An institution that has decided to hold a digital asset has made the easy decision. The hard one is a market-structure question rather than an investment one: through whom. A public ledger settles a transfer without asking permission, read for most of a decade as a forecast that intermediaries were being written out. The filings say otherwise. Between an investment committee and a bitcoin held on its behalf sit a distributor, an execution venue, a custodian and a wrapper issuer, and every one of them is paid.
What follows takes the access and execution chain only: who faces the client, who routes the order, who collects the fee. Custody law, segregation and insolvency belong to who actually holds it; sizing and vehicle selection to how institutions allocate.
Three seats on the main stage, and no custodian on it
The Paris 2026 programme put the question on its largest stage. Banks, Exchanges, and Asset Managers: Who Controls Institutional Access When Markets Move Onchain? ran on Day 1, Tuesday 2 June 2026, from 17:20 to 18:00 on the Hecto Main Stage at the Louvre Palace, in the Tokenisation of Finance track. The programme listed the three panellists as Amy Oldenburg, Head of Digital Asset Strategy at Morgan Stanley; Alex Kim, Chief Executive Officer of Upbit Global; and Matthew Sigel, Head of Digital Assets Research and Portfolio Manager of the VanEck Onchain Economy ETF at VanEck. Frank Chaparro, listed as Head of Special Projects and Content at GSR, moderated.
No abstract was published for that session, only a title, a stage, a time and four names, so what follows is not a report of what was said. It is a reading of the seating chart, which is itself an argument: a bank, a trading venue and a wrapper issuer were cast as the three parties who decide institutional access, and no custodian was given a seat. That absence is where our custody piece begins.
Two neighbouring slots on the same stage frame it. Earlier that afternoon Chaparro moderated a fireside, What Institutions Need From Blockchain Infrastructure, from 15:00 to 15:30, with Ronit Ghose of Citi and Evan Cheng of Mysten Labs. At 13:45 Kim had sat on the panel After the Flush: What October 10th Revealed About Crypto Market Fragility. Access and fragility were the same question about the same firms, hours apart.
What a bank is actually selling
The clearest account of what a bank's digital asset launch consists of is a disclosure, and it comes from the wealth channel rather than the institutional desk. Announcing on 16 July 2026 that E*TRADE from Morgan Stanley had completed its rollout of spot trading, the firm said eligible clients could buy, sell and hold Bitcoin, Ethereum and Solana "in a linked zerohash account at a competitive price of 50 bps". The same release then states: "Morgan Stanley Smith Barney LLC ... does not transact in or custody digital assets. All digital asset transactions and custody occur between you and Zero Hash LLC ... through a separate, non-brokerage account in your name at zerohash and outside of Morgan Stanley. Digital assets held through zerohash are not FDIC insured or SIPC protected."
The product is legible. The bank supplies the relationship, the brand and the consolidated view; a third party supplies the account, the asset and the legal counterparty. Chad Turner, Head of Morgan Stanley Wealth Management Platforms, described the rollout as bringing capabilities to clients "in an integrated way". The integration is at the screen.
This is lawful and deliberate. OCC Interpretive Letter 1184, signed by Acting Comptroller Rodney E. Hood on 7 May 2025, confirms that a bank "may buy and sell assets held in custody on a customer's behalf at the direction of the customer" and "may use a sub-custodian to provide custody services". The release also cites Morgan Stanley Wealth Management's Pulse Survey, in which the top response on what matters most when choosing a platform to trade crypto was "an established company they can trust". Trust is the product, not the balance sheet behind the asset.
The bank is at least four different businesses
Treating "banks" as one contender is the first error. Distribution is one model. Custody is a second: Citi launched Custody+ on 18 August 2026, saying it "expects to go live with digital asset custody later this year, starting with the custody of Bitcoin", so that "clients will access traditional and crypto custody capabilities within the same framework", against a Services business the bank says invests over US$2 billion annually in its platform strategy.
Principal trading is a third. In July 2025 Standard Chartered put deliverable spot Bitcoin and Ether onto its existing FX platforms through its UK branch, claiming to be "the first global systemically important bank to offer deliverable spot Cryptoasset trading for institutional clients", with clients free to "settle to their choice of custodian". That clause matters: the bank unbundles execution from safekeeping by default, the opposite of the venue model below. Crowdfund Insider reported in September 2026 that the service had reached institutions in the United Arab Emirates through the bank's DIFC unit, under Dubai Financial Services Authority supervision.
