In my last post I described how stablecoin infrastructure is reshaping the nature of an FX swap. The post goes through what is called the stablecoin sandwich: USDC and USDT sit in the middle as a de facto global settlement layer while two slices of traditional infrastructure hold it together at the edges, converting local currency into stablecoins on the way in and back out on the way out.
The promise of that architecture is that it kills the frictions of correspondent banking but, as usual, reality is subtler. In its current iteration – without local stablecoins – the frictions are simply pushed to the edges of the sandwich, where the on and off-ramps meet the local banking system. In particular, I see a few broad areas of frictions: licensing in each jurisdiction, KYC/KYB, AML (Customer acceptance policies, Travel Rule, transaction monitoring), integration with local banking rails, FX risk, and local liquidity.
This post is about one: rails integration. Every ramp, sooner or later, hits the same wall: to move real euros or dollars, it has to connect to a domestic payment system and it can do that either directly, as a participant in its own right, or indirectly, by riding an institution that already is one. That single choice determines its cost base, its settlement speed, and how much of its business depends on somebody else’s balance sheet.
The goal of this post is to investigate the trade offs of direct vs indirect integration – the costs and the benefits – and to determine in which scenarios one choice makes more sense than the other.
Direct Participation in Key Rails
Foreign exchange is overwhelmingly a two-currency story. According to the 2025 BIS Triennial Survey, global FX turnover reached $9.51 trillion/day, up from $7.5 trillion in 2022 with US dollar sitting on one leg of 89.1% of all trades and euro on 28.5%. Together, roughly 96% of all global FX turnover has either the dollar or the euro on at least one leg (the figures exceed 100% because every trade has two legs, but stripping out the euro-dollar pair itself leaves ~96%).
If almost the entire market touches one of two currencies, then the infrastructure that matters is the domestic plumbing behind those two currencies. That is why I focused on the euro and dollar rails: six systems, grouped into three functions that mirror each other: settlement (T2/Fedwire), instant (TIPS/FedNow), and clearing (STEP2/FedACH).
For each one I looked at the same things: what the rail is, the criteria to become a direct participant, the cost of doing so and the benefits it unlocks, paying particular attention to specific parameters that do most of the discriminating work (e.g. reach, intraday credit, settlement time, finality type).
T2 (Europe – settlement)
T2 is the Eurosystem’s real-time gross settlement system: every euro payment settles individually in central bank money, and that settlement is the moment of legal finality under the Settlement Finality Directive. Direct access was reserved for credit institutions until 6 October 2025, when authorised EEA payment and e-money institutions became able to hold an account directly.
The cost of going direct is not the fee, the real cost is operational burden and liquidity.
Operationally, going direct means becoming a piece of critical market infrastructure in your own right. You connect through the Eurosystem’s single gateway (ESMIG), pass both technical and operational certification, and stand up real-time monitoring able to flag a failed or delayed settlement the moment it happens. None of this is trivial: it carries continuous compliance and reporting obligations, and a long onboarding timeline (months rather than weeks).
In terms of liquidity: to settle payments you have to keep euros sitting in the account in advance, and those balances barely earn anything: every euro you are forced to pre-fund is yield given up. But there are two different regimes depending on whether you are a bank or an EMI: a bank can post collateral for Eurosystem intraday credit and run a thin buffer, while a payment or e-money institution cannot and must fully pre-fund every outgoing payment, giving up the yield on all of it.
The benefit is access to the strongest position on the euro leg: settlement in seconds, statutory finality, a direct line to every T2 participant and no dependence on a sponsor.
TIPS (Europe – instant)
TIPS settles SCT Inst payments individually, in central bank money, in about ten seconds, 24/7/365. The same October 2025 opening applies: authorised payment and e-money institutions can now hold a TIPS account directly. For an FX or on-ramp operator whose real problem is settlement lag, this is – in my opinion – the most relevant euro rail of the three.
As with T2, the real cost is again operational burden and liquidity, and both are harsher than T2’s because the rail never closes.
Operationally, 24/7/365 means exactly that: no maintenance window, no overnight downtime, and real-time reachability you have to sustain around the clock. Failed-payment monitoring and exception handling stop being business-hours tasks and become continuous ones.
In terms of liquidity, the always-on design bites twice. Funds sit in the account through the night and the weekend with no overnight sweep-back to somewhere they could earn, so the yield you give up is even larger than on T2. And the same two regimes apply: a bank can lean on intraday credit against collateral, while a payment or e-money institution must keep the account fully pre-funded at all times to cover outflows that can arrive at any hour.
