Digital Assets in 2026: The Infrastructure Transition Banks Cannot Delay

Digital assets are no longer an innovation bet. They are a settlement, custody, and balance-sheet architecture question. The strategic task for banks is infrastructure design under structural uncertainty.

---

Executive Summary / Key Takeaways

• Digital assets are an infrastructure transition, not an innovation theme. Crypto is a ~$3 trillion asset class, stablecoins have reached ~$300 billion, and tokenised real-world assets grew roughly 300% in 2025. The BCG flagship report published in May 2026 frames this as a strategic, governance, and balance-sheet question for every bank board — not a fintech curiosity (BCG, The Future of Digital Assets, May 2026).
• The biggest structural risk is not a single product. It is losing the client interface, balance-sheet relevance, and control of critical settlement infrastructure to non-bank platforms. In BCG's rapid digital expansion scenario, banks could face ~10% smaller balance sheets, ~14% lower revenues, and ~30% lower profits by 2035 versus a no-digital-asset baseline.
• The biggest structural opportunity is equally clear. Trading businesses could see up to ~4% RoE uplift, asset managers could unlock 15%–30% revenue growth, and retail and corporate banking could each capture hundreds of millions in annual incremental revenue — if banks design wallets, custody, tokenised funds, and programmable treasury services into their operating models now.
• The winning model is orchestration. The institution that can route across stablecoins, tokenised deposits, CBDCs, and legacy rails — with bank-grade custody, embedded compliance, and multi-chain architecture — will shape the next settlement stack. The institution that waits will connect to someone else's (Sebastien Rousseau, Stablecoins vs Tokenised Deposits).

---

Why 2026 Is the Year Digital Assets Became an Architecture Question #

The digital asset conversation has shifted. It is no longer about whether blockchain technology has a role in financial services. The question is where in the value chain — settlement, custody, issuance, collateral management, payments, treasury, distribution — a bank needs to own infrastructure, where it can partner, and where delay creates irreversible dependency.

Three data points explain why the conversation has changed. Crypto remains a roughly $3 trillion market capitalisation asset class with an annual revenue pool estimated at ~$90 billion — margin-rich and commercially meaningful for any institution with an institutional client base. Stablecoins have reached ~$300 billion in outstanding value, roughly 0.5% of global M2 money supply, with strong growth once major jurisdictions reach full regulatory clarity. And tokenised real-world assets, while still small at around $30 billion on public chains at year-end 2025, grew at approximately 300% year-on-year and are the category with the deepest structural relevance for banking over the next decade. By mid-May 2026, on-chain RWA value had reached $31.4 billion according to RWA.xyz data, driven almost entirely by institutional demand for on-chain yield, tokenised US Treasury products, and money-market funds.

BCG's May 2026 flagship report, The Future of Digital Assets, frames this as a strategic infrastructure transition comparable to the telecommunications shift from circuit-switched voice to packet-switched IP networks — a dual-rail migration that unfolded over two decades. The report argues that banks will operate legacy and tokenised infrastructure in parallel for years, and that the strategic objective is not to predict the winning rail but to remain systemically relevant regardless of which rail scales.

This article examines the BCG framework through the lens of practical banking technology architecture — the layer where strategy either becomes operational or remains a slide deck. It draws on previous analysis of stablecoins versus tokenised deposits, programmable liquidity, tokenised deposit services, and the DORA, AI Act, and data sovereignty compliance stack to connect the BCG vision to what banks actually need to build, govern, and evidence.

The Three Asset Classes: Different Economics, Different Architecture #

BCG separates digital assets into three categories that look similar from the outside but carry fundamentally different economics and strategic implications.

Digital Money: Stablecoins, Tokenised Deposits, CBDCs #

Digital money — privately issued stablecoins, bank-issued tokenised deposits, and central bank digital currencies — sits closest to the core of the financial system. The architecture question is which form of digital money serves which settlement use case, and how a bank protects its deposit franchise while capturing the benefits of programmable settlement.

Stablecoins are the dominant form today. Roughly 65% of the ~$300 billion outstanding is tied to crypto trading and DeFi activity, about 25% serves as store-of-value demand (primarily dollar exposure in emerging markets), and only around 10% is linked to real-economy payments — but that segment is growing fast. The BCG analysis estimates that real bilateral stablecoin payments for goods and services reached $350–550 billion in 2025, with B2B the largest segment at roughly 40%.

