-20 from the files in sources/. The newest evidence is the February 2026 investor letter (performance through December 2025, commentary on January 2026). The newest offering memorandum and Form ADV are from March 2022, and the only DDQ is from August 2020. No 2026 monthly returns, current exposures or current terms were available. This is a first-pass note, not full investment diligence, not ODD, and not a recommendation.

Source key: [L26] Investor Letter, 4 Feb 2026 · [Est25] Estimated Month End Returns, Dec 2025 · [PS25] Standard Portfolio Summaries, Dec 2024–Nov 2025 (monthly) · [Q2-25] / [Q3-25] Investor call decks, 2025 (image-only screen captures) · [EXP] Portfolio Summaries Oct 2023–Apr 2024 ("Exposure" files) · [OM22] Confidential Offering Memorandum, Onshore LP and Offshore Ltd, Mar 2022 · [ADV22] Form ADV 2A/2B, 31 Mar 2022 · [ADV21] Form ADV 2A, 31 Mar 2021 · [ADV20] Form ADV 2A (AUM as of 31 Dec 2019) · [Deck21] Presentation, Dec 2021 · [Deck20] Presentation, May 2020 · [L21] Annual Letter, 8 Feb 2021 · [L20] Annual Letter, Jun 2020 · [PS20] Portfolio Summaries Nov and Dec 2020 · [DDQ20] DDQ, Aug 2020 · [SuppA] Offshore Supplement, Sub-Class A shares (dated Apr 2019) · [Email21] Atreides IR responses to VS Partners ODD follow-ups, 19 Jan 2021, with attached subscription/redemption history and investor breakdown · [OTC] Novus analysis of Fidelity OTC Portfolio, Jun 2020

---

1. First Impression

What it is. Atreides runs a TMT- and consumer-focused long/short equity fund launched 15 May 2019 by Gavin Baker, who ran the $17B Fidelity OTC Fund from 2009 to 2017 [DDQ20 p.4, 7]. The fund runs high gross exposure (historically averaging 215%) at moderate net (roughly 25–60%), and it tries to win on "long-short spread" between fundamentally similar longs and shorts rather than on market direction [L26 p.3; Q2-25 p.4]. A crossover share class (A-3) can also hold up to 20% (plus 4% follow-ons) in private venture and growth companies [OM22 pp.8–9; PS25 Nov]. The master fund held about $4.0B in November 2025, up from $312M at launch. Firm assets also include a co-invest vehicle (about $1.0B), Atreides Ventures Fund I (about $189M) and the Valor Atreides AI Fund [PS25 Nov; L26 p.6].

Why it's worth a meeting.

• Full-cycle live record. There are 6.6 years of reported net returns: a COVID year, a growth-equity bear market, a recovery and an AI-led bull market. Since inception through December 2025, the public-only A-1 class has compounded at about 18.0% net against the NASDAQ at 19.1%. My figures from the monthly nets show realized volatility of about 17.5% against 19.7%, a beta of about 0.6 and down-capture of about 56% [PS25 Nov p.2; Est25].
• Strong recent stretch. The last three years were +21.9% (2023), +43.4% (2024) and +40.5% (2025 estimated). On a three-year basis the fund compounded at about 35% against the NASDAQ's 31% [PS25; Est25; L26 p.1].
• Institutional build-out. Service providers are institutional: Morgan Stanley Fund Services (administrator), KPMG (auditor), Goldman Sachs and Morgan Stanley (prime brokers), and Morgan Lewis and Ogier (counsel). The offshore fund has two independent directors [DDQ20 pp.10–11, 27; Deck21 p.16]. Headcount has doubled from 14 to 28 [L26 p.4].

What is unclear, and the biggest gaps:

• 2021–22 drawdown. From May 2021 to December 2022 the fund fell about 40% (A-1 net), and it did not regain its high until November 2024. The NASDAQ's drawdown over that period was about 33%. Long-short spread was negative in both 2021 and 2022 at gross exposure above 200% [Q2-25 p.3]. That outcome sits poorly with the stated aim of minimizing basis risk.
• Assets arrived near the peak. The firm said in June 2020 it would "pause fundraising at $750m" [L20 p.2]. Instead the fund grew to about $2.4B before soft-closing on 1 Feb 2021, which was just ahead of the drawdown [L21 pp.2–3]. Asset-weighted returns are likely well below the time-weighted record. This is an inference: no dollar-weighted data is in the sources.
• Edge sits mainly in the long book. ITD net long relative performance is +10.8% a year, while the net short book is −3.2% a year relative [Q2-25 p.3]. Spread was strongly positive in 2019, 2020, 2024 and 2025 and negative or flat in 2021–23.
• Expansion now. Two new public products (150/50 Technology and Low Net) that trade pari passu, plus Ventures Fund II, are being launched at about $4B flagship AUM. Single-name concentration is also rising, including AI subthemes and Unity, which drove January 2026's "worst month of long-short spread ever" [L26 pp.3, 5–6].
• Stale legal and ODD documents. The DDQ is from 2020 and the OM and ADV from 2022. The 2025 call decks are screen captures. The "Quarterly Top Positions" file is password-protected and could not be read.

