Summary

Morehead’s central thesis is that a fund multiple is meaningless without a timeline — what endowments need is velocity of capital. A 15X return over a 15–18-year VC fund can lose to three sequential 6-year 3X growth-equity funds, which compound to 27X — “better than 15X by like a factor of two.” His office rule: “you’re not allowed to talk about returns without also talking about time,” because “if you’re up 5X over 30 years, that’s horrible. And if you’re up 5X in five months, that’s amazing. I guess that’s SpaceX.”
“The single reason that privates exist is to make money, period, end of story” — so Baylor is winding down real assets and concentrating on VC, expansion/growth equity, and buyout. Growth equity is the biggest private allocation, annualizing around 30% against an 8–9% bogey; venture, by contrast, “is a pure diversification play for us.” About 2.5% of Baylor’s endowment is in Anthropic through managers, with no OpenAI or SpaceX exposure.
When software was down 50–60% from October 2025 into early 2026, Baylor began allocating into it based on a human-behavior read rather than a technical one. Morehead called friends running 500-person private businesses and asked if a vibe-coded app would replace their CRM — “not in a million years” — and recalled, with some uncertainty, the Salesforce CEO saying the best AI would be “93% right,” while “the issue with software is 100% right.” He thinks that, in some vertical industries, trusted incumbent software may become “the delivery mechanism for AI.”
The risk discipline is “I never wanna be all in. Things can always get worse.” Baylor allocates mechanically in 10% market-decline increments — down 20%, put roughly 20% to work; down 30%, another 20% — accepting money left on the table to avoid being fully invested before the bottom. Cash is priced at 8.5%: a 3.5% yield plus a 5% opportunity cost, based on the high odds of finding a 20% opportunity within four years. Cash was 15–16% pre-pandemic and is low today.
Tradeable read on AI infrastructure: the scarce asset has migrated from land to powered land to “permitted powered land,” and permitting pushback is the new bottleneck — something that “didn’t exist six months ago.” Data-center sites in Baylor’s book are up 50% in six months, a UK site is valuable “simply because we have a permit,” and power prices may rise until the supply constraint is solved over the next five to seven years. He’s bearish Europe broadly — “defense… Russia… behind on AI, because, because, because” — with macro hedges on European indices.
Manager discipline runs on a baseball-GM analogy: style drift gets you fired regardless of returns. “If I ever walk out on the field and I have two second basemen and no third baseman, the third baseman’s getting fired… I don’t care what your returns are” — though the line is drawn at genuine strategy switches, such as moving from post-product-market-fit companies to “two guys in a garage,” not artificial category lines. Position sizing starts from dollars per company: $2.5–3M in each underlying name for expansion and buyout, so a 5X actually matters; venture is somewhat different.
Baylor spends more time on asset allocation than manager selection: privates target 45% within a 35–55% band sized so a denominator effect never forces selling (“the number one thing to avoid is fraud, and the number two thing to avoid is forced selling”). Baylor supplements commingled vehicles with fund-of-one arrangements so it can dial single-name exposure like NVIDIA up or down. Full-year 2025’s 9.4% return versus Dartmouth’s 10.8% reflected a second J-curve from 60–70% higher private commitments in 2020, 2021, and following years, while a fund-of-one and another asset class were beginning to inflect upward; Morehead expects 18.5–19% this year. The endowment grew from $1.4B to $2.7B.

Deep dive

1. Demographics are a slow-moving train wreck — and they dictate how Baylor invests

Morehead’s starting point is structural: fewer US high-school graduates after the Global Financial Crisis plus visa friction affecting full-pay international students means “a lot of schools across the country did not meet their targets” for the class of 2030 — so for the next 10–15 years, endowment distributions must fill the revenue gap. Baylor began reorganizing around this roughly five years ago.
The office’s historical edge is the downside — flat in Q1 2026 against an S&P down 4%, with similar outperformance in Q4 2018, Q1 2016, and 2012. The five-year project has been fixing the other tail: “the market’s up 70% of the time. If you’re gonna trail to the upside, that’s gonna be problematic.” Finance faculty “actually laugh at me” when he describes trying to win both sides.
The mechanism is fund-of-one arrangements. Commingled funds deliver “the average risk-return profile” needed to keep 100 or 1,000 LPs happy; Baylor instead asks GPs to run the same strategy separately with visibility into the book — so when the next manager wants to add NVIDIA, Baylor can say “we’ve got plenty” or “make it three times as big.” “It’s actually worked exceedingly well over the last two, three years.”

