🇺🇸Money Transmitter Licenses in the US: Fifty Licenses Instead of One

Why No Federal Money Transmission License Exists

Anyone arriving from the EU or the UK expects a single authorisation and a passport across the bloc. The United States offers the opposite: money transmission sits in the states' police power rather than in banking law, and no federal money transmission license has ever existed. There is a federal obligation, but it registers rather than authorises.

FinCEN requires registration on FinCEN Form 107 (Registration of Money Services Business) within 180 days of establishing an MSB, renewal every 24 months by 31 December, and re-registration on a transfer of more than 10% of voting power or equity or a rise of more than 50% in the agent count.

The civil penalty for failing to register — $5,000 per violation under 31 U.S.C. § 5330(e), each day counting separately — applies as $10,556 after the inflation adjustment in 31 CFR 1010.821; failure to register is also a predicate of the federal unlicensed-transmission offence, which carries up to five years' imprisonment. None of it confers the right to operate in a single state: the federal MSB regime is covered separately.

What the states do share is a filing pipe, not a regulator. Almost all of them accept applications through NMLS, but fifty banking departments then reach fifty independent decisions on the same business. Montana is the only state with no money transmitter regime at all, which is why "50-state coverage" in marketing copy almost always means 49 states plus the District of Columbia and the territories.

The scale of the system explains why it survives its own inefficiency; the CSBS figures for 2023 and Q2 2024:

Concentration is close to absolute: ten firms hold nearly all of the industry's stored value.

The key parameters of the regime as of September 2026:

MTMA: the Countable Was Harmonised, the Contested Was Not

The attempt to build a common standard without a federal statute is the Money Transmission Modernization Act, a model law approved for state adoption by the CSBS Board of Directors (text). The CSBS enactment table updated in September 2026 lists thirty-five states with full or partial enactments, including Louisiana (H.B. 1230, effective 01.07.2026), Oklahoma (H.B. 3521, effective 01.11.2026), Alaska (effective 01.07.2027) and Delaware (effective on the earlier of 06.07.2027 or publication of final regulations); Michigan is weighing two competing bills, and the CSBS MTMA page still gives an older headline figure of thirty-one.

Two caveats attach to that count. First, enactment is legally heterogeneous: Maryland entered the tracker through a regulation, COMAR 09.03.14, and Rhode Island began in 2022 with banking bulletins before a partial statute in 2024; California, Connecticut, Rhode Island and Utah are recorded as partial adopters. Second, the much-quoted CSBS line about 99% of industry volume means that transmitters licensed in at least one MTMA state account for that share. It is a concentration statistic, not a coverage statistic: the large players are simply licensed everywhere. For a start-up operating in three states, none of which has adopted MTMA, harmonisation delivers nothing at all.

Exactly three things were unified, and all three are countable:

Type I within that coverage comprises cash and equivalents, including FBO accounts at a federally insured depository institution, and US government obligations. Control is unified separately: the right to vote 25% or more is control, a rebuttable presumption bites at 10%, immediate family interests aggregate, and an acquirer must obtain written approval before closing (§ 2.01(f), § 6.01(a)).

Beyond that lies the uncountable — and the uncountable is what determines whether a license is needed. The exemption list is not unified. Consumer-versus-commercial scope is not unified either. The virtual currency module (Article XIII) has been adopted by a handful — Minnesota, North Dakota and Maine, joined by Delaware's 2026 act, which includes it and takes effect in 2027. There is no de minimis threshold in the model law at all: exemptions run by activity type, never by volume.

Crypto on the balance sheet is addressed through interpretive guidance rather than the statute. Under Tangible Net Worth and Virtual Currency (approved 21.05.2025), virtual currency is an intangible asset under GAAP and is deducted in computing TNW while remaining in the asset base used to compute the requirement itself. In the CSBS worked example, of $25 million in BTC only $2.5 million matching customer BTC obligations escapes deduction — the other $22.5 million leaves capital. In March 2026 came Stablecoin Tangibility: fiat-backed stablecoins count as tangible assets where there is a contract with the issuer and an unconditional right to redeem at par, and stablecoins issued under the GENIUS Act meet the criteria automatically.

