The model has been run. The output is 68 pages. BCG published it on May 18 from New York, and the question it ends with, on page 64, is the same question that gets asked when an analyst has finished the numbers and is standing in a corridor deciding who to call first: does anyone in this building know what we are actually holding?
The answer, across the European banking system, is mixed. Some institutions know. Thirty-seven of them confirmed it by joining Qivalis, a MiCA-compliant, DNB-supervised, 1:1 euro-backed stablecoin consortium now representing 37 institutions across 15 European countries, with a H2 2026 launch target. BCG had named Qivalis on page 22 of the same report as one of two consortia where the interoperability condition for digital asset scale would be resolved. BCG listed the condition as outstanding. Qivalis listed it as done. The condition was outstanding on May 18 and had 37 confirmed signatures before the week ended.
The remaining institutions are monitoring the space.
BCG Split the Asset Class Into Three. Most Banks Have One Committee.
The taxonomy is where the mistakes start, and the mistakes follow a consistent pattern. Three distinct asset classes with different economics, different threat timelines, and different required responses are being processed through a single governance structure that meets quarterly and produces a paper.
Running crypto, digital money, and tokenized real-world assets through one digital asset strategy committee is the financial equivalent of using one insurance policy to cover a bicycle, a speedboat, and a nuclear reactor, on the grounds that all three are technically property.
Crypto is approximately $3 trillion in market cap at year-end 2025, with a ~$90 billion annual revenue pool. The slice accessible to regulated institutions, covering trading, derivatives, custody, prime services, and staking, is ~$55 billion globally. Revenue yield: 2–4% of assets annually, structurally higher than most traditional asset classes. The banks monitoring the space are watching that revenue depart, one quarterly meeting at a time.
Digital money is led by stablecoins at ~$300 billion, representing 0.5% of global M2 against a $57 trillion total. BCG puts the structural ceiling at ~15% of M2, or approximately $9 trillion, grounded in economics: fiat-backed stablecoins do not yield, do not create credit, and compete for transactional balances rather than savings. The distance between 0.5% and 15% is where transaction banking fees currently live, and where they will stop living if the distance closes.
BCG breaks the $300 billion by use case: 65% crypto trading collateral, 25% store of value predominantly USD in emerging markets, and 10% real-economy payments, the smallest slice and the fastest growing.
The BIS was founded in 1930 as the coordinating institution for the world's central banks. Its April 2026 framework assessed all four digital money instruments against three tests. Stablecoins failed all three. Their future beyond the crypto gateway is, in BIS's words, "unclear under current design." The unified ledger vision places tokenised central bank reserves and commercial bank deposits at the center of the next-generation monetary system. Stablecoins route around both.
The BIS argues stablecoins fail the elasticity test because they operate on a cash-in-advance constraint and cannot expand balance sheets like fractional reserve banks. Stablecoins' lack of elasticity is the deliberate design choice that makes settlement finality possible. The BIS is not wrong. It is applying the criteria of the system it was built to defend.
Under MiCAR, EMT issuers must hold 1:1 reserves in central bank money or HQLA and grant legally enforceable redemption rights at par. King's College London's 2025 analysis described singleness not as an absolute state but as a spectrum maintained by institutional design. Singapore's MAS published its own digital money framework in the same month and operationalised a hierarchy where certain permissionless blockchain assets qualify for Group 1 prudential treatment with adequate risk controls. BCG maps four regulatory archetypes on page 46. The EU sits in the bespoke regime quadrant. Singapore sits in the competitive hub quadrant. Both published in April.
Tokenized real-world assets are ~$30 billion today, grew 300% in 2025, and in BCG's progressive scenario reach $88 trillion by 2035, or 16% of global investable assets. From $1.9 billion in 2023 to $26.5 billion by March 2026. Fourteen times in three years. The first year looked like noise. The second looked like a trend. The third is running through the settlement infrastructure of institutions that do not use the word "pilot" for $7.3 trillion.
Three categories. Three P&Ls. The committee meets in Q3.
Twenty-Five Million European Clients Have Left. The Banks Have Noted This As An Area Of Interest.