---

2. Team

Facts. The firm had 14 employees at the February 2021 soft-close and 28 at February 2026 [L26 p.4]. The July 2025 org chart shows about 13 investment and trading staff and 12 legal, IR, finance and operations staff, all reporting to Baker [Q2-25 PDF p.41]. The Management Advisory Committee comprises Edouard Carmignac, Antonio Gracias (Valor) and James Star [DDQ20 pp.6, 9–10]. Visible departures across the sources are Dunyelle Rosen (GC/CCO, gone by September 2021), Liz Jacobson (IR head, gone by July 2025) and Stephanie Star (analyst, gone by July 2025). None is explained in the sources.

Read. Baker is the only named PM, the majority owner and the only key person in the OM. He approves every trade [DDQ20 p.18]. He also states the edge in individual terms: "The team is the only true edge," and process-driven edge is "a fairy tale" [L26 pp.29–30; L21 p.10]. The bench has deepened, with four MDs and a ventures partnership layer, but nothing in the sources shows anyone other than Baker with portfolio-level sizing or risk authority. Key-person risk is high. Baker's 15-month gap between leaving Fidelity (September 2017) and founding Atreides (January 2019) is not addressed in the reviewed materials.

---

3. Terms, Liquidity and Alignment

Sources: [OM22 pp.4–5; Deck21 p.16; DDQ20 p.14]. Founders' F classes (1.25–1.50% / 12.5–16.5%) are closed [OM22 p.5]. Platform sub-series A-2 and A-3 carry a higher management fee that is not quantified in the sources, and part of the A-4 management fee goes to the placement agent [OM22 pp.5–6]. The minimum is $3M [OM22 p.6]. Withdrawals pay within 30 days, with a 3% audit holdback on withdrawals of 97% or more [DDQ20 p.14]. On a key-person event (Baker's death, departure, 60 days' incapacity or "Cause"), LPs may withdraw all their liquid capital, quarterly and within one year at the latest. The private sleeve is excluded [OM22 pp.21–22]. Co-investments via Special Circumstances carry 0% management fee and 12% carry [Deck21 p.16].

Alignment, as evidenced.

• Positives. Baker invested a "significant portion" of his liquid net worth, later described as the "overwhelming majority" [DDQ20 p.13; Deck21 p.3]. Internal capital was 3% of fund AUM in January 2021 [Email21]. Management fees on privates are charged on the lower of cost or fair value, and performance fees on privates only on realization [Email21; SuppA PDF p.11]. Private-deal expenses are charged to A-3 only [Email21].
• Gaps and frictions. The current GP and employee commitment in dollars is not disclosed. Atreides Parties pay no fees, so GP money and LP money do not have identical economics [OM22 p.2]. The hurdle is low and clears easily in any up year. The fund bears a broad expense load: research and soft dollars, expert networks, alternative data, Form PF, D&O and regulatory-inquiry costs [OM22 pp.14–15; DDQ20 p.23]. Broken-deal costs that co-invest vehicles could have shared fall on A-3 unless the co-investors agree otherwise [OM22 p.15]. The OM reserves side letters and new classes with "more favorable" rights [OM22 pp.6–7; ADV22 p.6].

---

4. Performance and Drawdowns

Reported net returns. A-1 is the highest-fee public-only class, for a representative day-one investor [PS25 Nov p.2; Est25].

From 15 May 2019. 2025 is estimated and unaudited. ITD figures are my compounding of the monthly nets and agree with [L26 p.1]. A-4 history before 2022 is hypothetical [PS25 footnote].

