An allocator looking at digital assets in 2026 is rarely asking whether the asset class exists. What is left is harder to answer from public material: what the mandate permits before anything else, how much, through which vehicle, measured against what, and what the position does when everyone tries to leave at once.
The Paris 2026 programme gave that sequence 19 of its 77 scheduled items, on a track called Investing in Digital Assets, the second largest after tokenisation, which carried 33. Seventeen of the 19 had named speakers and carry session pages of their own; the two days are listed in full on the agenda. What follows treats that programme as evidence of which arguments institutions thought worth a stage on 2–3 June 2026, set beside the filings an allocator has to read.
One point governs everything below. Nothing in this research is a published target allocation. No endowment, sovereign fund, pension or bank cited here states a policy number for digital assets in a document of its own. Every percentage that follows is either derived by us from a regulatory filing or self-reported to a survey. That absence is itself the finding, worth holding on to whenever a headline reports what institutions "allocate".
The mandate gate closes before the sizing question opens
For a bank, the decision was largely made in Basel. The Committee's cryptoasset standard, SCO60, took effect on 1 January 2026, following amendments the Basel Committee published on 17 July 2024 with that implementation deadline. It sets a ceiling: "A bank's total exposure to Group 2 cryptoassets should not generally be higher than 1% of the bank's Tier 1 capital and must not exceed 2% of the bank's Tier 1 capital" (SCO60.117).
The clause that matters for anyone hoping a listed wrapper solves the problem comes just before it. Banks must apply that limit to aggregate exposures "including both direct holdings (cash and derivatives) and indirect holdings (eg those via investment funds, ETF/ETN, or any legal arrangements designed to provide exposures to cryptoassets)" (SCO60.116). A 1,250% risk weight applies to Group 2b exposures, and SCO60.83 extends it to funds of those assets, naming cryptoasset ETFs, and to other entities whose value is primarily derived from them. An exchange traded product and a treasury company's equity land in the same bucket. The wrapper changes the operational problem, not the capital charge.
Elsewhere the gate has been opening, though less far than the coverage suggests. The SEC approved generic listing standards for Commodity-Based Trust Shares on 17 September 2025, letting exchanges list qualifying products without first submitting a proposed rule change under Section 19(b). The FCA opened UK retail access to crypto exchange traded notes from 8 October 2025, with no Financial Services Compensation Scheme cover and its ban on retail cryptoasset derivatives left in place. The US Department of Labor's proposed rule of 31 March 2026 states there is "no per se rule respecting the inclusion of actively managed investment vehicles that are investing in digital assets", and sets out a safe harbour resting on six factors: performance, fees, liquidity, valuation, benchmarking and complexity. It remains a proposal.
That is the room the panel Banks, Exchanges, and Asset Managers: Who Controls Institutional Access When Markets Move Onchain? was convened into, on the Hecto Main Stage at 17:20 on 2 June 2026. The programme records Amy Oldenburg of Morgan Stanley, Alex Kim of Upbit Global and Matthew Sigel of VanEck on the panel, moderated by Frank Chaparro of GSR. Access, not conviction, was the framing, and the surveys agree: the EY-Parthenon and Coinbase survey published in March 2026 found regulatory uncertainty a primary concern for 66% of respondents, with 78% naming market structure as the area most in need of regulatory clarity.
Sizing, and where the single-digit figures come from
Almost every "institutions allocate 1 to 5%" line in circulation traces to one of two survey families, both of which need reading with the methodology open. The EY-Parthenon and Coinbase Institutional Investor Digital Assets Survey, published March 2026 with fieldwork from 13–30 January 2026, covered 351 institutional investors, 60% of them in the United States. It reports that "the proportion of firms allocating more than 5% of AUM to digital assets is expected to rise from 18% to 29% by the end of 2026", on a base of 317. Its 34 not-currently-invested respondents planned under 1% (38%), 1–5% (47%) or 6–10% (9%).
