Holding the Winners: Scott Wilson on the Research and Restraint Behind WashU’s Concentrated Endowment

1. Guest Background

Scott Wilson, chief investment officer of the WashU endowment, manages public and private investments through external managers, direct holdings and co-investments. In this approximately 57-minute episode of 20VC, Harry Stebbings asks how that combination works at institutional scale. SpaceX supplies the striking concentration example, but the conversation extends well beyond a successful position: it covers portfolio reconstruction, spending obligations, manager discipline, AI valuations and decisions Wilson wishes he had made differently.

Wilson traces his approach to Grinnell, where he worked under David Clay. He recalls arriving without a clear understanding of an endowment’s daily work and learning within a portfolio that combined external partners with unusually concentrated direct and co-investment positions. At WashU, he says, the essential method remains recognizable: identify capable partners, independently examine their most distinctive opportunities, commit additional capital selectively and allow successful businesses time to compound.

The scale makes the institutional context essential. Wilson reports starting with just under $8 billion and managing a pool now just below $20 billion. Separately, he says his nine years at WashU generated roughly $17 billion in profits and supported approximately $5 billion of university distributions. These are his interview statements, without an independently supplied performance reconciliation. Assets under management, investment profits and university payments describe different quantities; the change in the asset pool alone cannot establish an investment return.

2. What the Episode Covers

The interview begins by correcting an appealing turnaround anecdote. Wilson did not immediately sell more than three-quarters of WashU’s portfolio during his first week. He says he submitted redemption requests covering over half of it, following several months of preparation. The board wanted meaningful performance improvement within three years, making an early start necessary. Removing 36–37 hedge funds and using private-market secondary sales eventually produced more than 75% turnover over approximately 18 months. The distinction matters: deciding to exit, requesting liquidity and completing a transaction are separate events.

After the restructuring, Wilson approached company analysis and portfolio allocation as related but different tasks. A business does not become a fundamentally different research subject simply because it lists publicly. Over a ten- or twenty-year horizon, he argues, valuation frequency should have limited influence on the assessment of the business. Stebbings presses the more consequential distinction: access to cash. Wilson agrees that liquidity requirements drive his broad half-public, half-private target.

The episode’s spending illustration makes that constraint concrete. A university payout equal to 5% of total assets consumes the equivalent of 10% of the liquid portion when only half the portfolio is liquid. Expected private realizations cannot be relied upon to arrive precisely when needed. Wilson’s preference for 70–80% private assets therefore describes a hypothetical portfolio without those liquidity obligations, rather than an unconditional allocation recommendation.

His return comparison helps explain the preference. Wilson reports private compounding in the low-to-mid twenties over the preceding decade, against approximately 10% for public equities. The supplied interview does not include the methodology needed to verify that comparison independently. More revealing operationally is how his team spends its time. Direct and co-investments represent roughly half or more of private assets; everyone is a generalist, and Wilson estimates that 80–90% of team time goes to meeting individual companies.

External managers remain central to that work. They provide an already filtered starting point, after which WashU conducts its own verification. Wilson says the portfolio would have been around peer average without direct and co-investments, a retrospective attribution rather than a demonstrated controlled comparison. The team also initiates opportunities: it asks managers for their five most exciting companies, visits them and explores later financing rounds, employee tenders or direct ownership. Wilson reports better results from this approach than from waiting for incoming co-investment offers.

He also identifies a capability boundary. Seed and Series A investing are not the team’s strengths; customer conversations, an understandable product and assessable unit economics help it develop conviction. Nor does a long-term preference for private investments dictate every current purchase. Wilson says private assets approached 70% at fiscal year-end, but subsequent realizations are being reinvested publicly, including in sectors outside AI and its infrastructure that he considers relatively protected from AI disruption.

3. Core Views: Reasoning, Examples, and Limits

The foundation of Wilson’s concentration argument is the ability to decline. Before investing directly, WashU must understand and independently verify the thesis, underwrite the company, monitor it afterward and judge it the best marginal use of capital. An inability to distinguish a business from five competitors is grounds to pass. Wilson estimates that actual investments are closer to 0.1% of reviewed opportunities than 1%. That is an estimate of selectivity, not evidence that a particular rejection rate produces superior returns. His painful automated and additive manufacturing investments also show that substantial conviction can coexist with disappointing results. Their outcomes remain unfinished, and the manager network constrains the opportunity supply.

Concentration then requires the correct unit of measurement. Wilson’s example is a fund accounting for 2% of the endowment, with half that fund invested in one company. The endowment’s exposure is 1%, or 100 basis points. A restrictive fund-level limit can therefore obscure how much risk the institution actually owns. WashU aggregates company positions across vehicles to reveal the underlying exposure. This calculation improves visibility; it does not make a genuinely large position safe.

SpaceX illustrates that distinction. Wilson says it once exceeded 20% of the overall portfolio. The team chose continued ownership despite opportunities for liquidity, using continuation vehicles and board support to extend the holding. He separates the cost initially committed from the position’s later net asset value. Allocating more than 2–3% of portfolio cost to one company generally requires board notification, he says, while around 4% requires approval. A position growing through appreciation is consequently different from deciding to commit one-fifth of the institution’s money at inception. Both require governance, but they present different decisions.

