Every quarter, the world's biggest money managers are legally required to tell the SEC exactly what they own. Most people never read these filings. That's a mistake, because the one Harvard just submitted says something its headline coverage completely missed.

Harvard Management Company, the entity that runs Harvard University's endowment, now holds more money in gold than it holds in Bitcoin. Not as a rounding error. Not as a temporary blip. As a clear, disclosed, dollar-for-dollar fact sitting in a public filing that almost nobody outside of finance trade sites has actually read.

Here's why that matters more than another "institutions are buying Bitcoin" headline.

What Harvard's filing actually shows

Direct answer: Harvard held its Bitcoin ETF position steady in Q2 2026 at 3,044,612 shares worth $101.4 million, after two straight quarters of aggressive selling, while its combined gold holdings reached $171.2 million, a gap of roughly $70 million in gold's favor.

That pause in selling sounds like good news for Bitcoin bulls. It isn't quite that simple.

Look at where Harvard started. In September 2025, the endowment held 6,813,612 shares of BlackRock's iShares Bitcoin Trust. Over the following two quarters, it cut that position by 21%, then by another 43%. By the time Q2 2026 rolled around, more than half the original position was already gone. "Holding steady" this quarter isn't a comeback story. It's a pause after a retreat, and the position lost another $15.6 million in value during the quarter purely because Bitcoin's price fell, not because Harvard bought or sold a single share.

Meanwhile, Harvard's gold exposure told the opposite story. The endowment held $149.5 million in the iShares Gold Trust and another $21.7 million in the SPDR Gold Trust, for a combined $171.2 million. That's roughly $70 million more than its Bitcoin position. For a portfolio manager choosing between two classic inflation hedges, that gap is a statement.

One more detail rarely mentioned: Harvard didn't just deprioritize Bitcoin. It exited Ethereum entirely. An $86.8 million spot Ethereum ETF position from the prior filing simply doesn't appear anymore.

To be fair to Harvard's Bitcoin position, context matters here too. It ranks as just the 11th largest disclosed holding in a portfolio where SpaceX alone, now worth $2.21 billion after its June IPO, makes up 52% of everything publicly reported. And "publicly reported" is the key phrase. Harvard's total endowment sits around $57 billion, and the vast majority lives in private investments that never show up in a 13F at all. So this isn't Harvard declaring war on crypto. It's one visible slice of a much larger, mostly hidden portfolio, and that slice just got smaller relative to gold.

The institutions that refused to blink

Direct answer: While Harvard trimmed and Morgan Stanley cut its position, Abu Dhabi's two largest sovereign investors and JPMorgan moved in the opposite direction, holding steady or actively increasing their Bitcoin ETF exposure through the same quarter.

Consider the scenario from the other side of the trade. Imagine you're managing a $290 billion sovereign wealth fund with a multi-decade time horizon. A single quarter's 14% price drop barely registers as noise. That appears to be exactly how Mubadala Investment Company and the Abu Dhabi Investment Council treated it.

Mubadala held 14,721,917 shares worth $490.1 million, unchanged in share count, making Bitcoin its second-largest disclosed 13F position. The Abu Dhabi Investment Council held 8,218,712 shares worth $273.6 million, also unchanged, and Bitcoin remains its single largest reported holding. Combined, the two Gulf institutions control close to $764 million in Bitcoin ETF exposure. Their position's value dropped by roughly $118 million purely on price, and neither fund sold a single share to escape it.

Dartmouth College showed the same pattern on a smaller scale, holding its Bitcoin, Ethereum, and Solana ETF positions completely unchanged even as their combined value slipped from $14.6 million to $12.4 million.

Then there's JPMorgan, which didn't just hold. It added. The bank's disclosed IBIT position rose from roughly 8.3 million shares to 10.4 million, a 25% increase, and it more than quadrupled its Ethereum ETF holdings in the same filing. Morgan Stanley, by contrast, trimmed its IBIT stake by 4.5%, though it partly offset that by building a $43.3 million position in its own newly launched proprietary Bitcoin Trust. Two of Wall Street's largest banks, same quarter, same asset, opposite headline moves.

Why the "institutions are buying" headline is only half true

Direct answer: Aggregate institutional Bitcoin ETF holdings rose 7.5% in Q2 2026 even as total ETF assets shrank and the number of institutional filers fell, which means the "institutions are buying" narrative reflects a smaller group of larger holders rather than broad new demand.

Here's the part most coverage skipped entirely. Institutional Bitcoin ETF holdings climbed from 498,389 to 535,723 BTC equivalent, a 7.5% increase, and that number got repeated everywhere as proof of growing institutional confidence. But total spot Bitcoin ETF holdings across all owner types actually fell from 1,297,010 to 1,211,322 BTC, a 6.6% drop. And the number of institutions filing 13Fs with any Bitcoin exposure dropped too, from roughly 2,000 to roughly 1,900.

Put those three numbers together and a different picture emerges. Retail and smaller holders appear to have sold harder than institutions did. That mechanically pushed the institutional share of a shrinking pie up to a record 44.2%, from 38.4% the previous quarter. That's not necessarily new conviction. It can just as easily be math: when the denominator shrinks faster than the numerator, the percentage rises even if nobody's genuinely more bullish.

