A 13F filing proves exactly two things about a named institution’s crypto exposure: how many shares of a qualifying security it held on the last calendar day of a quarter, and what those shares were worth at that day’s closing price. It does not prove that the institution bought or sold anything with conviction, what it paid for the position, whether the position was hedged, or whose money it actually was.
That obligation comes from Section 13(f) of the 1934 Securities Exchange Act. It requires any institutional manager overseeing $100 million or more in qualifying US securities, with investment discretion, to file Form 13F – a threshold reported by CoinShares – no later than 45 days after each calendar quarter ends, a deadline confirmed independently by Yellow.com and crypto.news. The filer universe is broad: investment advisers, banks, insurance companies, broker-dealers, pension funds, sovereign wealth funds and corporations, per Yellow.com.
Crucially, not everything an institution owns shows up. Section 13(f) securities cover most US-listed stocks, ETFs, certain convertible debt and listed options – and nothing else, per crypto.news. Bitcoin held directly, in self-custody, is invisible to the form – a point made independently by crypto.news and Yellow.com. Yellow.com’s example: an institution holding 10,000 BTC in self-custody has no 13F reporting obligation for that position at all. Crypto only appears when it sits inside a regulated wrapper: a spot Bitcoin ETF, a spot Ethereum ETF, or crypto-related equity like Coinbase or Strategy.
The same filing category, four different stories
CryptoSlate’s comparison of five Q1 and Q2 2026 filings, published August 15, 2026, shows why treating any single quarter-over-quarter change as “buying” or “selling” fails. Mubadala held 14,721,917 shares of BlackRock’s iShares Bitcoin Trust (IBIT) on both March 31, 2026 and June 30, 2026 – no change in share count. Its reported value nonetheless fell from $565.6 million to $490.1 million. The Abu Dhabi Investment Council held 8,218,712 shares across the same two dates, with its reported value falling from $315.8 million to $273.6 million. Both positions lost about 13.35% of their reported value with zero shares sold – a pure effect of price, not conviction, according to CryptoSlate.
JPMorgan’s filing moved differently: its ordinary spot-ETF shares rose from 8,462,883 in the amended Q1 filing to 10,623,591 in Q2 2026, a 25.53% increase, per CryptoSlate. UBS’s ordinary spot-ETF holdings rose 13.20% to 414,191 shares over the same window, while its disclosed IBIT call-underlying equivalents jumped from 80,000 to 1,950,000 shares and its put-underlying equivalents fell about 52.75% to 143,300 – a shift in the reported options mix, not proof of 1.95 million shares actually purchased, CryptoSlate cautions, because Form 13F expresses options in underlying-share terms rather than contract counts. Morgan Stanley’s external spot-ETF units fell 3.99%, from 19,411,356 to 18,636,055 shares, between Q1 and Q2 2026, while a new line appeared in its Q2 filing: 2,570,627 shares of a Morgan Stanley Bitcoin Trust, not reported the prior quarter. CryptoSlate is explicit that the filings do not establish whether the same client capital moved from the external funds into the new branded wrapper.
Zoom out to the fund complex and a fifth pattern appears: CryptoSlate calculates, using Farside Investors’ daily data, that the broader Bitcoin ETF market recorded about $4.89 billion in net outflows during Q2 2026, with roughly $2.06 billion of that concentrated in the final five trading sessions of June. That aggregate flow data sits at the wrapper level and cannot be tied back to any one of the five filers above, CryptoSlate notes.
An earlier example: the Q1 2025 pullback
CoinShares’ own research, published June 5, 2025, documented a similar-looking but separately sourced pattern a year earlier. Aggregate 13F Bitcoin ETF holdings fell 23% quarter-over-quarter, to $21.2 billion at the end of Q1 2025 from $27.4 billion at the end of Q4 2024. CoinShares attributes part of that to price: Bitcoin itself fell 11% over the same quarter, while the broader US Bitcoin ETF market (institutional and retail combined) declined a smaller 12%, ending Q1 2025 with $92.3 billion in total assets. Institutional investors’ share of that total AUM slipped to 22.9% from 26.3% the prior quarter, per CoinShares. Within the 13F pool, advisors held 50% of assets and hedge funds’ share fell to 32% from 41% the previous quarter – a rotation CoinShares reads as tactical hedge-fund profit-taking rather than a broad institutional retreat.
What the filing structurally cannot show
Yellow.com’s account, published March 15, 2026, stresses that a 13F omits short positions and written options entirely – a limit CryptoSlate’s filing comparison, cited above, also notes. Yellow.com illustrates the consequence with Goldman Sachs: its Q4 2024 filing disclosed $527 million in IBIT put options alongside $157 million in calls – a structure that reads more like a hedged trading book than a directional bet. By Q4 2025, Yellow.com reports, some sources put Goldman’s puts and calls at $827 million and $160 million respectively, though other sources described the options as reduced – the two accounts do not agree, which itself illustrates how thin the public picture around options rows can be.
The form also cannot separate a bank’s own capital from client capital it manages with discretion, nor can it distinguish a deliberate position from an authorized participant’s temporary inventory built up while processing ETF creation and redemption orders – shares that may be gone again within days of the snapshot date, per Yellow.com. And because filings are due up to 45 days after quarter-end, the number is already stale when it becomes public. Yellow.com’s example: Goldman Sachs reported roughly $1.06 billion in IBIT at December 31, 2025, when Bitcoin traded near $88,400. By the time the filing was public in mid-February 2026, Bitcoin had fallen to roughly $68,700, and the identical share count was worth about $944 million – a 45% drop with no shares sold.
The common misreading
The recurring error is reading a quarter-over-quarter change in a 13F’s reported dollar value as proof an institution bought or sold Bitcoin exposure. As the Mubadala and Abu Dhabi Investment Council rows show, the value can fall while the share count stays fixed. As JPMorgan’s and UBS’s options rows show, a rising or falling number can describe one leg of a hedge rather than a net bet. And as Morgan Stanley’s filing shows, a fall in one product line and a rise in a related but differently branded product can look connected without the filing proving they are.
What this page does not tell you
It cannot tell you whether Morgan Stanley’s external-ETF decline and its newly reported branded trust involve the same underlying client accounts, a link CryptoSlate says its own filing comparison cannot establish. It cannot tell you whether any bank’s reported “sole discretion” position is proprietary capital, client capital, or authorized-participant inventory awaiting ETF creation or redemption – a distinction Yellow.com describes as structurally invisible on the form itself. And it cannot tell you what any filer does between quarter-end snapshots: a position opened and fully closed inside one quarter leaves no trace on a 13F at all.
Sources
Every fact above is attributed to one of these reports. Where they disagree, the article says so.
TheCoinrise reference desk (headline only)
TheCoinrise reference desk
TheCoinrise reference desk
TheCoinrise reference desk (headline only)
TheCoinrise reference desk
TheCoinrise reference desk
TheCoinrise reference desk
The post What a 13F filing actually proves about institutional Bitcoin ETF holdings appeared first on TheCoinrise.com.