When we present client data to traditional bank executives, the response is not disagreement. It is a careful stillness, followed by a request to see the methodology. The mental model these executives carry of their crypto-active clients is detailed and consistent: young, speculative, technically confident, almost certainly a niche, probably best addressed in the Q4 digital strategy review.
BCG cites a BlackRock October 2025 survey across 14 European markets: 22% of European investors hold cryptocurrency. Portugal: 43%. Netherlands: 40%. Switzerland: 34%. Spain: 29%. Banking sectors with full client visibility for decades have, in the majority of cases, responded by establishing a working group.
Chainalysis classifies Romania as an "acceleration zone": crypto adoption still expanding, not stabilising, at 45 to 50 percent year-on-year growth in trading volume. Approximately 25 to 29 percent of Romania's population are active crypto users: 5.65 million people, projected to reach 6.3 million by end of 2026. At the European Conference on Financial Services 2025, the president of Romania's Association of Fund Administrators disclosed the figure: over 600,000 Romanians hold crypto assets. The Bucharest Stock Exchange has 220,000 investors. Three Romanians hold crypto for every one Romanian who holds a stock. The banking sector has classified this as an emerging segment requiring careful monitoring.
Three Romanians hold crypto for every one Romanian who holds a stock.
The 25 million European clients holding crypto outside their bank have not left the bank. They still have mortgages, current accounts, and direct debits. They have simply stopped asking the bank for the menu on the item that matters most to them financially. The bank has interpreted this as evidence that they are not very interested in the menu.
BCG does the arithmetic. For an average G-SIB with 30–40 million retail and wealth clients: ~$40 billion of the off-bank crypto pool is linked to the existing client base. Recapturing 20% translates to ~$80 million in annual incremental revenue today, scaling to $340–$600 million per year as the pool grows. The 20% assumption is not aggressive. It is the minimum defensible figure for a bank that simply offers the service.
You know which category you are in.
A Table on Page 26. The United Kingdom Is Ranked Seventh.
BCG includes Table 2 on page 26: non-US foreign holders of US Treasury bills. The table lists Japan, China, the Cayman Islands, Luxembourg, Ireland, Canada — and then Tether, ranked seventh, ahead of the United Kingdom.
Tether is a private issuer of a dollar-pegged digital bearer instrument, incorporated in the British Virgin Islands, running on public blockchains, with no branches, no physical premises, and a reserves page. It holds more short-term US sovereign debt than the United Kingdom. The United Kingdom has a central bank, a Treasury, a Debt Management Office, an elected parliament, and approximately 67 million people. Both appear on the same table.
BCG separately estimates Tether generated $98 million in profit per employee in 2024. Most banks have more people in their compliance department than Tether has employees. This is filed under competitive dynamics.
$62 trillion in raw stablecoin transfers in 2025. After removing bots, protocol mechanics, and routing: ~$4.2 trillion in economically meaningful activity. Observable bilateral payments for actual goods and services: $350 to $550 billion, B2B the largest segment at ~40%, growing at above 100% CAGR. The gap between $62 trillion and $350 billion is where most current regulation is directed. The $350 billion growing faster than almost anything in financial services is where the banks positioning now are building toward. Both things are true.
Broadridge settled $365 billion per day in tokenized repo on the Canton Network in January 2026. $7.3 trillion for the month. Year-on-year growth: 508%. DTCC, Euroclear, LSEG, and Tradeweb participating. The platform did not exist in recognisable form eighteen months ago. Describing $7.3 trillion in monthly settlement volume as a pilot requires a definition of that word that BCG does not supply.
DTCC announced in December 2025 it is tokenizing DTC-custodied US Treasury securities on Canton for production rollout in H2 2026. DTC is the central securities depository of the United States. The infrastructure has a launch date.
Canton Network, the privacy-enabled permissioned blockchain built in 2023 by Goldman Sachs, BNP Paribas, DTCC, and 47 other institutions, settled that $7.3 trillion. The BIS's Project Agorá, with seven central banks and 43 financial institutions, is designing the same architecture with wholesale central bank reserves added as the settlement layer. The private sector did not wait for the framework.
Qivalis Has 37 Banks. BCG Named The Condition Eight Days Before It Was Met.