His reflections on adding capital expose a harder tension. Wilson considers insufficient investment in winners a major mistake: a company’s valuation rose from $2 billion to $5 billion, or its position had already become substantial, and the team stopped buying despite continuing conviction. He estimates billions in forgone gains. That is a retrospective opportunity-cost claim, not realized profit or a reproducible calculation supplied in the episode. He also acknowledges funding companies that should have been allowed to fail. The defensible principle is to reassess prospective returns at today’s price; neither a rising share price nor previous commitment establishes the answer.

Manager selection applies the same reasoning to repeated behavior. WashU rejects bundled commitments and has no mandatory three-fund policy. Wilson’s example of a manager apparently unable to reject transactions at 100 times revenue describes a pattern of valuation indiscipline, rather than proof that every expensive company is a poor investment. Deployment speed is similarly subordinate to expected value. Rapid investment can be acceptable when opportunities justify it; moving slowly cannot rescue unattractive economics. His preference for voluntarily smaller funds follows the arithmetic of a successful investment contributing meaningfully to the whole vehicle.

Stebbings challenges the corresponding skepticism about megafunds by pointing to extraordinary AI outcomes. Wilson accepts that exceptional company values can support large funds, while arguing that doubling a fund usually requires bigger checks at higher prices or additional, weaker investments. This is a mechanism that may make historical multiples harder to repeat, not a demonstration that large funds must fail. His AI concerns also distinguish adoption from durable profits: as usage costs rise, he observes, businesses may switch from frontier models to cheaper alternatives sufficient for their tasks. His predictions that half of relevant companies could disappear within ten years, or that models would be indistinguishable for 99% of tasks behind one interface, are judgments without supporting tests in the supplied evidence.

Execution further complicates an investment opinion. WashU’s Anthropic exposure sits inside partner funds holding many companies, preventing convenient sale of that single position. Wilson’s explanation of weak private distributions through private-public valuation gaps is likewise his interpretation. His exit examples resist a universal schedule: Cerebras was partly reduced and hedged around its IPO while some shares remained; Nubank was retained nearly five years after distribution; SpaceX warrants multiyear reductions because of its size. Finally, geographic opportunity remains company-specific. He sees selected European valuation advantages but concedes subjectivity, while his China approach depends on local partners, policy fit and shareholder cash returns. A regional discount does not complete the underwriting.

4. Learning and Application

An institution applying this episode should begin with a cash exercise. Put required spending against available liquid assets, then examine what happens if private distributions arrive later than expected. Keep anticipated realizations separate from cash already available. The interview’s 5% payout against a 50% liquid allocation shows why total-asset percentages can conceal the burden placed on the liquid portfolio. Wilson’s preference for longer spending smoothing also introduces a genuine tradeoff: changing distributions affects the university’s mission and budgeting, while incoming gifts influence sustainability. Investment flexibility cannot be the only objective.

The next practical tool is a company exposure register across funds and direct positions. Record invested cost, current value and means of obtaining liquidity separately. Use that register to distinguish decisions about fresh capital from decisions about appreciation-driven concentration, with explicit board notification and approval arrangements. Wilson’s willingness eventually to commit 1% of the portfolio is a useful question about conviction, but his roughly $200 million scale should not become another institution’s automatic position limit. Starting with $50 million and building over time describes his method of developing meaningful exposure. His target of half the portfolio in the top 50 look-through holdings similarly needs to be assessed against an institution’s own capacity and obligations.

Every proposed follow-on investment can then receive a fresh underwriting note: what does the current price assume, what will additional capital accomplish, what evidence would invalidate the thesis, and can the team monitor the business? This process addresses both mistakes Wilson acknowledges—anchoring to the original purchase price and continuing to fund deteriorating companies. A strong manager recommendation does not substitute for internal competence. Without the ability to verify and follow the investment, concentrated ownership imports responsibilities the institution may not be equipped to meet.

For sales, compare an executable offer with the uncertain benefit of waiting. Wilson’s secondary-sale account is especially useful: the team delayed during 2020–2021 to capture expected valuation increases, then regretted missing an opportunity at 80–90% of book. He reports a roughly 65% portfolio rise in 2021 followed by markdowns. A sale memo can specify an acceptable price, the reason for waiting and a date for reconsideration. A discount to reported value does not by itself establish a bad transaction. Equally, this retrospective example cannot prove that selling sooner is always preferable.

Partner selection should also examine structural compatibility. Wilson respected Sequoia’s investors but found its evergreen arrangement poorly suited to WashU’s co-investment model, taking the issue to the board. Before committing, an institution can therefore ask whether information access, additional company investment and eventual disposition support its own process, alongside assessing the manager’s investment ability.

Finally, AI-related underwriting can separate operating progress from favorable cyclical pricing. Wilson’s hypothetical increase in chip or memory prices from $3 to $30 illustrates how profits can expand without establishing permanent economics. Test whether the thesis survives lower selling prices rather than treating current earnings as automatically durable. Wilson explicitly cannot time the shakeout he expects, so the example supports scenario analysis rather than a trading signal. These applications organize investigation; his reported performance and market forecasts remain inputs to assess, not substitutes for asset-specific evidence.

Source

• Original episode: Inside WashU's Endowment Fund | SpaceX Made Up 20% of Out AUM | Scott Wilson