Agent-of-the-Payee: Where the Line Actually Runs

The trigger is not a business model but a fact pattern: the company receives money or monetary value from one person for delivery to another, and takes control of it. Hence the practical dividing line — FBO account versus direct settlement. In the payment facilitator model, money transmission arises when the PayFac receives funds from an acquirer or processor in order to distribute them to sub-merchants; direct settlement, where the networks pay sub-merchants directly, creates no exposure, while PayFac-settled flows through an FBO account do. The same question surfaces in any card programme and any BaaS chain where customer money sits under the operator's operational control even briefly.

The main way around it is the agent-of-the-payee exemption. MTMA § 3.01(b) requires three conditions at once: a written agreement, the payee publicly holding the agent out as its agent, and — the decisive one — payment being treated as received by the payee when the agent receives it. The exemption is enacted state by state, and the wording differs. California has its own version at Fin. Code § 2010(l): a preexisting written contract under which delivery to the agent satisfies the payor's obligation. Vermont, which adopted the model law in 2024, carries the three MTMA conditions in 8 V.S.A. § 2504(2) alongside a separate exemption for payment-system operators. A structure that relies on the exemption therefore has to be checked against each state's text, not against the model law, and the contract wording has to meet the strictest version in the footprint.

Payroll: Agent of the Payor, Not the Payee

Payroll is where the 2026 divergence is sharpest. The logic is mirrored: a payroll processor acts for the payor, not the payee, and falls outside ordinary agent-of-the-payee relief. Iowa enacted its own agent-of-the-payor exemption; Maryland exempts payroll agents of the payor from 01.10.2026 (S.B. 261); Nebraska from 26.02.2026 exempts small cases — fewer than 20 employees, or servicing fewer than 50 Nebraska-resident employees (L.B. 717); and Michigan is simultaneously considering H.B. 5544, which pulls payroll processors into the regime, and S.B. 835, which exempts them (CSBS tracker). In 2027 the same product will be licensable in one state and exempt next door.

Crypto: Three Parallel Regimes and the States That Matter

There is no single crypto regime — there are three, running at once. First: crypto assets fall inside the general money transmission definition and an ordinary MTL is required — Pennsylvania wrote this into its statute, requiring a licence for transmitting virtual currency from 26.08.2025 (DoBS notice on Act 7 of 2025). Second: a dedicated license on top of, or instead of, the MTL — BitLicense in New York and DFAL in California. Third: exemption, statutory in Wyoming and New Hampshire or set out in regulator guidance elsewhere, plus Montana, where there is nothing to regulate under.

The gap between a statutory exemption and regulator guidance is the gap between a statute and a memorandum. Wyo. Stat. § 40-22-104(a)(vi) exempts buying, selling, issuing and even taking custody of payment instruments in the form of virtual currency — unusually broad — and New Hampshire's RSA 399-G:3, VII exempts dealing solely in convertible virtual currency. Texas rests on Supervisory Memorandum 1037, revised on 28.01.2025, which brings fiat-pegged, reserve-backed, redeemable stablecoins within the money transmission analysis and offers no general guidance on other virtual currencies, leaving them to case-by-case determination. A memorandum can be rewritten in a day.

There is countervailing movement too. CSBS records the 2025 adoptions in Colorado, Mississippi and Virginia as full MTMA introductions without the virtual currency title that Minnesota, North Dakota, Maine and Delaware took (CSBS legislative update, August 2026), and Virginia went further: its new act excludes virtual currency from the definition of money (Va. Code § 6.2-1922). Modernisation conceived as harmonisation has, in several states, objectively deregulated crypto.

Twenty-one states on one grid

The decision "which states, in what order" needs more than the four or five jurisdictions usually quoted. Twenty-one states below on five axes plus the crypto overlay; a dash means no figure is given here — either none is published or it could not be confirmed from a primary source at the check date — not that no requirement exists, and "MTMA ladder" means the model-law formulas set out above apply as enacted.

The Nevada, Idaho and Ohio lines come from the statutes and the regulator's own page — NRS 671, the Idaho Department of Finance and ORC 1315.05 with 1315.07. The Florida, Wyoming, California and Hawaii figures are taken from the statutes cited in the grid; every figure should still be confirmed with the state regulator before filing.