On page 22, BCG identifies four conditions required for digital assets to reach scale. Three are either met or close: customer adoption producing real momentum, regulatory clarity functioning as a catalyst, bank-grade technology infrastructure now existing in production. The fourth was interoperability, assessed as "consortia forming but early stages," with two named: Swift and Qivalis.
Qivalis confirmed 25 new members, bringing the total to 37 institutions across 15 European countries behind a MiCA-compliant, De Nederlandsche Bank-supervised, 1:1 euro-backed stablecoin. Launch: H2 2026. ABN AMRO, Erste Group, Intesa Sanpaolo, Rabobank, Groupe BPCE, Bank of Ireland, AIB, and the National Bank of Greece are among the new arrivals. The condition BCG assessed as outstanding on May 18 had 37 confirmations before June.
For anyone tracking the Nordic market: Handelsbanken, Nordea, Jyske Bank, Swedbank, OP Pohjola, and Landsbankinn all confirmed. We have written about Nordic banks moving in formation toward the same destination at different speeds. This month they arrived at the same consortium simultaneously. The formation is noted. The lateness is noted. The arrival is welcome.
DNB, Norway's largest bank, is not on the list. The other Nordic countries are. DNB is building wallet infrastructure designed to connect to digital money from multiple issuers rather than taking a minority stake in one consortium. Ole Morten Sunde, Head of Emerging Business at DNB: "We have assessed that ownership in stablecoin issuance is not right for us at the current time, but we follow Qivalis in particular closely." DNB is developing tokenized deposits in collaboration with Norges Bank. The BIS named tokenized deposits as the preferred form of private money in April.
The ETF structural parallel is the right lens for what comes next. State Street, iShares, and Vanguard today control approximately 75–80% of global ETF assets: $12 trillion, up from $400 billion in 2005. ETFs account for 40–50% of daily US equity trading volume. State Street, iShares, and Vanguard were the first movers. Everyone else competes for the remainder. The window to be State Street in the ETF market opened in 1993 and had, in any commercially meaningful sense, closed by 2000. The window in tokenized fund infrastructure has the same dimensions. Thirty-seven institutions have walked through it. The committee meets in Q3.
What Each Business Line Actually Faces.
Transaction banking is the most exposed revenue pool across every scenario BCG models. Against the base case through 2035: $297 billion in fee compression in the rapid expansion scenario, $215 billion in the fragmented scenario, $146 billion in the most benign institutional evolution case. The mechanism is not complicated: programmable rails remove the intermediary steps between a payment leaving one account and arriving in another. You do not need to be a transaction banker to understand what removing the intermediary steps does to transaction banking revenues. The number is in the right-hand column.
Net interest income faces $203 billion in pressure in the rapid scenario, primarily through deposit leakage as stablecoins absorb transactional balances that underpin cheap funding. BCG is clear that wholesale migration is unlikely near-term. The stickiness weakens gradually. This is less visible on a quarterly earnings call and therefore easier to classify as a structural concern to be revisited annually.
Asset management is the upside story. For a $2 trillion AUM manager: 15–30% revenue uplift, equivalent to $1.2–$2.5 billion annually, as tokenized fund wrappers capture off-book assets, product mix shifts toward alternatives through fractionalization, asset velocity expands the investable base, and new distribution fees emerge from infrastructure ownership. Bottom-up outputs from a firm that does not publish directional gestures dressed as analysis.
Capital markets carries the most technically sensitive upside. BCG models up to 4% RoE uplift for trading businesses. For a G-SIB with $15 billion in trading revenues and $35 billion of tied equity: every 1% RoE is $350 million in additional trading profit. Maximum 4% uplift: $1.4 billion additional profit annually. Entirely dependent on settlement design. Netting-preserving model: 0 to +13% net RoE. Gross atomic settlement: -6% to +10%. The settlement architecture decision is a balance sheet strategy choice with a ten-figure revenue implication that will be made in the next eighteen months, either deliberately by a CTO and CFO who have read this report, or by default.
The Scenario Performs Well Until Page 21.