The deposit-flight discussion is real but bounded. BCG argues that absent a fundamental monetary regime change — stablecoins becoming legal tender, paying interest, and being insured — roughly 15% of global M2 money supply (about $9 trillion at 2025 levels) represents a natural upper limit for stablecoin adoption, constrained by yield dynamics, credit-system feedback, and institutional structure.

For a deeper analysis of the competitive dynamics between stablecoins and tokenised deposits — including counterparty risk, interoperability fragmentation, and the BIS unified-ledger framework — see my earlier article on stablecoins versus tokenised deposits. The BCG report reinforces that analysis: tokenised deposits are most beneficial for large-transaction banks with internal settlement and wholesale payment use cases, while stablecoins win wherever parties are not connected through efficient payment rails or already interact on public chains.

The architecture implication is that banks need to design for coexistence, not a single winner. The routing decision — which form of digital money settles which transaction — should be made by cost, speed, finality, jurisdiction, client preference, and resilience. This is precisely the orchestration model described in the DORA compliance stack.

Digital Real-World Assets: The Capital Markets Transformation #

Tokenised RWAs — bonds, funds, equities, commodities, alternatives represented as on-chain tokens with legally enforceable claims — are where the deepest structural relevance for banking sits. The BCG report projects that in progressive scenarios, about 16% of global investable assets could be tokenised by 2035, with money-market instruments (25%–40% penetration), securitised debt (20%–30%), and alternatives (25%–35%) leading adoption. Government bonds (3%–5%) and listed equities (3%–7%) lag due to governance, market-structure, and exchange-incumbency constraints.

The economic case is clearest in settlement and post-trade. For a global bank with $100 billion in daily repo volume, BCG estimates $150–300 million in annual savings through reduced idle collateral and accelerated settlement cycles. Broadridge's DLR platform reported January 2026 average daily volumes of $365 billion and 508% year-on-year growth — production-grade institutional adoption at scale. J.P. Morgan's Tokenised Collateral Network and BlackRock's BUIDL fund demonstrate that the first movers are not experimenting; they are building operational infrastructure.

The connection to programmable liquidity is direct. Tokenised assets without tokenised settlement leave frictions at the cash interface, while tokenised cash without tokenised assets is a faster pipe connected to slow, batch-based post-trade processes. The combined proposition is atomic delivery-versus-payment and near-real-time finality. BCG draws an analogy to ETFs: the innovation was not the asset but the operating model — intraday liquidity, in-kind creation and redemption, tax efficiency, and exchange-based tradability — and early adopters (State Street, iShares, Vanguard) captured 75%–80% of global ETF assets through self-reinforcing advantages.

For a detailed examination of the BlackRock BRSRV and BSTBL filings — which show how tokenised money-market funds work around stablecoin yield restrictions under the GENIUS Act — see my earlier article on BlackRock's stablecoin-adjacent yield strategy.

Crypto: Where the Revenue Is Today #

Crypto is the most commercially mature segment — roughly $3 trillion in assets with an estimated ~$90 billion annual revenue pool, compared with ~$400 billion in traditional bank and market infrastructure trading revenues on $300 trillion of investable assets. The revenue yield relative to market capitalisation is structurally high at 2%–4%, driven by trading velocity, volatility, leverage, and 24/7 global market structure.

The addressable pool for regulated financial institutions — trading, derivatives, custody and prime services, staking — totals approximately $55 billion globally. Banks have structural advantages in institutional client relationships, regulatory credibility, risk management, and collateral optimisation. The strategic question is whether participation is defensive (retaining client relationships and wallet share) or offensive (seeking incremental growth), and which participation model — agency, prime, distribution, or full-stack — matches the bank's risk appetite and delivery capacity.

The Impact on Banks: Revenue at Risk, Revenue Addressable #

BCG models four scenarios for the evolution of digital assets, ranging from private-led rapid expansion to regulatory constraint and defensive reset. In the most aggressive scenario, banks could face roughly 10% smaller balance sheets, 14% lower revenues, 9% market share lost to non-bank financial institutions, an 8 percentage point higher cost-to-income ratio, and 30% lower profits by 2035 compared with a baseline without digital assets.