What the grid decides

The first thing it shows is that adopting the model law does not make a state cheap or slow: it makes it predictable. Nevada wrote the MTMA formulas into NRS 671 almost verbatim, so its bond is capped at $500,000 and its tangible net worth starts at $100,000 — and its published fee ladder (a survey fee of no more than $500, an initial licence of $200 to $400, a renewal of no more than $400) is among the lowest in the country. California adopted parts of the same model law and still sets bonds up to $7 million and charges a $5,000 application fee. The variable that matters is not MTMA status but whether the state legislature also rewrote its bond schedule and its fee schedule.

The second is that the expensive states are expensive on different axes, so a budget built on one of them misprices the rest. Pennsylvania's cost is a flat $1,000,000 bond that does not scale down for a small programme. Ohio's is capital: a $500,000 minimum net worth under § 1315.05 and a security device that starts at $300,000 and runs to $2,000,000 under § 1315.07, with no statutory fee at all — the superintendent sets the annual fee under § 1315.13 and bills the applicant for the examination. A start-up with $600,000 of equity clears Nevada and Idaho comfortably, clears Pennsylvania only by buying a bond it cannot collateralise, and fails Ohio's capital test outright.

The third is that the crypto column and the money columns point in opposite directions. The states that modernised most recently — Virginia, Colorado and Mississippi in 2025 — took the model law without its virtual currency title, and Virginia went further and removed virtual currency from the definition of money. The states that adopted the title are Minnesota, North Dakota and Maine, with Delaware to follow in 2027. So a crypto business reading the grid top to bottom finds harmonised capital rules sitting above a crypto perimeter that is diverging, not converging, and the two states that decide the programme — New York with the BitLicense and California with DFAL from 01.07.2026 — sit outside the harmonisation altogether.

Where to file first therefore depends less on ambition than on which of four constraints binds:

Take a remittance operator with $600,000 of equity, average daily in-state liability of $400,000 in its three largest states and no crypto leg. Idaho and Nevada cost $100 and up to $500 in filings, and bonds of $10,000 and $400,000 respectively — Nevada's bond is 100% of average daily liability because that exceeds $100,000, and it stays below the $500,000 cap. Texas adds its own filing fees and volume-linked security. Pennsylvania adds a $1,000,000 bond regardless of volume, and Ohio requires $500,000 of net worth before the first customer — which, against $600,000 of equity, is the point at which the programme either raises again or drops two states. That is the real shape of a national rollout: not fifty fees, but three or four states that set the capital floor for the other forty-six.

New York's alternative to the BitLicense is a limited purpose trust charter, which carries both fiduciary powers and money transmission without a separate NY MTL. In California, from 01.07.2026 digital financial asset business requires a DFAL licence unless a complete application was filed on or before that date and is still pending (Fin. Code § 3201); a money transmitter licence is not among the DFAL exemptions in § 3103(b).

What Fifty States Cost, and How Supervision Actually Works

The entry budget is not a single fee schedule. It combines state application and renewal fees, surety-bond premiums tied to each state's required amount, registered-agent and corporate filings, NMLS administration, audits, counsel and an in-house compliance function. The total depends on the licensed states, payment volume, permissible-investment model and bond underwriting; a national rollout should therefore be budgeted as a multi-year compliance programme, not a fixed “licence price”. Current line items come from the Nationwide Multistate Licensing System and each state regulator, checked before launch.

Timelines are set by the worst link, not the average: each state reviews on its own clock and no regulator publishes a nationwide timetable, so the rollout plan is built around the slowest file.

The most underestimated cost is not a fee but 100% permissible investments. This is neither capital nor a reserve: it is an obligation to hold full coverage of customer obligations in eligible assets. The operator's float is not working capital, and industry-wide actual coverage stands at 153% (CSBS). Other regimes impose the same duty under different names — an EEA payment or e-money institution must safeguard client funds it still holds at the end of the business day after receipt (PSD2 Art. 10(1)(a)) — but none of them multiplies it by fifty licensing files. The eighteen payment regimes are lined up on capital, clock and safeguarding in the fintech licence map, and the crypto overlays in the crypto licence map.

Supervision: MMLA, Networked Supervision and the Call Report

Supervision, by contrast, is not fifty independent examinations. Under the Multistate MSB Licensing Agreement states coordinate the processing of licence and change-of-control applications — what the model law calls a multistate licensing process (MTMA § 2.01(s)) — and each participating state adds its own requirements.