Computed from the A-1 monthly nets (my figures, not the manager's): maximum drawdown was about −40.1% (May 2021 to December 2022), recovered November 2024 (about 3.5 years underwater). Worst month −9.07% (May 2022), best +12.74% (May 2025). Up-capture vs NASDAQ is about 65%, down-capture about 56%, beta about 0.59 and correlation about 0.66. Annualized return per unit of volatility is about 1.03 for the fund against 0.97 for the NASDAQ.

Claims vs. evidence.

• "Sharpe ratio more than 50% higher than the NASDAQ's" ITD [L26 p.1]. The footnote shows this uses gross daily returns of the public book [L26 endnote 4]. On net monthly returns the advantage is much smaller: about 6% on a return/vol basis, by my calculation. The manager's own deck shows an ITD Sharpe of 0.85 through June 2025 without a NASDAQ comparison [Q2-25 p.3].
• Crossover adds value. A-3 is ahead of A-1 ITD, but that depends on marks. A-3 gained +16.1% in December 2025 alone against +1.9% for A-1 [Est25]. The crossover private book's net IRR was 11.3% at November 2025 [PS25 Nov] and 16% one month later [L26 p.6]. Private marks are set by Atreides' Valuation Committee "in its sole discretion," with a third-party valuation agent reviewed by the administrator [PS25 footnote 21; Email21].
• Fidelity record. Holdings-based gross was 19.7% vs. NASDAQ 17.9%, with an actual net of 19.4% (2009–2017). Up/down capture was about 110/111% and Sharpe 1.14 vs. 1.19 [OTC pp.2–3]. That long-only record was mostly market exposure with modest excess return, and the report says the strategy "differed substantially" from Atreides.

Are drawdowns explainable? Partly.

• 2021–22. The manager attributes 2022 mainly to his own error. He came in at 24.8% net in March 2022, added risk in May–June, then cut risk in July 2022 and captured only 59% of the NASDAQ's upside that month. He calls this "my biggest mistake since starting Atreides" [L26 p.4 fn.6]. That does not explain the 2021 loss: −8.1% against NASDAQ +22.2%, with a gross long/short spread of −11.5% [Q2-25 p.3]. Nor does it explain why the 2022 loss reached about 85% of the index's at roughly 25–50% net. The sources contain no 2021 or 2022 annual letter.
• Behavior after. The letter cites "nearly three years improving our organization" and process changes after the soft-close. These were described in a 5-year anniversary letter that is not in sources/ [L26 p.4]. Visible changes are more longs (29 in October 2023 to 62–70 in late 2025), lower top-10 concentration (89% to 56–75%) and custom AI sub-theme betas [EXP; PS25; L26 p.31].
• Recent stress. In February–March 2025 the A-1 class lost about 11.3% (NASDAQ −11.7%) while adding AI risk into the decline [PS25; L26 p.2]. January 2026 was described as the worst month of long-short spread ever, with no number given [L26 p.3].

---

5. Strategy and Edge

Stated approach.

• Long book. It owns companies with the highest expected 5–7 year IRRs on EV/FCF, aiming for 20%+ IRR [DDQ20 pp.14–16; Email21].
• Short book. It pairs fundamentally and quantitatively similar names to strip out basis risk [L21 p.11; L26 p.30].
• Turnover. It tends to be high: 8.5x in 2020, with longs at 2.0x and shorts at 14.0x [Email21].
• Crossover research. Private and venture work, now about 20 datacenter-semiconductor and optical portfolio companies, is described as an information edge for the public book, through the Valor/Crusoe partnership and Ciena and Lumentum sizing [L26 pp.3–4].
• Current themes. The book has concentrated in AI infrastructure: Nvidia, Astera Labs, Trainium, optics, memory and Pure Storage. Software shorts are a large hedge, about −17% of the −97% short book in November 2025. Videogame platforms such as Unity and Roblox are held as AI beneficiaries [L26 pp.24–29; PS25 Nov].

Claim / evidence / still to diligence.

Where it should work: high-dispersion, innovation-led tape with liquid TMT names (2020, 2024–25). Where it should struggle: factor rotations in which growth or quality longs and their paired shorts decouple (2021–22), and fast reversals after it has cut risk (July 2022).

Read. The evidence supports real stock-picking ability on the long side over 6.6 years. It does not yet show that the short book adds value, or that the spread holds up through a regime shift.

---

6. Portfolio, Risk and Capacity

Shorts include baskets and index futures. Exposures use delta-adjusted options and notional swaps, and exclude privates.

Stated guidelines vs. observed.