Three structural forces drive these pressures: tokenisation shifts value to issuers and consumers with less need for intermediaries; tokenisation accelerates the secular trend of value migration from banks to non-bank financial institutions; and parallel operation of legacy and tokenised rails creates temporary cost duplication.

But the report is equally clear about the opportunities. For an average G-SIB with a large bank-owned asset manager, a bottom-up analysis identifies very sizeable upside despite these headwinds.

The connection to the AI operating system for payments is important here. The same decisioning infrastructure that routes payments across multiple rails — scoring by cost, speed, finality, and compliance — applies to the orchestration of tokenised settlement, stablecoin on- and off-ramping, and collateral mobility. The institutions that build this decisioning layer once, across both traditional and tokenised rails, will have a structural cost advantage over those that build parallel control stacks.

The Architecture Principles That Matter #

BCG's CTO view identifies seven principles for digital asset technology strategy. Reading these alongside the architecture baselines in the DORA compliance stack and cloud-native banking, a consistent pattern emerges.

Treat DLT as Infrastructure, Not a Product #

The most common failure mode is isolated business-unit pilots, each selecting a different chain, custody model, and vendor stack. The result is duplicated wallets, inconsistent security models, and fragile integrations. BCG recommends centralised platform ownership at group CTO/COO level, with business units owning prioritised use cases on top of a common control framework. This maps directly to the platform-engineering discipline described in cloud-native banking — Kubernetes plus VM coexistence, DORA-tested resilience, sovereign cloud, and proof that critical services survive provider disruption.

Design Multi-Chain from the Outset #

Public and private rails will coexist. Individual blockchains are building their own ecosystems with distinct user bases, applications, and liquidity pools. Lock-in to a single chain creates strategic dependency. The architecture must be chain-agnostic, with abstraction and interoperability built in so that new chains can be added without redesigning the system. Deutsche Börse's D7 platform and BIS Project Agorá are examples of this pick-and-mix approach in action.

Build for Regulatory Reversibility #

Rules will evolve. Configurable controls, auditability, clean migration paths, and safe upgrade mechanisms are not optional. This principle is central to the DORA compliance architecture: the ability to evidence control under stress, across on-chain and off-chain systems, during market volatility, and in coordination with third parties. The agentic engineering blueprint extends this: AI agents operating within digital asset workflows need the same spec-driven governance, consent mandates, and audit trails as any other automated banking process.

Keep the Control Plane In-House #

Policy, key management, contract governance, limits, evidence, and incident response belong inside the bank. Partner for production engines, network connectivity, and standardised tooling where network effects or specialist capabilities matter. Buy selectively when speed, licenses, talent, or distribution are more valuable than building from scratch — but only if integration into the bank's control model is realistic.

This is the same build-partner-buy discipline that applies to payments orchestration, cloud infrastructure, and AI model governance. The principle is consistent: own what differentiates, partner for what standardises, buy for what accelerates.

Risk Control: Why AML, Custody, and Smart Contracts Are Growth Constraints #

BCG's CRO view makes a critical point that is often underweighted in digital asset strategy discussions: the ability to scale revenue is constrained by the design of the control framework, not just by demand or capital. In traditional banking, controls sit around the product. In programmable markets, controls sit inside the product.

Financial Crime Shifts from Customer-Centric to Flow-Centric #

Stablecoins now account for roughly 84% of illicit on-chain transaction volume globally. Crypto-related fraud represents only about 10% of reported incidents but approximately 50% of total losses. AML and sanctions risk arises through transaction flows and ecosystem connectivity, not just direct customer relationships. Wallet behaviour, transaction context, clustering patterns, and network relationships become primary risk signals alongside customer profiles.

This is the same shift described in the AI payments operating system: fraud scoring, sanctions screening, and compliance controls move from periodic batch monitoring to continuous, real-time, flow-based intervention. The digital asset context amplifies this requirement because settlement is near-instant and irreversible.

Custody Is a First-Order Control Risk #

In digital asset markets, control of private keys determines control of assets, with immediate legal and operational consequences. Legal custody, technical control, and instruction authority can be misaligned — an institution may be the legal custodian while technical control resides with a technology provider or protocol. In the first half of 2025, more than $3 billion was stolen across 119 crypto hacks, with centralised exchanges accounting for more than half of losses. Many incidents involved compromises in key management and custody controls rather than blockchain protocol failures.