Two caveats travel with those numbers and are usually stripped off. The survey is co-branded by an exchange that sells the product being measured, and it screens by design: respondents are currently invested, previously invested, or planning to invest in the next twelve months. Nobody who decided never to invest is in the sample, so the quoted finding that 73% plan to increase allocations in 2026 is 73% of a group pre-selected for interest.
The second family is Invesco's Global Sovereign Asset Management Study, where the more useful evidence is an absence. The 2026 edition, published June 2026, drew on 144 senior investment professionals at 90 sovereign wealth funds and 54 central banks managing about US$29 trillion, interviewed face to face by NMG between January and March 2026. Across its 46 pages we found no occurrence of "digital asset", "crypto", "bitcoin" or "stablecoin". "Gold" appears 28 times, "AI" 93. The 2025 edition, same methodology, carried a whole theme titled "Digital assets: continued exploration amid structural potential". Read that precisely: the 2026 study does not mention the subject, which is not the same as saying sovereigns have stopped. And the 2025 figures rest on small bases, 64 sovereign funds behind the participation chart and 29 behind the one reporting cryptocurrencies 76% and stablecoins 48%.
A tension worth stating rather than smoothing over. Kathleen Wrynn, billed by the programme as Invesco's Head of Digital Assets, appeared on the Hecto Main Stage on 3 June 2026 in a fireside with Azeem Khan of Miden, Crypto Winter, Institutional Summer: Accelerating Digital Asset Market Maturity. Her firm's sovereign study had been in the field for the three months before she took that stage, and was published that same June with the topic absent. Both are true. A specialist digital assets desk and a sovereign investor's strategic asset allocation are not the same room.
The 2025 study did say where the money sits when it arrives: digital assets "are therefore generally not positioned as substitutes for traditional safe havens like gold, nor are they yet treated as standard diversifiers like equities or bonds", and even small experimental exposures usually require "explicit board-level sign-off". EY-Parthenon's respondents agreed on the filing cabinet: 40% book digital assets under alternatives, 32% as their own category, 19% as innovative or emerging technology, and 9% elsewhere, including currencies and real assets. No option on that list describes a core holding.
The denominator changes the answer more than the position does
Harvard Management Company is the cleanest worked example. Its Form 13F for the quarter ended 30 June 2026 shows iShares Bitcoin Trust at $101,355,133 on 3,044,612 shares. As a share of that table, $4,263,102,872 across 19 lines and dominated by one $2.21bn SpaceX position, it is 2.38%. Against Harvard's endowment of $56.9bn, the FY2025 figure as of 30 June 2025 published by its own research office, it is 0.178%. One position, two defensible denominators, more than thirteen times apart, a year out of step because no FY2026 endowment value exists yet.
Form 13F causes most of the distortion. It captures US-listed equities and ETFs, not private funds, direct holdings, foreign securities, derivatives or cash, which is most of an endowment. Every percentage derived from one is a share of the table, not the portfolio.
Mubadala Investment Company shows the same arithmetic at sovereign scale. Its IBIT position at 30 June 2026 was $490,092,617 on 14,721,917 shares: 1.41% of its $34.77bn 13F table, and 0.127% of the US$385bn it says it manages, per its results announcement of 9 April 2026 for FY2025. An elevenfold spread on one position, with a six-month date mismatch. Our post on what changed in Abu Dhabi covers the ADGM framework behind it.
The vehicle question, and what each one costs
The spot ETP is now the default. Summing net assets reported by all twelve US spot bitcoin products in their own Form 10-Q filings for the period ended 30 June 2026 gives $71,417,346,401. That is our summation of primary filings, not a published index, and what it shows is concentration: IBIT alone reported $43,386,012,125, or 60.75% of the total, and the three largest funds 86.6%. The eight spot ethereum products held $8,050,053,739 on the same date, 8.87 times smaller. Two months on, BlackRock's own page reported IBIT net assets of $62,524,453,295 as of 4 September 2026.