Layered on top is MSB Networked Supervision, under which states work together on a single, comprehensive examination of a company licensed in several states.

Reporting is the quarterly report of condition, the MSB Call Report filed through NMLS: under MTMA § 7.01 it is due within 45 days of each quarter-end and covers financial information, state-by-state transaction activity and permissible investments, while destination-country data goes only into the fourth-quarter report.

Criminal Liability: 18 U.S.C. § 1960 and the Blanche Memo

A state licensing breach in the United States is a federal crime. 18 U.S.C. § 1960 punishes whoever knowingly conducts, controls, manages, supervises, directs or owns an unlicensed money transmitting business, with a five-year maximum. There are three predicates, and the first decides most cases: subsection (b)(1)(A) reaches operating without a state license where doing so is an offence, "whether or not the defendant knew that the operation was required to be licensed". Knowledge of the licensing requirement is not an element, and its absence is not a defence.

The Deputy Attorney General's memorandum of 07.04.2025 (the Blanche memo) narrowed application to digital assets: the DOJ will not charge "regulatory violations", including § 1960(b)(1)(A) and (B), except where the defendant knew of the requirement and violated it willfully; the NCET unit was disbanded immediately. The door left open is exactly one word wide — "willfully" — and it creates an uncomfortable paradox: any internal memo reading "we may need an MTL in state X" becomes evidence of knowledge. That is a reason to hold such analysis under privilege, not a reason to skip it.

Enforcement: OKX, Samourai and Storm

The narrowing did not stop enforcement. Aux Cayes Fintech, the OKX operator, was prosecuted in the Southern District of New York on charges filed on 24.02.2025 (No. 1:25-cr-00069), including operating an unlicensed money transmitting business.

The Samourai Wallet founders were prosecuted in the same district (United States v. Rodriguez, No. 1:24-cr-00082) although the service was non-custodial. The lesson is unforgiving: a federal regulator's view that an operator is not a money transmitter is no defence under (b)(1)(A), because the predicate there is state law.

The open question is United States v. Storm (Tornado Cash, No. 1:23-cr-00430, S.D.N.Y.). Until that case resolves, whether a developer of non-custodial software can be a money transmitter remains a live legal question rather than a rhetorical one.

The Federal Bypass: GENIUS Act, OCC Charters and Letter 1192

For the first time there is a lawful route to not obtaining fifty licenses at all, and it is built from two independent parts.

GENIUS Act: Preemption for Issuers Only

The first part is the GENIUS Act, Public Law 119-27 of 18.07.2025. Its § 5(h) expressly supersedes and preempts "any State requirement for a charter, license, or other authorization to do business" as applied to a federal qualified payment stablecoin issuer. State-qualified issuers face a $10 billion outstanding ceiling, after which § 4(d)(2) allows 360 days to move into the federal framework.

The statute takes effect on the earlier of 18.01.2027 (18 months after enactment) or 120 days after the principal regulators issue final implementing rules; the Act required those rules within one year of enactment, by 18.07.2026 (§ 20 and the rulemaking section).

The critical limit: preemption runs to issuers only. Secondary intermediaries, wallets and exchanges are not addressed — § 3(h) exempts only peer-to-peer transfers and self-custody — and from 18.07.2028 § 3(b)(1) imposes a countervailing restriction on them: they may not offer a stablecoin that is not from a permitted issuer. That sits above the MTL rather than replacing it; the regime itself is covered separately.

The OCC Trust Charter and Letter 1192

The second part is the national trust charter: on 12.12.2025 the OCC conditionally approved the conversion of Fidelity Digital Assets, a New York trust company that had held state money transmitter licences, into a national trust bank (Conditional Approval No. 1355).

The unlock came with OCC Interpretive Letter 1192 of 12.05.2026: the bank "is not required to comply with state money transmitter licensing requirements", and the conclusion extends to "any similar money transmitter licensing requirements that purport to apply to the Bank, including by limiting their licensing exemptions to a subset of national banks".

The reasoning is two-part: the "prevent or significantly interfere" standard, plus visitorial powers under 12 U.S.C. § 484 — state MTL statutes give states examination and document-production powers that are fundamentally inconsistent with the OCC's exclusive jurisdiction. The second argument is broader than the first: it reaches supervision, not merely licensing.

