Harvard Management Company's third-quarter 2025 filing showed 6,813,612 shares of BlackRock's iShares Bitcoin Trust, worth about $442.9 million. That made the bitcoin ETF 21.04% of the portfolio in that filing, and its largest single holding. The 21% is the number that travelled.
Harvard's endowment closed fiscal 2025 at $56.9 billion. So the same position was roughly 0.8% of the endowment. Both figures are accurate. Only one of them is an allocation.
That gap is the problem with reading institutional crypto adoption off headlines. A Form 13F covers only section 13(f) securities: exchange-traded equities, certain options and warrants, closed-end funds and some convertible debt. It leaves out private equity, hedge funds, real assets, bonds and cash. For an endowment that runs most of its money in private strategies, a 13F is a keyhole. Harvard's entire filing came to $2.105 billion across 18 positions, against a $56.9 billion pool.
What the filings actually show
The State of Wisconsin Investment Board is the other case worth reading slowly. Its February 2025 filing reported 6,060,351 shares of the same bitcoin ETF, worth roughly $321.5 million. Its May 2025 filing reported none. The whole position went during the first quarter of 2025.
Wisconsin was the fund cited through 2024 as proof that public pensions had arrived. It bought, it added, and it left, inside about four quarters. Harvard's position moved quickly too: roughly 1.9 million shares at the end of June 2025, then 6.8 million three months later. That's a 257% increase in one quarter.
Neither pattern looks like a strategic allocation with a policy weight and a rebalancing band. They look like positions. And a 13F is a snapshot taken at one quarter end. It says nothing about the weeks in between, and nothing at all about what a fund holds outside listed securities.
What the surveys actually asked
The most quoted institutional crypto survey is the one EY-Parthenon runs with Coinbase. The 2025 edition polled more than 350 institutional investors. It reported 76% already invested in spot crypto or spot crypto exchange-traded products, and 83% intending to increase allocations. The 2026 edition, again more than 350 investors and fielded in mid-to-late January 2026, reported 73% planning to increase: 5% significantly and 68% by less.
Read the respondent list before the percentages. EY describes the sample as asset managers, asset owners, family offices, private banks, hedge funds and VC firms. Pension funds and endowments sit inside "asset owners", one of six categories. The published summaries I could open don't split the headline numbers by category, don't give a response rate, and don't describe how the sample was recruited.
That matters, because this isn't a census. It's a panel of firms willing to answer a digital-assets questionnaire run by a crypto exchange and a consultancy. Nobody's hiding that. It's on the page. It just doesn't survive the trip into a headline reading "institutions are piling in".
It's worth comparing that with how a regulator writes up a survey. The FCA's 2025 cryptoasset research carries a "survey limitations" section on page six. It states the online panel may overstate ownership, names digital exclusion as the reason, and flags response bias. It gives base sizes under every chart. That's the standard the institutional surveys are being read against, and they don't meet it.
There's a timing wrinkle as well. EY describes the 2025 fieldwork as taking place just after the US election and before the executive order on digital assets. That's a narrow and unusually bullish window. Stated intentions from that window aren't positions held a year later. Wisconsin's round trip happened inside roughly the same period.
Where crypto sits in a real institutional portfolio
The two best allocation datasets don't carry a crypto line at all.
The 2025 NACUBO-Commonfund Study of Endowments covers 657 US colleges, universities and affiliated foundations holding $944.3 billion, for the year to 30 June 2025. On a dollar-weighted basis the average endowment held 45% in alternative strategies, 31% in publicly traded equities, 11% in fixed income, 10% in real assets and 3% in "other". Crypto, wherever it sits, is inside that 3%, alongside everything else the study declines to name. The same cohort returned 10.9% in FY25 and 7.7% a year over ten years. Whatever produced that, it wasn't a digital-asset sleeve.
The Thinking Ahead Institute's Global Pension Assets Study for 2026 covers 22 markets holding $68.3 trillion, of which $62.0 trillion sits in the seven largest. Their average allocation at the end of 2025 was 48% equities, 31% bonds, 19% other assets and 3% cash. Equities are down nine percentage points over twenty years; "other" is up six. The summary edition I read doesn't mention crypto, bitcoin or digital assets once.
There's a second lesson buried in the NACUBO tables. "Institutional" isn't one constraint set. Endowments under $50 million held 60% in publicly traded equities and 8% in alternative strategies. Those over $5 billion held 25% and 51%. The median participant ran $253.6 million. The small ones look far more like a household portfolio than like Harvard, and they aren't the ones running direct crypto custody either.
Absence isn't proof of zero. Both studies bucket a great deal into "other". But if $68 trillion of pension money had a meaningful digital-asset sleeve, the people who publish the allocation tables would have added a row. Flow reporting has the same problem in a different shape, which is why it's worth knowing what CoinShares crypto fund flow data really measures before reading a weekly inflow number as adoption.
