20VC: Scott Wilson on WashU’s $19BN Endowment, SpaceX and AI
Topics
Early-Stage VC Investing Dynamics
Macro Asset Allocation in High-Interest Regimes
AI Venture Funding and Bubble Risks
Battle for Frontier AI Moats
Pioneers
Scott Wilson
Summary
- As WashU’s managed pool grew from just under $8 billion to just under $20 billion, Wilson describes a strategy of concentrating behind managers’ best companies. He reports $17 billion of profits and roughly $5 billion paid to the university over nine years; without directs and co-investments, performance would have been peer-average. His playbook: find something unique, then “double, triple, quadruple down on it.”
- Liquidity requirements set the limit on Wilson’s preference for private investments. A roughly 5% annual payout consumes 10% of liquid assets when half the portfolio is liquid; without that obligation, he would probably target 70–80% privates. Private investments have compounded in the low-to-mid-20s over the last decade, versus roughly 10% for public equities, but current liquidity proceeds are returning to public markets.
- SpaceX’s rise above 20% of the portfolio was a deliberate decision to let a winner compound. Board support and continuation vehicles enabled WashU to retain the position through its IPO. Wilson’s biggest regret runs in the opposite direction: failing to add to winners because their prices or portfolio weights had increased, a mistake he says cost “billions of dollars.”
- WashU’s model favors smaller managers with company access over mega-funds with abundant capital. Wilson prefers a $1 billion fund returning every 18 months to a $10 billion fund returning every three years: larger checks usually bring higher entry valuations, while additional investments dilute selection. Stebbings counters with Thrive’s OpenAI and Cursor outcomes; Wilson concedes trillion-dollar outcomes make mega-fund returns possible, while questioning long-term returns on AI capital expenditure.
- Wilson blames private-market valuation premiums for weak distributions, rather than an unavailable IPO market. “You can’t go out public at the valuation that the private markets will give,” he argues. He thinks Anthropic could have raised privately above $2 trillion, but is unconvinced public investors would pay the same; WashU is redeploying liquidity into public businesses it considers relatively protected, “if not immune,” from AI.
- Model interchangeability leaves Wilson skeptical about AI’s eventual profit pools. With identical interfaces, he argues, users probably could not distinguish the top five models for 99% of tasks, while cheaper alternatives undermine willingness to pay. WashU nevertheless retains Anthropic exposure inside broader partner funds; Wilson expects a shakeout but says, “I don’t know what the timeline is.”
- Europe and China offer selective opportunities where Wilson thinks price understates business value. He finds European companies on small earnings multiples that he believes would command large revenue multiples in the US: “Price is what you pay, value is what you get.” In China, he accepts geopolitical risk selectively, favoring companies that can return capital and operate below the political radar or with political support.
Deep dive
1. WashU’s overhaul paired wholesale exits with concentrated reinvestment
- Wilson brought Grinnell’s mix of external managers, directs and co-investments to WashU: find good partners, then “double, triple, quadruple down on” exceptional opportunities. Over nine years, he reports roughly $17 billion of profits on an initial pool just under $8 billion, alongside approximately $5 billion paid to the university. Today’s managed pool is just under $20 billion.
- Stebbings’ opening claim of a first-week 75% fire sale gets corrected: Wilson submitted redemptions for over half the portfolio, following months of preparation. Turnover ultimately exceeded 75% over 18 months, involving private secondaries; all 36 or 37 hedge funds went. The board wanted improvement within three years, while restructuring an endowment resembled “turning around an aircraft carrier.”
- The performance distinction is direct exposure, Wilson says: without directs and co-investments, WashU would have landed around its peer-group average; with them, it sits “at the top, if not the very top” over the last decade. He rejects Stebbings’ suggestion that this simply proves superior company picking: trusted managers have already identified the opportunities.
2. The payout obligation constrains private-market ambition
- Wilson considers public and private ownership economically similar over 10–20 years; marking frequency should not determine investment merit. Stebbings pushes back that 12–15-year fund lives make liquidity a fundamental difference. Wilson agrees liquidity drives the roughly 50/50 target, while arguing that Stripe going public should not suddenly change its economic classification. Privates reached 70% at fiscal year-end, before substantial positions became liquid.
- The constraint is arithmetic: a roughly 5% annual payout, smoothed over a rolling five-year average, takes 10% from the liquid portfolio when only half the assets are liquid. Private realizations cannot be relied upon precisely when needed. Without payouts, Wilson would probably target 70–80% privates; given freedom to change one policy, he would probably smooth distributions over a longer period.