2. Box the privates first — the band exists to prevent forced selling

Baylor spends more time on portfolio construction than manager selection, “which is unique in the space.” Current mix: roughly 45–47% private and 53–55% public. The privates decision comes first because “the private side is gonna suck your liquidity and hogtie your ability to allocate” — set it, box it, and accept that “it’s really, really hard to move a private book around.”
The 35–55% band around the 45% target is engineered so a denominator effect never forces liquidation: “the number one thing to avoid is fraud, and the number two thing to avoid is forced selling. That’s a disaster.” In the last part of 2022, when technology slid, the private side reached roughly 51–52% — uncomfortable but never constraining.

3. Velocity of capital: why 15X can be the wrong answer

The signature critique: funds have stretched from 10–12 years to 15–18, “much to the chagrin of all LPs,” and “it’s not clear to me that the GP incentives are aligned with the math that runs endowments.” A 15X return over 15–18 years can lose to redeploying through three 3X six-year growth funds — 27X. “Students can’t pay their tuition with returns. They have to pay with dollars.”
He sees exactly why GPs hold winners — a 6X looks better in marketing than a 3X, and “that suggests that the next fund will be raised” — but “I’m not optimizing for the best business for the GP. I’m trying to optimize for the biggest pile of money for our students.” When velocity of capital “starts to asymptotically approach wherever it’s going to be,” he wants to move on.
Stebbings’ blunt follow-up — why do VC at all if growth equity gives you 3X in six years? — gets a candid answer: the question is debated internally, but laddering return horizons matters: some returns arrive in 6–10 years, others in 3–5, and others in 1–3. The office rule is that returns may never be discussed without time attached.

4. Venture is diversification; growth equity is the engine

Asked directly whether venture is “just a pure diversification play,” Morehead answers “It is for us” — while noting that about 2.5% of the endowment is in Anthropic through managers, with no SpaceX or OpenAI exposure. He explicitly credits the managers, not himself.
Baylor came late to the brand names — “when you knock on the door, they kinda don’t answer” — so VC lives in “newer upstart-y names,” while “the ladies in our office have had exceptional, absolutely exceptional returns” in expansion/growth equity, annualizing around 30% against an 8–9% bogey. Growth equity also wins on zeros: “if there are fewer zeros, then everything else doesn’t have to cover for the things that don’t work.”
On 2021/2022 mulligan vintages: “that just kind of comes with the territory.” Baylor sets an allocation across PE, expansion capital, and VC, then evaluates whether the overall portfolio clears its expected return hurdle rather than judging one vintage in isolation.

5. The software trade: human behavior beats engineering knowledge

Morehead’s claimed edge isn’t technical — “much of the stuff that comes out of Silicon Valley is over my head, but I do know how people think.” Against the “software is dead, somebody’s gonna vibe code this” narrative, he phoned friends running 500-person private businesses — including a business he thinks may be the only vertically integrated potpourri maker in the world — and asked if they’d tear out their CRM for something unproven: “not in a million years.” He recalled, with some uncertainty, the Salesforce CEO saying the best AI would be “93% right… but the issue with software is 100% right.”
The resulting thesis is that, in some vertical industries, trusted incumbent software may become “the delivery mechanism for AI.” SaaS companies worth $20–50B “aren’t stupid”; they are unlikely simply to let their existing software go to zero.
With software “on sale to the tune of 50, 60% from October of ‘25,” Morehead’s response was: call businesses, hear that the disruption thesis was not true for their operations, and say, “I’ll own that.”
Execution ran through manager Sean Barrett: daily calls for four weeks, trading articles at all hours, and Morehead pushing concentration — “you have this name and another name… Which one has better risk-adjusted opportunity?” His division of labor: “I’m making a decision based on human behavior… but I’m relying on the manager to be expert in their individual field.” That’s the allocator’s job, Buffett-and-Munger style — deciding who gets the incremental dollar.

6. Never all in: mechanical buying and the 8.5% price of cash

The scar-tissue lesson: “whenever you’re trading, you for sure are gonna lose money… sometimes for a long period of time… the takeaway is I never wanna be all in. Things can always get worse.” Even in the software buy, there was no line in the sand — it was down 50–60%, but “who’s to say it’s not gonna be down 70, 80%?”
The mechanism is 10% increments: 0–10% down is “normal stuff” for an infinite-life portfolio; down 20%, put roughly 20% to work; down 30%, another 20%. “The reality is we actually never get all the way invested before it rebounds… we leave money on the table. That’s true” — the payoff is never being fully committed as the decline continues.
Cash is explicitly priced: the odds of finding a 20% opportunity within four years are “really high,” so cash earns its 3.5% yield plus a 5% opportunity cost — 8.5%. Pre-pandemic, finding nothing attractive, Baylor sat on 15–16% cash; today balances are low “because we keep finding 20, 30% annualized things to do.”
On concentration risk in downdrafts: none. “We own everything from sunscreen to helium to technology… we’re infinitely more diverse than the S&P 500. It’s not even close.”