Issuing the bank's own onchain money is the fourth. J.P. Morgan's Kinexys milestones record that its USD-denominated deposit token, JPM Coin, ticker JPMD, "became available for J.P. Morgan's institutional clients' use on Base", described there as the Ethereum layer 2 built within Coinbase. Here nothing stands between client and asset, because the bank's liability is the asset.
One further Citi product spoils the tidy story that bundled functions are an exchange problem. In June 2026 it issued digital depositary receipts over private company shares on infrastructure operated by SIX, with Kaleido the first company whose equity was tokenised, and said this was the first time a global financial services company had both issued and acted as custodian for tokenised depositary receipts representing private companies. That is the combination-of-functions question usually aimed at trading venues. It attracts less commentary when a G-SIB does it.
The venue that is also the vault
The exchange case has official analysis behind it. In November 2023 the Financial Stability Board defined multifunction crypto-asset intermediaries as firms offering combinations of services "typically conducted by separate legal entities in traditional finance". Its comparison annex is blunt: in traditional finance "exchanges and broker-dealers are usually separated", whereas many such firms provide "a combination of trading venue, brokerage, settlement, and custody services without separation of legal entities", with the listed risks including misappropriation of clients' funds and firms "trading against or ahead of their customers". Combining functions that are "typically restricted or separated for traditional finance", the FSB concluded, "appears prima facie inconsistent with the principle of 'same activity, same risk, same regulation'".
None of it is hypothetical. The sharpest evidence is filed by a large bitcoin fund about itself. The iShares Bitcoin Trust 10-K for FY2025 names Coinbase Custody Trust Company as Bitcoin Custodian, Anchorage Digital Bank as Additional Bitcoin Custodian, and Coinbase, Inc. as Prime Execution Agent, described in the filing as "an affiliate of the Bitcoin Custodian". It discloses the consequences without euphemism. Orders to buy or sell bitcoin "may be routed to the Prime Execution Agent's platform where such orders may be executed against other Coinbase customers or with Coinbase acting as principal". The agent "does not engage in front-running, but is aware of orders or imminent orders and may execute a trade for its own inventory". When acting as principal, the filing concludes, "the Coinbase Entities may have an incentive to favor their own interests and the interests of their affiliates over the Trust's interests".
Two further disclosures bear on execution quality. The venues an order can reach include Bitstamp, LMAX, Kraken, the Prime Execution Agent's own platform and "four additional non-bank market makers" the filing leaves unnamed, and the agent "shall have no liability, obligation, or responsibility whatsoever for the selection or performance of any Connected Trading Venue". The Sponsor also "has no control over, and for security reasons the Prime Execution Agent does not disclose to the Sponsor, the percentage of bitcoin that the Prime Execution Agent holds". The wrapper cannot see how its own custodian holds the coins.
The trading business on the other side of that arrangement is shrinking. Coinbase's Q2 2026 Form 10-Q reports transaction revenue of $599.2 million for the quarter, down from $764.3 million a year earlier, out of $1,154.3 million in total net revenue. Institutional transaction revenue was the exception, rising from $60.8 million to $100.1 million. It is still the smaller part: $100.1 million of $599.2 million. Coinbase's own results release put its crypto trading volume market share at 10.3% for the quarter, which it called a new all-time high. The venue holding the keys for that bitcoin fund makes most of its transaction revenue somewhere other than its institutional business.
What the wrapper charges, and which way it is moving
The asset manager's contribution is the least technical and the most durable: a legal wrapper, a listing, and a line item an existing mandate can already hold. The iShares filing prices it exactly. The Sponsor's Fee accrues daily at an annualised 0.25% of net asset value, having been waived down to 0.12% on the first $5.0 billion of assets until 10 January 2025. Net asset value went from $51,519,566,547 at the end of 2024 to $67,401,155,244 at the end of 2025.
Then the direction reverses. BlackRock's own product page shows net assets of $61,602,737,818 as of 8 September 2026, below the year-end figure. The wrapper business is not a straight line. Note too that 0.25% a year and 50 basis points on a trade are not comparable quantities. The cost of access is quoted in incommensurable units, and nobody is obliged to convert them for you. Basket mechanics sit in our piece on ETF market plumbing.
Disintermediation is a claim about which function, not whether
The strongest sceptical case is that ledger finality removed one intermediary, the settlement agent, and left the expensive one untouched: credit. A white paper dated February 2026 from Ripple Prime, the Hidden Road business, argues this and must be read with its interest declared: the firm sells the unbundled prime brokerage model it advocates. Within that caveat the structural description is useful. "Most digital asset exchanges are vertically integrated, combining execution, custody, and credit within a single platform", it says, with some venues also rehypothecating client collateral "at no cost", against a foreign exchange market where execution, credit and settlement sit in separate firms.