The hedge fund filings that don't mean what they look like

Direct answer: Several of the largest reported Bitcoin ETF share-count drops, including Citadel and Macquarie, reflect options books and market-making inventory rather than straightforward bearish bets, making headline percentage changes misleading without deeper context.

One common mistake, understandable but costly, is reading a hedge fund's falling share count as a bearish signal on its own. Citadel's ordinary IBIT position dropped 59.66% to 514,614 shares. Read in isolation, that looks like a retreat. But Citadel's filing also shows call options tied to about 24.65 million underlying shares and put options tied to roughly 17.86 million, an options book that dwarfs the "sold" headline number by more than 40 times over. Characterizing that as a simple bearish trade means ignoring most of the actual position.

Macquarie's ordinary shares fell 61.78% to 1,581,934, worth about $52.7 million, but Macquarie also functions as an authorized participant, meaning it creates and redeems ETF shares as part of its market-making role. Jane Street sits on the opposite end of the same mechanism. Its reported position swung the other way, rebounding from 5.87 million shares to 24.88 million, worth roughly $828.2 million, which sounds dramatically bullish until you remember it's also an authorized participant whose inventory naturally fluctuates with trading flow.

Tudor Investment Corporation, Paul Tudor Jones's fund, added 109,446 shares for an 18.9% increase, worth $22.9 million. But the same filing discloses $23.8 million in put options and $4.93 million in call options sitting alongside that position, which points to a hedged approach rather than a straightforward bullish bet. Brevan Howard, for its part, cut most of its ordinary IBIT shares outright, one of the quarter's clearer directional retreats.

Bloomberg Senior ETF Analyst Eric Balchunas has flagged this exact problem publicly: 13F filings don't separate proprietary trading from client positions, hedging, or pure inventory, and they exclude short positions and most private holdings entirely. The stronger signal, in his view, comes from actual share-count changes rather than headline dollar figures, and even that requires knowing who's filing and why.

A simple framework for reading the next 13F season

Instead of asking "are institutions bullish or bearish," it helps to sort holders into three buckets.

• Sticky holders: endowments and sovereign funds like Harvard, Mubadala, Abu Dhabi Investment Council, and Dartmouth. Their share counts barely move quarter to quarter, so a change here signals genuine conviction, in either direction.
• Tactical holders: banks and asset managers like JPMorgan and Morgan Stanley, who actively adjust exposure and disclose clearer directional intent, even if the mechanics differ between them.
• Noisy holders: market makers and heavily hedged funds like Jane Street, Citadel, Macquarie, and Tudor, whose raw share-count swings often reflect options structuring or authorized-participant inventory more than a simple buy or sell decision.

Read through that lens, the Q2 2026 season looks less like a unified institutional verdict on Bitcoin and more like three separate experiments running at once, with Harvard's gold preference standing out as the one genuine surprise.

Key takeaways

• Harvard's Q2 2026 filing shows $171.2 million in gold versus $101.4 million in Bitcoin, a roughly $70 million gap that only appeared after two prior quarters of heavy Bitcoin selling.
• Abu Dhabi's Mubadala and Abu Dhabi Investment Council held a combined $764 million in Bitcoin ETF exposure completely unchanged through the same quarter.
• JPMorgan increased its Bitcoin ETF stake by roughly 25% and quadrupled its Ethereum ETF holdings, while Morgan Stanley trimmed its ETF position but built exposure through its own proprietary Bitcoin fund instead.
• The widely repeated "institutional holdings rose 7.5%" statistic is real, but it coincides with a 6.6% drop in total ETF assets and a shrinking number of institutional filers, meaning it reflects concentration among fewer, larger holders more than fresh broad demand.
• Several of the quarter's biggest reported "sells," including Citadel and Macquarie, are better explained by options books and market-making activity than by simple bearish conviction.

FAQ’s

What is a 13F filing?

It's a quarterly disclosure the SEC requires from institutional investment managers overseeing more than $100 million in assets, listing their U.S. equity and ETF holdings as of quarter-end.

Did Harvard sell all its Bitcoin?

No. It held its position steady in Q2 2026 after cutting it sharply over the two prior quarters. It still holds $101.4 million in Bitcoin ETF shares, just less than its $171.2 million in gold.

Are institutions collectively bullish or bearish on Bitcoin right now?

Neither, cleanly. Sovereign wealth funds and JPMorgan held or added exposure, Harvard and Morgan Stanley trimmed, and several hedge fund filings reflect options activity rather than a directional bet at all.

Should I change my own Bitcoin holdings based on this data?

This article is for informational purposes only and isn't financial advice. 13F data offers a partial, delayed picture that excludes private holdings, short positions, and true intent behind options-heavy filings, so it's worth treating as one input among many rather than a signal to act on directly.

DISCLAIMER

This article is based on publicly available Q2 2026 SEC 13F filings and third-party reporting believed to be accurate as of publication. 13F filings disclose only long U.S.-listed equity and ETF positions at quarter-end; they exclude private investments, short positions, options intent, and holdings outside U.S. jurisdiction, so they offer a partial and delayed picture of any institution's true exposure. Nothing in this article constitutes financial, investment, or tax advice. Cryptocurrency and precious metals markets are volatile, and past institutional behavior is not a predictor of future price movement. Always conduct your own research and consult a licensed financial advisor before making investment decisions.