BCG's four scenarios contain a number that institutional strategy summaries have handled selectively. The bank doing nothing, Scenario 4 "Constraint and Defensive Reset," shows +89% profit growth through 2035 versus Scenario 1's +32%. Revenue growth in the defensive reset: +60%. In absolute terms, the bank watching from the sidelines outperforms the bank building for the transition across the first five years of the modelling period.
BCG includes this number on page 20. They include a graph on page 21.
An investor calls his fund manager in 2006 demanding his money returned. The housing market is fine. He has seen the model. The model shows the cliff. He is not persuaded that fine and correct are the same word. He locks the investors in. BCG has now published something structurally similar, on 68 pages rather than a spreadsheet, and the trajectory on page 21 performs the same function as the model the fund manager ran at his standing desk at 2 am.
"Drift is not neutral. It is unmanaged exposure." — BCG, page 64
Translation: thirty percent of profits, by 2035, against the scenario where the bank acts.
Both things are true. The near-term trajectory is fine. The committee meets in Q3.
JPMorgan Froze Two Stablecoin Startup Accounts In December. The GENIUS Act Had Been Law For Five Months.
Congress passed the GENIUS Act in July 2025. JPMorgan froze accounts for two stablecoin startups in December 2025. Reasons: sanctions exposure, weak KYC controls, fraud signals the bank could not defend to its regulators. The GENIUS Act established what stablecoins are legally permitted to be. It was silent on the question of whether you had hired anyone who knew how to operate one.
The failure pattern is documented. Banking partnerships for stablecoin platforms break along three fault lines: identity verification that is point-in-time rather than continuous; geographic risk that cannot be proactively explained to a compliance reviewer; transaction patterns that diverge from the bank's original risk model without documented justification. Ross Freiman-Mendel, writing from Persona in February 2026: "Banks don't freeze accounts randomly. When they do, they're reacting to specific patterns that create compliance risk they can't defend to regulators." That last line is doing more work than it appears.
BCG's CRO section reaches the same destination from a different direction. In programmable markets, AML becomes flow-based and wallet-centric. Screening direct counterparties leaves the risk that arises several hops away, through mixers, bridges, and layered structures, entirely invisible to a system designed for a different topology. Stablecoins account for ~84% of illicit on-chain transaction volume globally (Chainalysis 2026). The compliance infrastructure built for account-based banking was designed for a different problem. It is being deployed on this one with considerable optimism.
First half of 2025: $3 billion stolen across 119 crypto hacks, most attributable to custody and key management failures rather than protocol exploits. Malicious approval exploits alone: ~$1.5 billion. Not coding bugs. Governance failures in contract approval logic.
BCG's instruction: smart contracts should be governed like high-risk financial models, with pre-deployment testing, continuous monitoring, and explicit intervention authority. This is identical language to what the EU AI Act requires for credit scoring, fraud detection, and AML profiling from 2 August 2026. AMLA directly applicable from 10 July 2027. Both documents are asking for the same architecture from the same institutions. August is twelve weeks away.
Most Institutions Are Building Two Of The Three Layers. The Third Is Where The Regulators Are Going.
A foundation model approach to AML, credit, and fraud has reached production at several major institutions. A single model trained on the bank's own transaction history learns the underlying grammar of how money actually moves through accounts. Every downstream task runs from that shared foundation. Production results: Revolut, 130% improvement in credit scoring, 65% improvement in fraud recall over previous task-specific systems. Stripe's Payments Foundation Model: fraud detection on card-testing attacks from 59% to 97% in a single deployment. Live systems, not benchmarks.
The gains come from the first two layers of a three-layer architecture. Layer one learns the behavioural patterns. Layer two makes specific decisions from that shared understanding. Most institutions currently stop at layer two.
Layer three explains the decisions. It does not make the original call. It generates the reasoning in structured terms that a compliance officer can read, a regulator can audit, and a court can challenge. A CEO in a morning meeting once asked a risk officer to explain what the firm was holding. The risk officer began. The CEO interrupted: explain it as you might to someone who needs the answer in plain language. The answer, it emerged, was that nobody in the room could provide that explanation. The positions had been built across layers one and two. Nobody had built layer three. The EU AI Act regulator arriving in August will ask the same question.
