Treasury Yield Architecture for Regulated Fintechs: The EMD2 Method 1(b) Yield Map (2026)

Inside almost every EMI and PI we speak to, the same operating belief is in place: customer e-money funds cannot earn yield. The CFO assumes the safeguarding regime is uniformly cash-account-only. The treasurer plans float as a static cost centre. The board reads quarterly that the firm is safeguarding €40M, €120M, €400M of customer balances and accepts that those balances generate exactly zero return. With prime money market fund yields at roughly 4% in 2026, that operating belief is leaving genuinely material P&L on the table.

The belief is also wrong. Under EMD2 Article 7 and the parallel PSD2 Article 10, safeguarding is not a single regime. It is a choice between Method 1 — segregation — and Method 2 — insurance / comparable guarantee. Within Method 1, there are two sub-methods: Method 1(a) — segregation in a credit-institution account — and Method 1(b) — segregation by investment in secure, low-risk, liquid assets defined by the firm's home-state NCA. Method 1(b) is the regime that permits yield. The permitted-investments universe is NCA-defined — it is not a free-form treasury mandate.

This article is what we call the EMD2 Method 1(b) Yield Map: the structural comparison of Methods 1(a), 1(b) and 2; the NCA-by-NCA permitted-instruments universe; the deposit-guarantee-exclusion fact that quietly inverts the conventional view of which method is "safer"; the indicative 2026 post-cost yield achievable at each tier; the supervisory evidence pack required for an NCA approval; and the PSD3 forward look on whether the yield universe tightens. It is supervisor-sensitive content — read the disclaimer below before treating any of it as actionable.

The text of EMD2 Article 7¹^[1] — and the equivalent PSD2 Article 10²^[2] applicable to payment institutions — sets out a binary structural choice. Method 1 is segregation: customer funds are kept legally and operationally separate from the EMI's or PI's own funds, by either depositing them in a credit-institution account or investing them in secure, low-risk, liquid assets. Method 2 is insurance: an authorised insurer or credit institution provides a guarantee equal to the customer-fund balance, payable in the event the institution fails to meet its safeguarding obligations.

Customer funds sit in a designated safeguarding account at an authorised credit institution. The account is in the EMI's or PI's name but is contractually labelled as held on trust for, or segregated from, the institution's own funds. This is the default architecture almost every EMI begins with: it is operationally simple, requires no NCA-side approval beyond the standard authorisation, and the safeguarding bank typically pays zero interest on the segregated balance. The float earns nothing.

Customer funds are invested in instruments defined by the home-state NCA as "secure, low-risk, liquid assets". The instruments are held in a segregated account at an authorised custodian. The choice of instruments is constrained — it is not a corporate-treasury mandate — but where the constraint is met, the assets earn yield and the yield accrues to the EMI or PI (subject to the firm's customer terms and any local interest-on-customer-balances disclosures). This is the regime that opens the treasury opportunity.

An insurer or credit institution issues a guarantee equal to the customer-fund balance. The institution still holds the funds operationally — typically at a credit institution — but the segregation requirement is satisfied through the guarantee rather than the legal separation. Method 2 earns no yield and adds a fixed annual premium — typically 25–60 bps of the safeguarded balance — paid to the insurer. It is the most expensive of the three options and is rarely chosen at scale; it appears mostly in EMIs whose Method 1 access is constrained by the local credit-institution market.

The structural punchline: only Method 1(b) generates yield. Method 1(a) is yield-zero by market convention; Method 2 is yield-zero plus a recurring premium drag. Any treasury programme that intends to monetise customer float is operating under Method 1(b) by definition.

EMD2 leaves the precise definition of "secure, low-risk, liquid assets" to the home-state NCA. The result is a patchwork: the instrument universe a Lithuanian EMI may invest in differs from what an FCA-authorised UK EMI may invest in, which differs again from the conservative position taken by BaFin or the Central Bank of Ireland. Three NCA positions matter most for current EEA and UK EMI / PI populations.

The Bank of Lithuania³^[3] — the EEA's most-used EMI authorisation jurisdiction — defines the Method 1(b) permitted universe in three categories. First, publicly-traded bonds issued by governments, central banks, or corporates, with a long-term credit rating of BBB+ or higher from a recognised rating agency. Second, short-term deposits — maturity ≤ 1 year — held with credit institutions. Third, collective investment schemes that themselves invest exclusively in the above two categories. The bond and deposit ratings, maturity, and CIS-look-through requirements are the operative constraints.