- His preference reflects experience rather than a permanent allocation rule: private investments compounded in the low-to-mid-20s over the past decade, versus roughly 10% for public equities. He finds private opportunities less picked over. Today, however, proceeds are being reinvested in public markets, particularly sectors WashU considers “relatively protected, if not immune, from AI.”
3. Direct investing requires company work and proactive access
- WashU’s generalist team spends 80–90% of its time meeting individual companies, reversing the manager-heavy schedule Stebbings describes among typical LPs. Partner conviction is the starting point, not sufficient underwriting: WashU must independently verify the thesis, monitor the investment and consider it the best marginal use of capital. Wilson estimates it executes closer to 0.1% of opportunities reviewed than 1%.
- Knowing where the team lacks an advantage matters. If it cannot distinguish a company from five competitors, it declines. Seed and Series A are generally difficult because WashU wants customers to interview, a product to understand and unit economics it can assess. “We’re not good super early stage investors.”
- Access is proactive: ask a manager for their five most exciting companies, meet them and pursue a next-round lead, employee tender or direct cap-table position. Wilson sees “a huge negative selection bias” in waiting for an IR team to offer an SPV allocation. WashU has done substantially better with this offensive approach than with incoming co-investment opportunities.
- Directs and co-investments constitute roughly half, perhaps more, of privates; Wilson would take that to 80%, subject to a natural equilibrium with manager relationships. Meaningful conviction means willingness to invest 1% of the total portfolio—now close to $200 million—although capacity often requires starting around $50 million and adding over time. A small fund commitment can make sense if it opens larger underlying company opportunities.
4. Concentration needs governance and a willingness to add
- SpaceX exceeded 20% of the portfolio at one point. Wilson credits Lee Fixel, the board and Adam for supporting the position; continuation vehicles helped WashU retain shares through the IPO despite liquidity opportunities. Cost limits are different: investing more than 2–3% in one name would probably require board notification; Wilson thinks 4% requires approval. Allowing appreciation to compound depended on supportive governance.
- Half the portfolio now sits in its top 50 underlying positions, with the top 30 intended to drive performance. Wilson argues diversification benefits diminish rapidly beyond 20–30 names. He also considers fund-level concentration limits mostly artificial: a manager representing 2% of WashU with half its capital in one company creates only a 1%, or 100-basis-point, university exposure.
- His biggest mistake was not adding to winners already above 1–2% of the portfolio. An investment made at a $2 billion valuation could still be the highest-conviction opportunity at $5 billion, yet the higher price and larger position discouraged additions. “That was my mistake, and it’s cost us billions of dollars.”
- Concentration has also hurt: substantial investments in automated and additive manufacturing have been painful. “Maybe we’re just poor at underwriting manufacturing companies.” Wilson says neither story is necessarily finished, but acknowledges the pain—and elsewhere admits adding capital to businesses WashU probably should have let die.
- Wilson thinks the direct-investing model scales better than a traditional fund-of-funds: adding managers makes each one matter less. WashU is still probably over-diversified, but he hopes to preserve its concentration even if the portfolio doubles in size.
5. Manager fit and valuation discipline outweigh brand
- WashU backs billion-dollar-plus funds but none above $10 billion, with a strong—not absolute—bias against that scale. “They don’t need our capital,” Wilson says of mega-funds. Stebbings counters that Thrive’s OpenAI and Cursor investments demonstrate unprecedented outcome expansion. Wilson concedes trillion-dollar-plus private outcomes make venture returns possible even in very large funds.
- His structural objection remains: doubling fund size requires larger checks, typically at higher valuations, or more investments, whose quality falls below the best ideas. AI’s capital intensity creates deployment capacity but leaves eventual returns on that expenditure unclear. Wilson would prefer a $1 billion fund every 18 months to $10 billion every three years; fast deployment is acceptable when expected returns justify it.
- Sequoia’s evergreen structure did not fit WashU’s co-investment model, prompting a board discussion and a decision to step away. Stebbings protests: “You don’t fricking turn down Sequoia.” Wilson concedes, “I’m sure it was a mistake if you look at the returns they’ve generated,” but adds, “we’ve done just fine.” He refuses commitments bundled into unwanted strategies.
- Asked how many funds WashU adds, Wilson says they added one last year and expect to add one next year; adding one or two is generally “one in, one out,” though not a hard line. Valuation discipline is the leading reason to leave. Wilson recalls one partner and says he is not sure the partner found a 100-times-revenue deal he was willing to turn down; such a deal would be exciting, but not the use of capital WashU wants. Stebbings objects that future winners often look expensive. Wilson cites Lee Fixel as pragmatic: willing to pay for exceptional quality but able to step away when things get too frothy.