Under the EU AI Act from 2 August 2026, a system without the decision tracing layer cannot be deployed for high-stakes decisions. Under GDPR Article 22, already binding, a credit denial that cannot be specifically explained is a liability. Under AMLA from 10 July 2027, opaque transaction monitoring systems face scrutiny. The institutions that build all three layers will have systems operational in September 2026. The institutions that do not will be rebuilding under time pressure, in production, after model risk validation objects.
The EU AI Act requires a verifiable record of training data, model versions, deployment approvals, and inference decisions. The document-based approach functions on inspection day. It does not hold in enforcement, where the regulator needs to reconstruct exactly which model version was running at a specific moment and what it had been trained on.
A permissioned blockchain ledger, the same infrastructure type underpinning regulated digital asset custody, provides a tamper-evident, append-only record of all of the above simultaneously. The governance layer a bank needs for digital asset operations satisfies the AI Act's evidentiary standard for AI model governance at the same time. One architecture. One decision. One ledger.
Most banks are currently designing these as separate programmes. The plumbing is identical. The bank is paying two separate plumbing teams to install the same pipe.
What Went Under The Radar.
Circle raised $222 million in a presale for Arc on May 11, a week before the BCG report. Lead investor a16z wrote that the infrastructure USDC currently runs on "wasn't built with big institutions in mind." Canton Network has been making the same argument since 2023. The Arc investor list includes BlackRock, Apollo, and Intercontinental Exchange. Circle's Q1 revenue of $694 million missed expectations by $28 million in the same quarter. USDC is 80% of the business. BCG's condition 3 for digital asset scale, bank-grade infrastructure in production, now has two competing answers. Arc is the hedge.
BCG maps tokenization penetration by asset class through 2035. Commodity funds: 40–50%. Money market instruments: 25–40%. Alternatives: 25–35%. Securitized debt: 20–30%. Listed equities: 3–7%. Government bonds: 3–5%. The highest tokenization potential sits in the illiquid, operationally intensive categories where current infrastructure is most visibly broken. The lowest rates belong to the assets most commonly featured in tokenization marketing materials. Tokenization solves problems. The assets at the top of the list have more problems.
Société Générale's digital asset unit, SG-FORGE, integrated its regulated EUR and USD stablecoins with Canton Network in May 2026, targeting institutional workflows for tokenized collateral, repo financing, and bank-grade settlement. The BIS described stablecoins' integrity architecture as structurally deficient in April. A major European bank began connecting them to the institutional settlement network in May.
BCG models the telecommunications transition without nostalgia. BT launched its 21st Century Network programme in 2004. Circuit-switched and packet-switched networks coexisted for over 15 years. The UK's formal circuit switch-off: 2025, nearly 20 years after large-scale IP migration began. Banks running dual rails for a decade is not a failure of ambition. The question is not whether to run two systems simultaneously. It is whether the old system's near-term profit comparison continues to fund the new system, or continues to argue against it.
BCG closes the ten-step section with five questions for annual CEO review:
- Where does irrelevance risk now exceed execution risk?
- Which regulatory or interoperability assumptions could break the strategy?
- Are we infrastructure price takers or price setters?
- Does our liquidity model hold in atomic settlement stress?
- Are we preserving optionality, or accumulating hidden lock-in?
That last question is doing more work than it appears.
The Cointegrity Perspective.
This is the layer we work in. Not the price action. The regulatory architecture, the compliance decisions, and the meeting where senior appetite is clearly present and the compliance team has not yet been given the framework to move with the board.
That meeting is where programmes stall. BCG has now quantified what stalling costs.
The full BCG report is a free download at bcg.com. The five questions on page 64 are worth your morning coffee. If the conversation about what follows is relevant, it starts at cointegrity.io.
Cointegrity is a compliance-first infrastructure and advisory firm operating across digital assets, regulatory technology, and AI governance for financial institutions. This deep-dive article is published in May 2026 in connection with BCG's flagship report "The Future of Digital Assets" (May 18, 2026). BCG report available at bcg.com.
Related internal resources: Bitcoin, Ethereum, Stablecoin, Blockchain.