In practice this means a Lithuanian EMI under Method 1(b) can hold a portfolio of investment-grade EU government and corporate bonds, short-dated bank deposits, and prime / government MMFs that themselves invest in those instruments. The practical 2026 expression for most EMIs is dominated by short-dated bank deposits and AAA-rated MMFs — the bond bucket adds duration risk that most EMI treasurers prefer to outsource to a CIS manager rather than run in-house.

The FCA's CASS 15⁴^[4] regime came into force on 7 May 2026 — three days before this article publishes. CASS 15 reorganises UK safeguarding around statutory-trust and CASS-style segregation principles already familiar from the investment-firm regime, and it explicitly addresses the permitted-assets question for the UK equivalent of Method 1(b). The default permitted assets remain narrow — credit-institution deposits and government instruments — but the FCA has signalled willingness to approve additional assets on a firm-by-firm basis. The standout addition is AAA-rated LVNAV money market funds — the EU regulatory MMF category for low-volatility net-asset-value short-term funds — which the FCA has explicitly recognised in policy commentary as eligible subject to firm-side consumer-protection demonstrations.

The CASS 15 architecture matters beyond the UK because it is the most recent major-market regulator review of the safeguarding-yield question. LVNAV MMFs are now supervisor-recognised as a Method-1(b)-equivalent instrument tier in the UK, and EU NCAs are likely to take the FCA's reasoning seriously when they revisit their own permitted-investments lists under PSD3.

At the conservative end sit BaFin, the Central Bank of Ireland, and a number of smaller-market NCAs. The default expectation is that EMIs and PIs operate Method 1(a) — credit-institution account — and any move to Method 1(b) requires a substantive permitted-investments policy submitted in advance, ongoing reporting on holdings, and a narrower instrument universe than Lithuania's. In practice the conservative NCA universe converges around: short-dated sovereign debt of the home jurisdiction or an EU AAA-rated sovereign; short-term deposits at a tier-1 EU credit institution; and AAA-rated EU MMFs (LVNAV or public-debt CNAV). Corporate bonds and lower-rated sovereign exposures are typically not approved.

The result is a real jurisdictional gradient in achievable yield. A Lithuanian EMI operating Method 1(b) at the boundary of its permitted universe can run a meaningfully higher-yielding portfolio than a German or Irish EMI operating at the equivalent boundary — at the cost of a more complex ALCO process and a less defensible position if the NCA later tightens the universe.

The table is indicative — every firm should verify the current published position with its home-state NCA before designing a Method 1(b) portfolio. The point of the map is the shape of the cross-NCA difference, not the cell content as a static reference.

The conventional intuition treats Method 1(a) as the "safest" option: customer money sits at a regulated bank, the firm earns no yield, but the funds are presumed protected. In its August 2019 opinion the European Banking Authority⁵^[5] addressed the question directly and concluded the opposite: EMI and PI balances at credit institutions are excluded from EU deposit guarantee scheme coverage. They are not eligible deposits under the Deposit Guarantee Schemes Directive. If the safeguarding bank fails, the EMI's customers do not have direct DGS recourse on the safeguarded balance — they have a claim against the failed bank's insolvency estate alongside other unsecured creditors, with the EMI as the formal account holder.

The implication for the safeguarding-method choice is significant. Method 1(a) concentrates the entire customer-fund balance at one (or in larger EMIs, two) credit institutions, with no DGS protection sitting underneath. Method 1(b) — properly designed — disperses the same balance across short-dated sovereign exposures, multiple credit-institution deposits below diversification thresholds, and AAA-rated MMFs that themselves diversify across hundreds of underlying instruments. From a pure customer-protection standpoint, a well-designed Method 1(b) portfolio can be more diversified and less concentration-risky than a Method 1(a) deposit at a single safeguarding bank.

This does not flip the choice automatically — Method 1(b) introduces investment-policy governance, custodian risk, mark-to-market mechanics, and the operational complexity of running a treasury portfolio under supervision. But it does undermine the lazy assumption that Method 1(a) is the conservative default; in many supervisory conversations from 2024 onwards, NCAs have been the ones pointing this out.

Rates have stayed higher for longer than most 2025 forecasts assumed. The Fed paused after the first cut of the cycle and consensus now expects only one further cut in 2026. The ECB has tracked a similar trajectory at lower absolute levels. The result for short-dated USD instruments is a yield curve sitting in the 3.5–4.5% range for institutional MMF and short-bank-deposit tenors; for short-dated EUR instruments it is a 1.75–2.75% range. A multi-currency EMI safeguarding USD, EUR, and GBP balances will therefore see materially different per-currency yields and needs to think about the architecture currency-by-currency.