The Limits of Letter 1192

The OCC expressly stated that the letter is not a preemption determination for purposes of 12 U.S.C. § 25b and that the conclusion depends on the facts as represented; subsidiaries and agents are not addressed, and the dedicated crypto regimes — BitLicense and DFAL — are untouched.

Note also that the whole construct is the product of the OCC's current leadership: a federal charter is cheaper than fifty licenses today, but politically more volatile than they are. The adjacent questions are direct access to Federal Reserve settlement and the trust charter itself.

When a Family Office Becomes a Money Transmitter

One diagnostic question comes first: is the company a counterparty of a licensee, or a licensee itself? A family office moving the family's own money is not a money transmitter — the "on behalf of the public" element is missing. Exposure begins the moment third-party money appears. Four structures burn people repeatedly:

• an internal treasury platform serving several related but not legally affiliated families;
• an SPV pooling co-investor funds ahead of a closing;
• crypto custody "for friends";
• FX netting between portfolio companies.

Section 1960(b)(1)(A), meanwhile, requires no knowledge of the licensing requirement.

Taking equity in a licensee is itself a regulatory event: 25% of the votes constitutes control, the presumption is rebuttable from 10%, and prior written approval is required from every state. For an international family office that means fingerprinting, FBI and state criminal history checks, credit reports and structure disclosure down to beneficial owners — and where a control person has resided abroad within the last ten years, an investigative background report by an independent search firm covering credit, court, criminal, employment, media and regulatory history (MTMA § 5.04(b)). Families with non-US residence histories should assume this applies to them by default. The workaround is standard rather than inventive: subscribe through non-voting classes and sign a passive investor attestation — a route the model law expressly provides: a passive investor that cannot elect a majority of managers, is not employed by the licensee, has no controlling influence and attests to this rebuts the presumption of control (MTMA § 2.01).

When diligencing the licensee itself, ignore the licence count in the website footer. Three things are useful: the ratio of permissible investments to outstanding obligations from the MSB Call Report — the industry sits at 153%, so anything under 110% at a specific firm needs explaining; participation in MSB Networked Supervision, which means a coordinated examination rather than scattered ones; and the dates of recent examinations and enforcement actions in NMLS Consumer Access. Separately, look at balance sheet composition: crypto assets that are not fiat-backed are deducted from tangible net worth as intangibles, so a sizeable crypto balance at a licensee means capital is weaker than the reported figures suggest.

Four Forks in Licensing Strategy

Whether a license is needed at all. The order of analysis is fixed: does the company receive money or monetary value from one person for delivery to another; does control over funds arise (FBO versus direct settlement); does an agent-of-the-payee or payment-processor exemption apply, and in which wording, state by state. The answer is almost never nationwide, and the design has to be built for the worst state.

Fifty licenses or a federal charter. Since 2026 this is a genuine choice rather than rhetoric: IL 1192 lifts the MTL from a national bank and extends to any similar state requirement. But a charter means OCC supervision, capital requirements, a multi-stage application process, a conditional approval that does not by itself confer the right to operate, and BitLicense and DFAL still standing. Add the political volatility of the construct.

Filing sequence. Start with the home state and the fast, low-fee, low-bond jurisdictions (Idaho, Nevada), run the MMLA track in parallel, and place New York and California early in the calendar because they run separate crypto regimes on top of the MTL — but last in the dependency graph, since their questions reshape the entire programme. Model the M&A scenario separately: any change-of-control transaction becomes fifty parallel filings, with a decision due within 60 days of a complete application in model-law states and deemed approval on regulator inaction (MTMA § 6.01), plus FinCEN re-registration on a transfer of more than 10% of the votes. That, not due diligence, sets the closing timeline.

A stablecoin product has two deadlines, not one. For the issuer: GENIUS Act commencement no later than 18.01.2027 and the $10 billion ceiling on state-qualified status. For a distributor, wallet or exchange: 18.07.2028, after which no stablecoin from a non-permitted issuer may be offered — while nobody has lifted the MTL obligations. The wider frame of building on one's own licence versus someone else's is set out in the piece on renting a licence.