• Gross and net. Gross was expected at 125–250% and net at 25–65% [Deck20 p.20]. Observed gross peaked at 260% in Q1 2021 and reached 253% in November 2025 [Q2-25 p.4; PS25].
• Position counts. Guidance went from 10–40 longs (2020) to 35–50 longs with at most 30 single-name shorts at 2% maximum (2021) [Deck20 p.11; Deck21 p.10]. Long count reached 70 in 2025.
• Position size. The cap is 15%, and it is internal, not in the OM [DDQ20 p.18].
• Concentration. In November 2025 the book was about 39% gross and 24% net in semiconductors and another 57% gross in tech hardware [PS25 Nov]. The largest disclosed single positions were AMD at 8.7%, Nintendo at 6.4% and GitLab at 6.3% (mid-2025) [Q2-25 PDF pp.37–39].
• Private sleeve. Utilization was 21.1% of crossover NAV, or $800.6M at mark against $545.8M at cost across 77 companies (November 2025) [PS25 Nov].

Capacity. Capacity commitments have shifted three times:

• June 2020: pause at $750M [L20 p.2].
• February 2021: soft-close at about $2.4B [L21 pp.2–3].
• February 2026: the firm says it is now "strong enough to support thoughtful, measured expansion." It is adding two pari passu public products and a second ventures fund, and says the crossover fund has "minimal capacity" for new privates [L26 pp.4–6].

At roughly $4B and about 250% gross, the book carries about $10B of gross public exposure, by my estimate. Its stated LLCC framework says a fund can safely hold only one of liquidity, leverage, concentration or crowdedness at a time [L26 p.31]. Leverage is structurally high, and AI-theme crowding was rising through 2025. How the firm reconciles the two is a key question.

Other conflicts disclosed [ADV22 Items 4, 10]:

• Separately managed accounts and non-discretionary model advice to third parties, which raises allocation, sequencing and information-leakage questions.
• Baker is a strategic advisor to a Valor-affiliated PE fund, and Gracias advises Atreides on Designated Investments. The firm co-invests with Valor, including the Valor Atreides AI Fund.
• Baker sits on private portfolio-company boards.
• Proprietary information is shared with other market participants for idea exchange.
• James Star is an advisory-committee member and also the interested director on the offshore board. He sits on the Master Fund Advisory Board too, whose approval is needed to designate an existing holding as a Designated Investment [DDQ20 p.10; OM22 p.8].

---

7. Key Counterparties and ODD Snapshot

Daily three-way reconciliation, shadow NAV in Enfusion, a monthly Valuation Committee and multi-person wire controls with administrator release are described [DDQ20 pp.18–21]. ADV Item 9 reports no disciplinary events as of March 2022 [ADV22]. Every service-provider entry is three to five years old and should be reconfirmed. The private-valuation agent, any SOC 1 reports and current insurance limits are not identified.

---

8. Known / Unknown / Ask Next

---

9. Diligence Questions

Incentive Alignment

• How much partner and employee capital sits in each Atreides vehicle today, and how has it changed since 2021? Is any of it in the new 150/50 and Low Net products?
• Will the new products carry their own fees on capital that moves between vehicles at the 5-day switch? How will allocation conflicts between the flagship, the new products, SMAs and model-advice clients be managed?
• What side letters with fee, liquidity or transparency terms exist, and is there an MFN?

Edge & Repeatability

• Short alpha is negative ITD. What would make the short book a source of return rather than only a hedge, and what changed after 2021–22?
• What share of 2024–25 alpha is attributable to the in-house "AI Beneficiary" factor versus selection within AI?
• Give two or three public positions where privates changed the sizing decision, including cases where it hurt, such as GitLab.

Performance & Drawdowns

• Walk through 2021 month by month. Why was the long/short spread −11.5% at about 250% gross?
• Did the growth in exposure from late 2020 to early 2021 coincide with the large inflows? What were dollar-weighted LP returns for the 2020–21 cohort?
• What was the January 2026 result, and what has 2026 looked like since?

Team Stability

• Who besides Baker can initiate, size or exit a position? What happens to the book in a key-person event?
• Why did the original GC/CCO and IR head leave? Has any investment MD left?
• How are MDs paid, and do the ventures partners share in carry from AVF and VAAI?

Risk, Financing & Capacity

• What are the hard limits on gross, net, single-name and theme exposure, and have any been breached?
• At $4B and about 250% gross, how many days would it take to cut gross by half? What borrow and financing concentration exists across GS and MS?
• What is the realistic capacity of the flagship strategy once the new pari passu vehicles are added?