Fees on an identical underlying span more than tenfold. From the filings: the Morgan Stanley Bitcoin Trust prospectus sets a unitary delegated sponsor fee of 0.14%, Grayscale Bitcoin Mini Trust charges 0.15%, Bitwise 0.20%, iShares 0.25% and Grayscale Bitcoin Trust 1.50%. The two Grayscale products carry identical 10-K wording apart from the number. In FY2025 GBTC shareholders paid a sponsor's fee of $280.6m on net assets of $14.50bn at 31 December 2025, while IBIT shareholders paid $174.6m on a fund reporting $67.40bn at the same date, four and a half times larger.
Filing counts suggest allocators noticed. Searching SEC EDGAR full text for 13F-HR filings naming "iShares Bitcoin Trust", we counted 798 in August 2025 against 868 in August 2026, up 8.8%, while the whole pool of 13F-HR filings containing the word "Bitcoin" was flat at 1,216 and 1,213. Filings naming "Grayscale Bitcoin" fell from 493 to 412; "Fidelity Wise Origin Bitcoin" held at 364 and 363. "iShares Ethereum Trust" rose from 12 to 50, fourfold off a tiny base, leaving roughly 17 bitcoin filings for every ethereum one. These count filings containing a phrase, not institutions, so the level is unreliable even though the year-on-year direction is not. What they suggest is that the number of filers is not growing, and that the money is consolidating into the largest vehicle, which is not the cheapest one available.
The treasury-company proxy is a different instrument
Paris put the treasury question on three times: a workshop on the X Ventures Masterclass Stage at midday on 2 June, On Trail: The Bitcoin Treasury Company Model; a Taostats Stage fireside that same afternoon, The Role of Digital Asset Treasuries and the Investment Case Ahead, with Nico Pasquariello of Cantor and Hyunsu Jung of Hyperion DeFi, moderated by Hadley Stern; and a Roundtable Lounge session on the morning of 3 June, Digital Asset Treasuries Roundtable.
The filings describe something that is not a wrapper. Strategy Inc reported 845,050 bitcoin as of 30 August 2026 at an aggregate purchase price of $63.73bn, an average of $75,412; in the week to 30 August it bought 4,603 BTC at an average of $80,318, above its own running cost. It funds purchases by issuing equity, and not all the proceeds reach the asset: of $602.8m net raised that week, $369.7m bought bitcoin, $50.7m funded dividends on its STRC preferred, $151.8m repurchased STRC and $30.0m went to cash. Roughly 61% reached the underlying.
The capital stack matters more than that ratio. Its Q2 2026 10-Q reports approximately 846,000 bitcoin carried at $49.67bn, stockholders' equity of $30.89bn, and Series A perpetual preferred with a $15.46bn liquidation preference ranking ahead of the common. An ETP holder owns bitcoin pro rata; a treasury-company shareholder owns the residual behind that preference. The vehicle also sells: about 1,395 bitcoin in the first half of 2026, and 1,690 BTC in the week to 9 August at an average of $64,262, below average cost, with the proceeds funding repurchases of the STRC preferred. On 31 August 2026 it said management would recommend holding the STRC dividend rate at 12.00% until that stock "has demonstrated sustained, healthy trading near $100 per share".
Allocator behaviour has diverged accordingly. Counting both names to control for the 2025 rename, "MicroStrategy" plus "Strategy Inc" appeared in 788 13F-HR filings in August 2025 and 646 in August 2026, down about 18%, while IBIT filings rose 8.8%. Summing two phrases double-counts any filing naming both, so read the direction rather than the level.
Benchmarked against what
The Department of Labor's six factors put benchmarking in writing, and it is the question the surveys never answer, because a satellite sleeve has no obvious index. The filings hint at the comparison allocators actually make. At 30 June 2026 Harvard held more gold than bitcoin: iShares Gold Trust at $149.5m plus SPDR Gold Trust at $21.7m, against $101.4m of IBIT. Nine months earlier it ran the other way, gold at $235.1m against bitcoin at $442.9m. Invesco's 2025 study insisted digital assets were not a substitute for gold. One endowment's filings show the two lines moving as though somebody treated them as comparable.