BCG argues that custody should be elevated to the same level of scrutiny as systemically important payments or settlement infrastructure. The post-quantum migration adds a temporal dimension: cryptographic key material protecting digital asset custody today will need to be migrated to post-quantum algorithms within a timeframe that is shrinking faster than most institutions have planned for. The G7 published its migration roadmap in January 2026, and BIS Project Leap has demonstrated feasibility in live payment systems.

Smart Contracts Must Be Governed Like High-Risk Models #

Smart contracts encode assumptions, thresholds, and behaviours that determine outcomes automatically and at scale. Coding errors, parameter miscalibration, or unanticipated interactions between contracts produce systematic effects comparable to model risk — without the buffers traditionally provided by human review. In 2025, malicious approval exploits accounted for approximately $1.5 billion in losses. BCG recommends pre-deployment validation, continuous monitoring, controlled change management, and technically enforceable escalation and intervention authority — integrated into the broader risk framework.

Regulation: Converging Intent, Diverging Architecture #

BCG identifies four regulatory archetypes — market-driven (US, Canada, Brazil), bespoke regime (EU, UK, Japan), competitive hub (Singapore, Switzerland, Hong Kong, UAE), and sovereign-controlled (Mainland China). Policy intent is broadly aligned around financial stability, market integrity, AML controls, and operational resilience, but translation into enforceable rules remains uneven.

The EU's MiCA framework is the most prescriptive, requiring authorisation for e-money token issuers, fully segregated reserve backing, enforceable redemption at par, and a prohibition on EMT issuers paying interest. The US debate has shifted from legislative design to implementation, with the GENIUS framework defining issuer categories and reserve standards while negotiations continue on yield mechanisms and broader market structure.

Two issues remain particularly controversial: whether stablecoins can pay a yield, and what constitutes acceptable backing assets and liquidity management under stress. These debates are proxies for deeper questions about deposit substitution, run dynamics, and the sustainability of issuer business models.

For banks, the practical implication is designing for persistent multi-regime reality rather than a single global end state. The DORA compliance stack and UK Payments Forward Plan analysis both reinforce this: modular operating models with parameterised controls, so that the bank can ring-fence, re-route, or exit markets as rules evolve.

What This Means by Bank Type #

Global Banks #

Global banks should create platform-level orchestration so that each market, rail, token, and custody model does not become a separate operating model. BCG recommends centralised DLT architecture ownership, multi-chain design, and a common control framework. The first-mover advantage is strongest in settlement infrastructure, collateral mobility, and tokenised fund distribution, where scale creates self-reinforcing network effects. The sequencing should be: ledger-agnostic platform and controls first, then high-value use cases (repo, collateral, tokenised funds), then gradual expansion of ledger exposure.

Regional Banks #

Regional banks face an affordability question around building new DLT infrastructure when the benefits may be limited absent interoperability standards or a shared settlement venue. BCG argues that for banks that are not among the largest globally, a tokenised deposit without interoperability is likely a feature rather than a strategic asset. The better strategy is an interoperability roadmap: participate in industry consortia, prioritise the client wallet as the trust anchor, and focus on use cases where local market knowledge and regulated custody beat scale — treasury visibility, fraud prevention, Open Banking payments, and stablecoin on- and off-ramping.

Fintechs and PSPs #

The market now offers production-grade building blocks at both technology and operating service levels. The best propositions will bring orchestration, compliance evidence, or data intelligence — reducing complexity for banks rather than adding another isolated rail. This is the same principle identified in the agentic payments analysis: the fintech that provides consent infrastructure, liability boundaries, and audit-ready evidence alongside payment execution will outperform the one that merely provides a faster pipe.

Asset Managers #

Tokenisation is a structural growth story, not a defensive one. The primary levers are increased capture of currently non-fee-bearing assets into managed structures, product-mix shifts toward higher-fee alternatives through fractionalization and scalable distribution, and expansion of the service stack through programmable fund lifecycle management. BCG draws the ETF analogy explicitly: incumbents initially underestimated ETFs because they appeared economically redundant, but the innovation was the operating model. The same dynamic applies to tokenised funds.

The Five Practical Steps #

BCG concludes with a ten-step guide. Distilled to five immediate priorities for the next twelve months:

Quantify the economics. Within 90 days, model by business line: revenue pools at risk, revenue pools addressable, deposit and liquidity sensitivity, and the expected cost of dual rails. Without quantified exposure, digital assets remain abstract.