---

10. Reference Check Prep

Current investors. How did Atreides communicate during 2022, and did the explanations match the outcomes? Any surprise on side pockets, gates or marks in A-3? How has reporting changed since 2023?

Former investors. Did you redeem during or after the 2021–22 drawdown, and why? Were withdrawals honored on time? Were any gates or holdbacks applied?

Former employees. Is investment debate real, or does Baker decide? How are risk overrides handled? Why did people leave? Did growth in privates or ventures pull attention from the public book?

Service providers. Did MSFS or the valuation agent challenge private marks? Have there been NAV restatements or late audits? How responsive is the operations team?

---

11. Suggested Materials Request List

1. Current DDQ (2025 or 2026) and current Form ADV Part 2A/2B
2. Current OM, all sub-series supplements (including platform and A-4) and new-product term sheets
3. Monthly net returns for all classes, 2026 YTD; audited financial statements 2022–2025
4. AUM and gross flows history by vehicle; dollar-weighted returns by cohort
5. Annual letters for 2021–2024, including the 5-year anniversary letter
6. Annual attribution by long, short, theme and analyst, plus BARRA factor attribution
7. Exposure history (gross, net, beta-adjusted, sector, liquidity) and written risk limits
8. Private-book schedule with valuation methodology, third-party valuation agent, marks history and realizations
9. GP and employee capital by vehicle
10. Org chart with joiners and leavers since 2021, and the succession or key-person plan
11. Current service-provider list, SOC 1 reports and the insurance schedule
12. Side-letter and MFN summary; allocation policy across the flagship, new products, SMAs and model clients
13. Reference list covering current and former LPs, former employees and the administrator
14. An unprotected copy of the "Quarterly Top Positions" file

---

12. Sources Reviewed

All files below are in Atreides_Foundation/sources/. Dates are document dates unless noted.

• Investor Letter, 4 Feb 2026 (34 pp.)
• Estimated Month End Returns, Dec 2025 (1 p.)
• Standard Portfolio Summaries, Dec 2024 and Jan–Nov 2025 (12 files)
• 2025 Q2 and Q3 Investor Call decks (image-only screen captures; reviewed visually, partial)
• Portfolio Summaries ("Exposure" files), Oct 2023–Apr 2024 (7 files)
• Reflections on Generative AI, Jun 2023 (thematic; not relied on)
• "Atreides Ventre Capital Call," 12 Jul 2022 (.pptx: slide images with typed meeting notes on private marks, authorship unclear; not relied on)
• Confidential Offering Memoranda, Onshore LP and Offshore Ltd, Mar 2022; privacy notices, Mar/Apr 2022
• Form ADV Part 2A/2B, 31 Mar 2022; 31 Mar 2021; and an undated version with AUM as of 31 Dec 2019
• Presentation, Dec 2021
• Administrator new-issue certification e-mail, MSFS, 29 Apr 2021 (routine)
• Annual Letter, 8 Feb 2021
• "Investor breakdown" e-mail thread, Atreides IR to VS Partners, 6–19 Jan 2021, with the ITD subscriptions/redemptions table and investor-mix table
• Portfolio Summaries Nov and Dec 2020; "Portfolio Summaries through 11.30.2020.xlsx" (not separately parsed)
• Sample Portfolio 2020.12.xlsx (VS Partners PORT characteristics run, 31 Dec 2020)
• Code of Ethics, 31 Dec 2020; Compliance Manual table of contents, May 2020
• Confidential Offering Memorandum (offshore, about 2020), Supplements, and Subscription Documents for Sub-Class A shares (supplement dated Apr 2019)
• DDQ, Aug 2020
• Annual Letter, Jun 2020
• Presentation, May 2020
• Novus analysis of the Fidelity OTC Portfolio (Jul 2009–Sep 2017), generated 12 Jun 2020

Not readable: "Quarterly Top Positions.xlsx" is password-encrypted.

Not used as evidence: "Atreides Alpha Beta 2025.xlsx" and "Data Analytics.xlsx" sit outside sources/ and were treated as prior work.

No Coverage File exists for this fund.

Not available:

• Any 2026 monthly data
• Annual letters for 2021–2024
• Audited financials
• A DDQ, OM or ADV dated after March 2022
• Any outside or web source; none was consulted

Method: All computed statistics (drawdown, capture, beta, return/vol) are mine, derived from the reported monthly net returns in [PS25 Nov p.2] and [Est25].