The other benchmark is the fund's own return. Mubadala reported assets under management up 17% to US$385bn for FY2025, with a five-year annualised return of 10.7% and ten-year of 10.3%. That is the bar an incremental sleeve must clear net of governance cost, and it explains how a 0.127% position sits in a portfolio for seven quarters without becoming a strategic allocation.
Liquidity under stress, where the decision is actually made
The most instructive record is an exit. The State of Wisconsin Investment Board filed a Q4 2024 13F on 14 February 2025 listing iShares Bitcoin Trust at $321,501,621 on 6,060,351 shares, 0.82% of a $39.2bn table. The next filing contained no occurrence of the word "Bitcoin", and neither did its Q2 2026 filing, a $49.55bn table of 2,529 entries. Exactly four SWIB filings have ever contained the word, the first of them filed on 14 May 2024 for the quarter ended 31 March 2024.
What makes it interesting is what it kept. That Q2 2026 filing still lists Strategy Inc, Coinbase Global, Riot Platforms, Galaxy Digital and Circle Internet Group. The board did not exit the theme. It exited the direct vehicle, and it did so first.
Two other trajectories complete the picture. Harvard first appeared with IBIT at 30 June 2025 on 1,906,000 shares, peaked at $442.9m on 6,813,612 shares that September, then cut the count by 55% across the two quarters to 31 March 2026 and left it unchanged at 30 June; a brief iShares Ethereum Trust position, $86.8m at 31 December 2025, was gone within a quarter. Mubadala ran the other way, and it started earlier: 8,235,533 shares at 31 December 2024, 8,726,972 through the first three quarters of 2025, then 12,702,323 at that year end and 14,721,917 at 31 March 2026, a count it left unchanged at 30 June as the mark fell from $565.6m to $490.1m. One allocator cut and stopped. One added and stopped. Neither has published a reason.
That is the conversation the panel After the Flush: What October 10th Revealed About Crypto Market Fragility was built for, on the Hecto Main Stage on 2 June 2026, with Victor Jung of Hamilton Lane, Alex Kim of Upbit Global, Stephan Lutz of BitMEX and Avtar Sehra of STBL, moderated by Artem Sinyakin of OAK Research. The surveys point the same way: 49% of EY-Parthenon's respondents said recent volatility had strengthened their firm's emphasis on risk management, liquidity and position sizing.
What the record does not settle
Three honest limits. Aggregate figures for the spot ETP market vary by billions between vendors, and no dated primary total for September 2026 could be verified, so the 30 June filings are the number we stand behind. Fidelity Digital Assets still serves its institutional digital assets study under a publication date of 15 June 2020, on fieldwork run from November 2019 to early March 2020, which is worth checking before it is quoted as current. And nothing above is advice. For a bank whose Tier 1 capital cannot absorb a 1,250% risk weight, or an allocator needing same-day liquidity in a market that has repeatedly failed to supply it, the defensible size may be zero, and the filings show a large public pension reaching that conclusion.
These positions were argued in one building. The programme carried Franklin Templeton, VanEck, Morgan Stanley, State Street, Coinbase Asset Management, Dragonfly, Arrington Capital, Haun Ventures and Spartan among others. The Great Crossover: Asset Management for the Onchain Investor put Franklin Templeton chief executive Jenny Johnson opposite Adam Back of Blockstream on the Hecto Main Stage on 2 June 2026, moderated by Eleanor Terrett; Web3 Venture in 2026: What Still Gets Funded put Arrington Capital and Spartan on the Taostats Stage the next morning, moderated by Marc Baumann. The full roster is in our post on institutions on stage; why the programme counts as evidence rather than advertising is in how the programme is built.
That roster is archived, bound to Paris on 2–3 June 2026. The next edition is the inaugural Gulf one, at the Louvre Abu Dhabi in the Saadiyat Cultural District on 3–4 December 2026, capped at 2,000, admission by application and review. Details are on the Abu Dhabi edition page, and allocators can request access there.