Lock in ambition and governance. Choose an ambition archetype for each relevant business line — defensive integrator, scaled participant, or infrastructure shaper. Appoint one accountable executive sponsor. Establish a board-level review cadence tied to scenario indicators.

Secure the client interface now. Prioritise secure wallet and custody capabilities, plus client-facing orchestration of on- and off-ramping, advice, and reporting. The wallet protects the client relationship regardless of which underlying money instrument wins.

Industrialise 2–3 high-value use cases. Typical candidates: cross-border treasury and programmable liquidity, tokenised funds and alternatives, repo and collateral mobility, and selective crypto services for existing clients. Avoid tokenisation in search of a problem.

Build the bank-grade control plane. Stand up the common DLT platform, key management, contract governance, AML tooling, reporting, partner framework, and exit paths. The first production use case should run on the same controls that later scale-up will depend on.

Conclusion #

Digital assets in 2026 are not an innovation question. They are a settlement, custody, compliance, and balance-sheet architecture question. The BCG report frames this as a structural transition — comparable to the telecom industry's dual-rail migration — in which the strategic objective is not predicting the winning rail but remaining relevant across multiple plausible futures.

The institutions that will shape the next settlement stack are those designing for orchestration now: multi-rail, multi-chain, multi-money, with bank-grade controls embedded in the product rather than layered around it. The institutions that wait will find themselves connecting to infrastructure designed by others, on terms set by others.

The architecture principles are consistent across every topic covered in this series — payments, treasury, compliance, cloud, AI, and now digital assets. Workflow first, data as the control plane, orchestration across rails and platforms, embedded compliance and evidence, and unit economics that justify scaling. The technology labels change; the design discipline does not.

Questions? Answers.

Why is this topic urgent in 2026?

Because the convergence of regulatory clarity (MiCA, GENIUS Act, UK frameworks), production-grade institutional platforms (Broadridge DLR, Kinexys, BlackRock BUIDL), and measurable stablecoin payment volumes has moved digital assets from experimentation to operational infrastructure. The dual-rail cost clock is now running.

What is the biggest implementation risk?

The biggest risk is architectural fragmentation: separate business units building separate DLT pilots, each with different chains, custody models, vendor stacks, and control frameworks. This creates duplicated cost, inconsistent security, and no path to scale.

What should a bank build first?

A secure client wallet and a ledger-agnostic control plane — key management, policy enforcement, monitoring, audit evidence, and integration patterns. These create optionality regardless of which chains, tokens, or money instruments scale.

How does this connect to existing compliance obligations?

Digital asset controls should be designed as an extension of the same architecture that supports DORA operational resilience, EU AI Act transparency, and data sovereignty — not as a parallel programme. The same evidence-production discipline, the same audit-readiness standard, the same governance model.

How should success be measured?

By revenue captured from addressable pools, cost of dual-rail operation, control-framework readiness (tested, not documented), client wallet adoption, and balance-sheet efficiency gains from tokenised collateral and settlement.

References #

• BCG, (2026). The Future of Digital Assets — BCG Flagship Report, May 2026 ⧉.
• RWA.xyz, (2026). RWA Market Data ⧉.
• Sebastien Rousseau, (2026). Stablecoins vs Tokenised Deposits in 2026 ⧉.
• Sebastien Rousseau, (2026). Programmable Liquidity in 2026 ⧉.
• Sebastien Rousseau, (2026). Tokenised Deposits in 2026 ⧉.
• Sebastien Rousseau, (2026). BlackRock BRSRV and BSTBL Filings Decoded ⧉.
• Sebastien Rousseau, (2026). DORA, the EU AI Act, and Data Sovereignty ⧉.
• Sebastien Rousseau, (2026). AI as the Operating System of Payments ⧉.
• Sebastien Rousseau, (2026). Cloud Native Banking in 2026 ⧉.
• Sebastien Rousseau, (2026). Securing the Ledger: Post-Quantum Migration ⧉.
• Sebastien Rousseau, (2026). Agentic Payments in Banking ⧉.
• Sebastien Rousseau, (2026). UK Payments Forward Plan ⧉.
• Sebastien Rousseau, (2026). Agentic Engineering for Banks ⧉.

Last reviewed 2026-05-30.