Indicative 2026 yields by Method 1(b) instrument tier, post-cost (gross MMF or deposit yield less custodian and management fees, illustrative; actual yields vary by manager, share class, and currency):

The practical 2026 picture, as The Payments Association⁶^[6] set out in its overview of MMFs as a strategic EMI treasury tool: an EMI safeguarding €100M of EUR-denominated customer balances, moving from Method 1(a) (zero yield) to a Method 1(b) AAA EU MMF allocation at the conservative end of the universe, can plausibly add €1.8M–€2.4M of annual yield — a meaningful proportion of operating profit at typical EMI margins. Scaled to a €500M-float EMI, the same architecture is a €9–12M annual line item that did not exist in the firm's P&L the year before.

Two caveats. First, customer-disclosure obligations — and in some jurisdictions explicit interest-on-customer-balances rules — may require a portion of the yield to be passed back to the end-customer. The architecture choice cannot be modelled without the customer-T&Cs analysis. Second, MMF NAVs are not perfectly stable; the LVNAV structure permits limited NAV deviation before forced conversion. Operational architecture must include same-day liquidity to cover redemption flows.

Switching from Method 1(a) to Method 1(b) — or expanding the Method 1(b) instrument universe beyond the NCA's defaults — requires a structured submission. The headline components are consistent across NCAs even where the specific universe differs:

• A board-approved permitted-investments policy specifying eligible instruments, rating thresholds, maturity limits, single-issuer concentration limits, and CIS look-through requirements.

• A custody architecture diagram identifying the safeguarding custodian, account naming convention, segregation legal opinion, and the operational flow that demonstrates customer funds remain segregated through the investment cycle.

• A documented liquidity-risk framework — what proportion of the safeguarded balance is held in same-day-liquid form to cover redemption peaks, what the trigger is to liquidate longer-tenor positions, and how the firm models customer redemption stress.

• An ALCO governance document — composition, frequency, decision authorities, escalation triggers, and the position the function takes in the firm's three-lines-of-defence map.

• The customer-disclosure pack — terms and conditions amendments, public disclosures, and any per-jurisdiction interest-on-customer-balances treatment.

• Ongoing reporting cadence to the NCA — typically monthly or quarterly portfolio composition, breach reporting, and annual policy review confirmations.

Realistic timeline: a well-prepared firm with an existing safeguarding-bank relationship can move from Method 1(a) to a basic Method 1(b) MMF allocation in 4–8 months of NCA-side work. Adding bond-ladder or wider corporate exposure typically extends to 9–14 months because the NCA wants to see the policy run through at least one supervisory review cycle at the simpler tier first.

The PSD3 / Payment Services Regulation package — currently in late legislative stages and expected to apply from 2026–2027 onwards — collapses the EMD2 / PSD2 distinction and rewrites large parts of the safeguarding architecture. Three changes matter for the Method 1(b) yield universe.

First, mandatory diversification. Above a threshold (still under technical-standards work), customer funds must be spread across at least two credit institutions. This formalises something most large EMIs already do informally and is consistent with the Three-Bank Resilience Standard. The implication for Method 1(b) is that single-custodian designs become non-compliant — multi-custodian Method 1(b) is the baseline.

Second, central-bank safeguarding becomes permitted. An EMI or PI may, subject to NCA permission and central-bank willingness, hold safeguarded funds directly at the home-state central bank. This is operationally a Method 1(a)-style architecture but with a higher-quality counterparty than a commercial credit institution. It does not generate yield in itself but it does narrow the case for Method 1(a) at commercial banks: if the firm wants pure-safety segregation, the central bank is the cleaner answer; if the firm wants yield, Method 1(b) remains the only path.

Third, 24-hour segregation — the timeline for moving received customer funds into the safeguarding pool tightens from the current five working days to next business day. Daily reconciliation is formalised. Method 1(b) portfolios designed around weekly or monthly rebalancing will need to move to daily flows, which has cost implications for the bond-ladder tier and reinforces MMFs as the simpler operating tool at scale.

What PSD3 does not do, on current drafts, is narrow the underlying permitted-instruments universe. The technical standards on Method 1(b)-equivalent assets are due in 2026–2027 and may reshape specific instrument eligibility, but the principle of a yield-bearing segregation route is preserved. PSD3 hardens the architecture around Method 1(b); it does not eliminate the regime.