Four constraints a household doesn't share
Horizon and spending rule. Harvard distributed $2.5 billion from its endowment in fiscal 2025, about 4.4% of the prior year's value, funding 37% of operating revenue. Across the NACUBO sample the average effective spending rate was 4.9%, up from 4.8%. That's a smoothed, committee-set number. An endowment isn't perpetual because it never spends. It's perpetual because it spends a rule rather than a need. A household's withdrawal is a need, and it arrives whether or not the risk asset is down 70%.
Governance. Some 46.2% of NACUBO respondents used an outsourced chief investment officer in FY25. The rest run investment committees, policy statements and delegated mandates. A 0.8% position gets sized, minuted and reviewed by people paid to do it. That machinery is the reason the number stays at 0.8%, and it's why a written policy set before the trade matters more than the trade. The evidence on pre-committing through an investment policy statement points the same way.
Custody. The first UK defined benefit scheme to allocate to bitcoin did so in October 2024 on Cartwright's advice, at 3% of assets, holding the asset directly rather than through a fund. Cartwright describes trustee training, a due diligence process and a custodial arrangement designed around securing the asset. Set that beside the FCA's picture of UK households: 74% of crypto owners leave their holdings on the exchange they bought from, and 13% use cold storage.
Tax. An endowment is tax-exempt. A UK household isn't. HMRC treats selling, exchanging or otherwise disposing of a cryptoasset as a capital gains event, and the annual exempt amount is £3,000. Swapping one coin for another counts. So a rebalancing rule that costs an endowment nothing can cost a household a tax event every time it trims.
The case against this reading
The strongest objection is that this is too cynical. Institutional adoption, on that view, is a genuine signal of maturation. Regulated custody, listed products and published risk frameworks are real things, not governance theatre.
There's evidence for it. BlackRock's Investment Institute published a portfolio-construction case in December 2024 that argues that a 1% to 2% bitcoin weight in a multi-asset portfolio is defensible. The reasoning is a risk budget rather than a price target. At 1% to 2%, bitcoin contributes 2% to 5% of total portfolio risk, which they compare to holding one average Magnificent Seven stock at index weight, near 4%. At a 4% allocation the contribution jumps to about 14%, and bitcoin starts driving the portfolio.
That's a serious framework, and it's the same logic used to size a strategic gold allocation. It comes with the authors' own caveats, which the coverage tends to drop. Since 2015 bitcoin has had six drawdowns of more than 70%. And they flag the reflexive problem: if broad adoption is the thesis, broad adoption also removes the structural catalyst for further large gains.
Both things can be true at once. The infrastructure improved. What hasn't followed is the weight. A risk-budgeted 1% to 2% sleeve is a defensible institutional position, and it is also a very long way from the impression left by "83% intend to increase allocations".
What a household is actually deciding
The FCA's 2025 research found 8% of UK adults held cryptoassets, down from 12% in 2024. Among owners, 33% said they bought as part of a wider investment portfolio and 30% said they bought as a gamble that could make or lose money. That gambling share rose four points on 2024, though it's still below the 40% peak in 2022. Twenty-one per cent held between £1,001 and £5,000, and 11% held between £5,001 and £10,000.
The caveats belong next to those figures. The FCA's survey is online-only, run through a YouGov panel: 2,353 nationally representative interviews between 5 August and 2 September 2025, plus a boost sample of 1,053 current or former owners. The FCA says the online method may overstate ownership, because digitally excluded adults are less likely to be able to buy these products, and that panel response bias is possible.
The operational risk sits in the same survey. A quarter of UK crypto users say they've encountered suspicious activity, and 7% say they've personally been a victim. Nine per cent funded a purchase with a credit card or another credit facility, down five points on 2024. None of those lines has an institutional equivalent. A pension trustee doesn't buy on an overdraft, and doesn't lose the position to a phishing email.
Read against the institutional data, the household picture is close to a mirror image. Institutions run tiny weights with heavy machinery around them. Households run larger relative weights with almost none. So the question that transfers isn't "what do institutions allocate". It's what someone can actually hold through a 70% drawdown, which is a question about risk capacity rather than a risk-tolerance score.
What would change the conclusion
Three things would.
First, a named line item. If NACUBO or the Thinking Ahead Institute adds a digital-assets row to its allocation table and it reads above roughly 1%, that's a change in institutional practice rather than a survey artefact. Neither publishes one now.
Second, persistence. Wisconsin's round trip and Harvard's 257% quarter both look tactical. Positions that survive three or four years of quarterly filings, including a drawdown, would say something different about intent.
Third, a survey with published methodology. A sample frame, a response rate, and results split by institution type, with pension funds and endowments reported separately from hedge funds and VC firms. Several hundred respondents is plenty for a useful read, provided you know who they are.
Until one of those turns up, the short version stands. Institutions built the plumbing. A handful took positions worth well under 1% of assets. At least one has already sold out entirely. And the conditions that make a small sleeve cheap for an endowment, a tax exemption, a smoothed spending rule, a committee and a professional custodian, are precisely the ones a household is missing.