6. Private premiums and interchangeable models complicate AI returns
- Wilson sees excessive capital chasing late-stage private champions. He thinks Anthropic could have raised above $2 trillion privately, but is not convinced public markets would assign the same valuation; he makes a similar comparison for OpenAI. With public markets near all-time highs, weak DPI reflects a pricing mismatch: “It’s just you can’t go out public at the valuation that the private markets will give.” He points to Aura’s attempt to go out, Cerebras trading poorly after its IPO, and SpaceX being close to its IPO price as examples of the gap.
- Stebbings asks why strong companies should list when private financing and employee tenders already work. Wilson accepts reasons to remain private but stresses public-market depth: SpaceX raising “85 billion or whatever” at IPO would be difficult to replicate privately. Sovereign funds, pensions and SoftBank can write only so many enormous checks; he thinks many such capital sources have been exhausted.
- On AI businesses, Wilson thinks “probably half of these companies don’t exist 10 years from now”; moats and eventual profit pools remain unclear. His test: give the top five language models identical interfaces, and users probably could not distinguish them for 99% of tasks. “I can’t describe another industry like that that has a big, deep profit pool.”
- Enterprise adoption initially favors expensive frontier models, he argues, but rising usage prompts buyers to discover cheaper models suffice for most tasks. Stebbings asks why WashU has not sold its Anthropic exposure. Wilson explains it sits inside partner funds containing roughly 40 other investments, so they would not sell that one holding in isolation; he values those relationships and could hedge after listing if the price warrants it.
7. Exits depend on valuation, position size and lessons from missed sales
- WashU still holds NewBank almost five years after its distribution, while oversized SpaceX requires gradual sales over multiple years despite perceived 5–10-year upside. “It’s better to be not stupid than clever”: supporting the university requires taking some chips off the table. A much smaller SpaceX position, Wilson says, could simply be held for another decade.
- Cerebras, sourced through Eclipse, was large enough at IPO to justify trimming and hedging before distribution. Wilson finds its roughly $25–30 billion valuation difficult to justify from trailing-12-month results; investors are underwriting delivery over the next four or five years. He nevertheless retains confidence in Andrew and the team over the next decade and intends to keep some exposure.
- Wilson put a large package of LP interests up for a secondary sale around mid-2020 but delayed selling to capture anticipated December-quarter write-ups. Available prices were probably 80–90 cents on the dollar: “We were just essentially being greedy.” He recalls being up, he thinks, 65% during 2021, then writing back down a bunch of those gains; waiting for a better exit forfeited attractive liquidity.
- Wilson expects a shakeout but offers no timing. His cyclical example: a chip or memory product moving from $3 to $30 sends incremental revenue to profits, making peak earnings look cheap. Some historically poor businesses will eventually become poor businesses again; “how long the cycle actually lasts, I think that’s anybody’s guess.”
8. Europe and China reward selective underwriting at different prices
- Recent European activity includes an unnamed Spanish fintech through a partner. Wilson finds good companies, including outside technology, whose growth and unit economics would attract large revenue multiples in the US but command small earnings multiples in Europe. Regional market weakness does not, in his view, eliminate high-quality individual businesses.
- Stebbings’ objection: if Europe always trades at a discount, perhaps that is simply its price. Wilson separates price from value: buying at 100 something he thinks worth 300 and compounding strongly beats paying five times the multiple for comparable growth. “Price is what you pay, value is what you get.” He acknowledges valuation remains subjective.
- China exposure is not large, but probably exceeds peers’, and WashU has added capital. Scott says they have been active since Jamie Dimon called China uninvestable. Stebbings challenges political control and the destruction of education-sector businesses. Wilson favors politically supported or below-the-radar activities and counters with developed-market intervention: he cites UK offshore oil-and-gas assets facing tax rates rising from 25% to 75%, calling that “basically confiscating half the company,” and similar changes in Norway.
- WashU has two permanent staff in Shanghai; Wilson says investors “certainly have to invest through local partners.” Stebbings questions the roughly 1:9 dollar/RMB split and restrictions on chip investments. Wilson acknowledges agency problems but argues entrepreneurs mostly prefer dollars, with VIE structures enabling offshore assets. Recent activity favors co-investments and one-offs, with particular attention to companies’ ability to return capital to shareholders.