7. Public prices are legitimate; private marks should be conservative

Stebbings pushes on the “casinoization” of public markets — SpaceX at $1.8 trillion on an Elon premium — and Morehead concedes irrationality but not legitimacy: “there are tens of millions of people trading on that information, whereas on the private side there’s, like, three… That doesn’t mean that they’re right. It just means that it incorporates all available information.” The exchange illustrates the private-mark problem with a valuation reset based on a few people saying they tried the product and liked it.
Marks discipline comes from his trading-book past: mispricing corrupts psychology (a position marked $30M but worth $10M makes you refuse a $20M premium bid). Morehead says Baylor’s evidence its marks are conservative is that he thinks gains in the six-to-nine months before takeouts average 60–90%, versus what he thinks is a 30–50% market norm.
On venture as a learning academy: “Not for me. I actually learn a lot from the public-side managers” — citing 2016’s autonomous-car hype: “we’re 10 years on and what do we have, like 50,000 cars on the road? Like, please.”

8. The baseball GM: drift gets you fired, dollars-per-company sizing

The style-drift rule, verbatim: “if I ever walk out on the field and I have two second basemen and no third baseman, the third baseman’s getting fired, full stop… I don’t care what your returns are.” A fully invested equity manager who wakes up with 10% cash is fired — “I don’t wanna be the guinea pig” for an untested macro instinct. But Morehead isn’t policing “artificially generated category limitations”: moving from post-product-market-fit companies to “two guys in a garage” is a firing offense; moving from B to late A is “who cares.” Stebbings, expecting misalignment, concedes: “we’re actually aligned completely.”
Sizing starts from what matters to the endowment: a manager’s 7X company sale returning $400K prompts “What? Who cares.” For expansion and buyout, Baylor now targets $2.5–3M per underlying company — 10 companies means a $30M commitment — so a 5X returns $15M: “that’s enough to matter.” Venture is somewhat different because its company set is larger.
On mega-scale allocators: Notre Dame at $20B says the wall they expected at $10–15B never came, but somewhere before Harvard/UTIMCO scale a $20M check 50X-ing to $1B is only 2% — which is what the a16z-type platform model is tapping into. Big venture platforms per se: “it just gets harder… the law of large numbers.”

9. Permitted powered land, bearish Europe, and the quickfire book

The AI pushback is real at the dirt level: data centers are built “in my neck of the woods, not in Silicon Valley,” and the scarce asset went from land to powered land to permitted powered land as citizens — angry about power and water prices in arid Texas and Arizona — put up yard signs and permitting boards say no to get reelected. Enough projects aren’t happening that “the power companies are coming to those who do have permits and saying, ‘We can get you power sooner than we thought.’” Data-center sites in the office’s book are up 50% in six months; a UK site is valuable “simply because we have a permit.” Morehead says China does not face the same process because authorities “just build it where they need it.”
Europe: flatly not bullish — “the defense structure of it… Russia… behind on AI, because, because, because” — yet Baylor allocates to European long-short managers precisely because there will be winners and losers, while running “some of our bigger macro hedges on European indices.”
Quickfire calls: leaned into software while taking energy length off when crude went north of $100 around the US-Iran war and Strait of Hormuz; added to private-equity sponsors in March or April; private credit is the most overhyped asset class — “credit exposure that looks and acts a lot like equity to the downside, but you don’t have upside equity returns.” Most admired peer: Brown and Jane’s team — “real investors… things that take a lot of courage.” Fund he most wants: Benchmark. Next decade’s excitement: biotech “solving diseases as opposed to simply treating symptoms,” plus navigating the $1B→$5B office inflection. Full-year 2025’s 9.4% versus Dartmouth’s 10.8% reflected a second J-curve from 60–70% higher private commitments in 2020, 2021, and following years, while the fund-of-one and another asset class were beginning to inflect; “we’ll be 18 and a half, 19% this year without any SpaceX or Cerberus.”

AI:AM Highlights: Astra as AGI, OpenAI's Pause, Mythos @ Mozilla & Human Agency vs Technocapitalism