Key Takeaways
- • Scott Wilson, CIO of Washington University's endowment, argues that direct co-investments, not fund commitments, separate top-quartile endowments from average ones. He says a fund-of-funds model would have left WashU at the peer average, while its directs place it at or near the top.
- • WashU turned over more than 75% of its roughly $8 billion portfolio in its first 18 months and redeemed all 36-37 hedge funds. Per Wilson, the endowment has generated about $17 billion of profit on the original $8 billion and paid about $5 billion to the university, and it now manages just under $20 billion.
- • Wilson targets roughly half the private book in directs and co-investments, which he would raise to 80% absent a 5% payout requirement. He treats look-through exposure as the unit of analysis, sees diversification benefits fade beyond 20-30 names, and notifies the board at 2-3% of portfolio cost.
- • SpaceX exceeded 20% of the portfolio at one point. Wilson calls his biggest mistake failing to add capital to winners, estimating the cost at "billions of dollars," and says the endowment should have sold a 2020-2021 secondary package at 80-90 cents on the dollar.
- • Wilson is skeptical of AI and late-stage private valuations, including Anthropic's reported private valuation above $2 trillion. He expects roughly half of today's AI-related companies to be gone in ten years and doubts model providers have durable moats.
Summary
In a roughly 57-minute episode of 20VC with Harry Stebbings, Scott Wilson, CIO of Washington University's endowment, lays out his portfolio construction philosophy. He contends that private markets are structurally less efficient than public markets and that directs and co-investments drive outperformance. By his account, 80-90% of his team's hours go to meeting individual companies, compared with the 70-90% spent on managers at a typical LP. The team has no dedicated direct-investing group, and every member is a generalist. Wilson concedes the edge is not superior deal origination. The process starts from a partner's conviction and then requires independent verification, underwriting, a monitoring plan, and confirmation that the position is the best marginal use of capital. He estimates the team invests in roughly 0.1% of what it sees.
Wilson took over nine years ago, when the endowment was just under $8 billion. In his first 18 months he turned over more than 75% of the portfolio through secondary sales, with redemptions filed for over half the portfolio in week one. He says the board gave him three years to show results and that changing an endowment portfolio is "more like turning around an aircraft carrier than a speedboat." The 5% annual payout is smoothed over a rolling five-year average and must come from liquid assets. Wilson cites the performance gap as justification for the constraint, noting that privates compounded in the low-to-mid 20s over the last decade versus about 10% for public equities.
On concentration, Wilson says the quantitative benefit of diversification trails off beyond 20-30 look-through names and that WashU has roughly half its portfolio in its top 50 positions, which he admits is still probably over-diversified. The board is notified above 2-3% of portfolio cost and must approve at 4%. He credits Lee Fixel, the board, and the team for the SpaceX outcome, in which the position exceeded 20% of the portfolio. The team used several continuation vehicles to hold through the IPO. Wilson opposes per-fund concentration caps, such as a 10% limit, which he calls "mostly artificial." For post-IPO holdings he favors gradual, multi-year selling. For Cerebras, the team hedged before distribution and still holds some shares, which he describes as "betting on the come."
Wilson identifies failing to add to winners as his largest error, a bias he attributes to anchoring on entry valuations. He also cites a mid-2020 secondary package that the team chose not to sell, expecting higher December-quarter marks. Venture markets then peaked and reversed, and he says the endowment was offered roughly 80-90 cents on the dollar. He also names two high-conviction direct losses, an automated manufacturing company and an additive manufacturing company.
WashU holds roughly 20-30 funds, adds one or two per year, and avoids funds above $10 billion, though Wilson describes that bias as not absolute. He argues that large funds dilute returns and that a manager returning capital in 18 months is preferable to one raising far more capital on a slower cadence. On AI, Wilson argues that a 2025-era valuation premium in private markets may not be validated by public buyers, citing Cerebras' weak trading since its IPO and SpaceX trading near its IPO price. He expects a shakeout over the next decade and says the top five large language models could be indistinguishable to users on 99% of tasks.
Wilson took over nine years ago, when the endowment was just under $8 billion. In his first 18 months he turned over more than 75% of the portfolio through secondary sales, with redemptions filed for over half the portfolio in week one. He says the board gave him three years to show results and that changing an endowment portfolio is "more like turning around an aircraft carrier than a speedboat." The 5% annual payout is smoothed over a rolling five-year average and must come from liquid assets. Wilson cites the performance gap as justification for the constraint, noting that privates compounded in the low-to-mid 20s over the last decade versus about 10% for public equities.
On concentration, Wilson says the quantitative benefit of diversification trails off beyond 20-30 look-through names and that WashU has roughly half its portfolio in its top 50 positions, which he admits is still probably over-diversified. The board is notified above 2-3% of portfolio cost and must approve at 4%. He credits Lee Fixel, the board, and the team for the SpaceX outcome, in which the position exceeded 20% of the portfolio. The team used several continuation vehicles to hold through the IPO. Wilson opposes per-fund concentration caps, such as a 10% limit, which he calls "mostly artificial." For post-IPO holdings he favors gradual, multi-year selling. For Cerebras, the team hedged before distribution and still holds some shares, which he describes as "betting on the come."
Wilson identifies failing to add to winners as his largest error, a bias he attributes to anchoring on entry valuations. He also cites a mid-2020 secondary package that the team chose not to sell, expecting higher December-quarter marks. Venture markets then peaked and reversed, and he says the endowment was offered roughly 80-90 cents on the dollar. He also names two high-conviction direct losses, an automated manufacturing company and an additive manufacturing company.
WashU holds roughly 20-30 funds, adds one or two per year, and avoids funds above $10 billion, though Wilson describes that bias as not absolute. He argues that large funds dilute returns and that a manager returning capital in 18 months is preferable to one raising far more capital on a slower cadence. On AI, Wilson argues that a 2025-era valuation premium in private markets may not be validated by public buyers, citing Cerebras' weak trading since its IPO and SpaceX trading near its IPO price. He expects a shakeout over the next decade and says the top five large language models could be indistinguishable to users on 99% of tasks.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.