The EU Anti-Money Laundering Regulation (AMLR) applies from 10 July 2027. For Method 1(b) yield architecture the AMLR effect is indirect but real: supervisory inspection cycles will include AML harmonisation overlays that increase the documentation density EMIs and PIs already maintain, and any deficiency in the safeguarding-investment governance pack will surface in the same inspection. The practical advice is to build the Method 1(b) policy and the AMLR-aligned governance pack as a single inspection-readiness exercise rather than as two separate workstreams.

Yes — under EMD2 Article 7 Method 1(b), customer e-money funds may be invested in secure, low-risk, liquid assets defined by the firm's home-state NCA. The yield accrues to the EMI subject to customer-disclosure obligations. Method 1(a) — credit-institution account — is the most common architecture and earns no yield; Method 2 — insurance — earns no yield and adds an annual premium cost. Method 1(b) requires a documented permitted-investments policy and, in most NCAs, explicit notification or approval.

The instrument universe is defined NCA-by-NCA. The Bank of Lithuania permits investment-grade publicly-traded bonds (BBB+ minimum), short-term (≤ 1 year) credit-institution deposits, and CIS investing in those categories. The FCA, under CASS 15, permits credit-institution deposits, government instruments, and AAA-rated LVNAV money market funds by approval. BaFin and the Central Bank of Ireland take a narrower view, typically requiring AAA EU sovereign exposures, short-term tier-1 EU bank deposits, and AAA EU MMFs.

No. The EBA's August 2019 opinion confirmed that EMI and PI balances at credit institutions are excluded from EU Deposit Guarantee Scheme coverage. If the safeguarding bank fails, the EMI's customers do not have direct DGS recourse on the safeguarded balance — they hold an unsecured claim against the bank's insolvency estate via the EMI as account holder. This is a material risk fact and undermines the conventional view that Method 1(a) is the safest option.

CASS 15 came into force on 7 May 2026. It reorganises UK e-money and payment institution safeguarding around statutory-trust and CASS-style segregation principles, formalises daily reconciliation and reporting, and explicitly opens an approval pathway for additional permitted assets — most notably AAA-rated LVNAV money market funds. Firms must update their safeguarding architecture, segregation legal opinions, and internal governance to align with the new regime; the FCA has signalled willingness to engage on firm-by-firm permitted-asset extensions.

In the 2026 environment, post-cost yields of 1.65–2.65% on EUR balances and 3.40–4.30% on USD balances are achievable across the AAA MMF and short-bank-deposit tiers. Investment-grade short-dated bond ladders push the EUR range to 2.10–2.85% and USD to 3.70–4.60% but require more governance overhead. For a €100M EUR-denominated EMI float, a conservative MMF allocation typically adds €1.8M–€2.4M of annual yield versus Method 1(a). At €500M float the figure scales to €9–12M.

On current drafts, no. PSD3 hardens the architecture — mandatory diversification across multiple credit institutions, 24-hour segregation, formal daily reconciliation, central-bank safeguarding option — but preserves the principle of a yield-bearing segregation route. The technical standards on permitted instruments are due in 2026–2027 and may reshape specific eligibility at the margin, but the regime survives. Firms designing today should anticipate multi-custodian Method 1(b) as the future baseline.

• EMI Safeguarding Architecture — the structural foundation on which Method 1(b) yield architecture sits.

• The Three-Bank Resilience Standard — the diversification pattern that anticipates PSD3's mandatory-multi-custodian baseline.

• Multi-Currency Treasury for CASPs — the per-currency yield-environment lens that pairs with the Method 1(b) Yield Map.

• Cost of Banking for a Regulated Crypto Firm 2026 — the cost side of the safeguarding architecture, against which yield is measured.

• Banking Access for Regulated Fintechs — our service: safeguarding architecture design, NCA-side approval workstreams, treasury counterparty introductions.

The EMD2 Method 1(b) Yield Map is not a marketing phrase. It is the structural answer to the misconception that customer e-money funds cannot earn yield — and the practical map of where, at which NCA, under what governance, and at what indicative post-cost yield they can. Build the policy first. Get the NCA conversation right. Then design the portfolio against the home-state instrument universe and the currency-by-currency yield environment. The treasury opportunity is real, but the supervisor sets the boundary — and a Method 1(b) programme that fails its first inspection unwinds the entire P&L gain in a single quarter.