Its quantified estimate is an interested party's number, not neutral research: default swap rates "of approximately 11%" applied by offshore exchanges and many bilateral liquidity providers, "roughly 7% above the risk-free rate", which the paper converts to about "$192 dollars per million per day". The behaviour behind it is harder to dispute. Clients "are unwilling to leave assets on exchanges or with bilateral LPs longer than necessary", so they prefund, trade and withdraw, and "'instant' today is only possible because exchanges use free client collateral". Intermediaries neither vanish nor are parasitic. The charge has migrated from an explicit fee into an execution spread and a funding cost, where it is harder to see and harder to negotiate.
The answer changes at the border
There is no global answer to whether one firm may run the venue, hold the asset and trade against the client, because the bodies that set the terms have not converged.
- IOSCO. Summarising IOSCO's May 2023 consultation report, the FSB records that recommendation 2 asks providers to address conflicts arising from vertical integration, "including the possible need for measures such as legal disaggregation and separate registration", and that recommendation 3 states a provider "should accurately disclose each role and capacity in which it is acting at all times".
- ESMA. Its final report of 31 May 2024 on the MiCA conflicts standards takes the opposite emphasis, clarifying that the draft technical standards do "not per se prohibit vertical integration" and that providers "may be vertically integrated and offer a variety of crypto-asset services and other activities", provided conflicts are identified, prevented, managed and disclosed. Its Q&A 2578 nonetheless says a custodian should "avoid commingling clients' crypto-assets with crypto-assets held on behalf of entities of the same group", citing Article 75(7) and the risk that an affiliate's information advantage lets it withdraw ahead of other clients.
- The SEC. Chairman Paul S. Atkins has declared himself "a strong proponent of 'super-apps' in finance that allow for the custody and trading of a variety of asset classes within a single regulatory license". That speech of 12 November 2025 contains no discussion of conflicts arising from combining those functions; his remarks of 29 January 2026 at the joint SEC and CFTC harmonisation event announced no rules on integrated venue-and-custody models; and the Regulation Crypto Assets proposal of 18 August 2026 addresses offerings and disclosure, not intermediary conduct.
The map beneath is more fragmented still. The FSB's peer review of 16 October 2025 records that "Canada, Hong Kong and the EU prohibit proprietary trading by CASPs operating trading venues", while Armenia, The Bahamas, Bermuda, Chile, Indonesia, the Philippines, Thailand and Türkiye "allow proprietary trading for CASPs but subject them to specific requirements". In Switzerland, "licensing requirements for each activity must be met when combining functions", while "some jurisdictions, including Australia, Canada, and Korea, have yet to develop detailed regulatory approaches for CASPs offering multiple services". The same brand may therefore be free to trade against you in one entity and barred from it in another. The question is never what the firm does, but which entity signed your agreement, and who licensed it.
And it changes by asset
The second reason there is no single answer is that "the asset" is four different legal objects. For a bearer asset such as bitcoin, the contest is genuinely between the venue-custodian and the wrapper, as the iShares filing shows. For a stablecoin, the intermediary is the issuer, because redemption at par is a claim on a company rather than a property of the ledger, as argued in the institutional view of stablecoins. For a tokenised bank deposit such as JPMD, the intermediary is the bank whose liability the token is. For tokenised private shares, as in the Citi receipts, there is no public venue at all and access is the issuer's to grant.
Questions that separate a counterparty from a brand
- Which legal entity is my counterparty, in which jurisdiction, and is it the one holding the licence being cited?
- Who holds the keys, and is that entity an affiliate of the firm executing my orders?
- Expressed in one unit, what is the all-in annual cost of access: fee, spread and funding drag together?
What the panel could not settle
Forty minutes with a bank, an exchange and an asset manager was never going to resolve a question on which a standard setter, an EU authority and a US regulator have taken visibly different positions over three years. What it did was frame the problem correctly: access, not conviction, is the binding institutional constraint, and whoever controls access captures the margin.
That programme is archived, and published at sessions and agenda. The argument is not. It moves to the Louvre Abu Dhabi in the Saadiyat Cultural District on 3–4 December 2026 for the inaugural Gulf edition, a capped room with no paid speaking slots, and returns to the Musée des Arts Décoratifs at the Louvre Palace in Paris on 2–3 June 2027. Admission to either is by application and review.