The benefit is the best combination available in euro: instant settlement and statutory finality in central bank money, all day every day, reaching every SCT-Inst-enabled PSP in SEPA now that instant reachability is mandatory for EU PSPs. The old €100,000 amount cap is gone, removed by the Instant Payments Regulation, so the rail is no longer limited to small-value flows.
STEP2 (Europe – clearing)
STEP2 is EBA Clearing’s pan-European automated clearing house: it batches and nets SEPA transfers, but the net positions still settle on T2.
A direct participant must be a bank with access to a T2 account – its own, or one sponsored by a liquidity provider. That is the crucial point: direct membership at the clearing layer still requires a settlement leg underneath it, so “direct on STEP2” can still mean “sponsored on T2.”
On cost, per item it is the cheapest euro rail, but direct participants also contribute to the ASI guarantee fund, a recurring capital cost indirect participants avoid. The benefit is reach: a direct push to roughly 4,800 PSPs across SEPA with no correspondent chain.
Fundamentally STEP2 is a reach rail, not a speed rail. It enables broader reach but doesn’t speed up settlement which still relies on T2.
Fedwire (US – settlement)
Fedwire is the dollar analogue of T2: the Federal Reserve’s real-time gross settlement system, settling each payment individually in central bank money.
Direct access requires a Federal Reserve master account, in practice reserved for federally-insured depository institutions. However a May 2026 executive order and a proposed “Payment Account” would open limited direct access to a broader tier of institutions – basically the US trying to do what the EU did in October 2025.
As in Europe, the cost of going direct is not the fee: the real cost is operational burden and liquidity.
Operationally, the real gatekeeper is the master account itself, and its complex approval process. In particular, FedLine connectivity and the Fed’s security and resiliency requirements are extremely severe, given the continuous compliance obligations that come with being plugged straight into the settlement layer.
Liquidity-wise, the same bank-versus-non-bank asymmetry is emerging. A master-account bank can run a thin buffer by drawing daylight overdraft – the Fed’s intraday credit – while the proposed non-bank Payment Account explicitly excludes intraday credit and caps closing balances at $1bn, so a non-bank must fully pre-fund its outflows.
The benefit is similar to T2: settlement in seconds, “statutory finality” (under Regulation J, not a single statute like the EU’s SFD – see more here), a direct connection to every US bank and no dependence on a sponsor.
FedNow (US – instant)
FedNow is the dollar analogue of TIPS: real-time gross settlement in central bank money, 24/7/365, on the same master-account terms as Fedwire. The cost and benefit profile mirrors TIPS: the same always-on operational burden, the same round-the-clock pre-funding, the same bank-versus-non-bank intraday-credit asymmetry.
Two differences are worth flagging.
The first is the value cap: FedNow will not process a payment above a set ceiling – raised from $1m to $10m in November 2025 – so large payments still have to fall back to Fedwire. Europe scrapped the €100,000 limit entirely in October 2025, leaving TIPS with no upper bound at all. The second is reach: FedNow’s participant roster is still growing, so not every US bank can yet be reached instantly, whereas in the EU instant reachability is now mandatory.
FedACH (US – clearing)
US ACH is the dollar counterpart to STEP2, run by two operators – FedACH (the Fed) and EPN (The Clearing House) – on deferred net settlement that ultimately clears through Fedwire.
A direct participant is an ODFI/RDFI, i.e. a depository institution. This is the sharpest EU-US gap in the whole comparison: there is no non-bank direct tier at all. A non-bank reaches ACH only as a Third-Party Sender, riding an ODFI’s routing and settlement number, so here “direct access” simply isn’t possible unless you are a bank.
Per entry it is the cheapest dollar rail, but the ODFI carries liability for every transaction, so a sponsoring bank passes that risk back to the Third-Party Sender through reserves and oversight. The benefit is the broadest low-value USD reach, near-universal across US institutions. The cost is finality: ACH has no statutory finality and entries stay returnable for up to two business days.
Conclusions
Based on the analysis of the different rails above, a few patterns emerge:
- The real cost of direct access is liquidity and operational burden
- Direct access isn’t the same as levelling the field: a bank is still significantly better placed than an EMI, because it can lean on intraday credit rather than fully pre-fund
- Clearing systems really help in terms of reach, but not in terms of settlement time.
Based on those patterns, when a fintech has to decide which rails to focus on to optimise its operations, I think two key decisions must be made.
The first: which is the most critical layer for your flow?
- Large or high-stake payments → clearly RTGS (T2/Fedwire) where individual settlement and statutory finality matter
- Time-critical payments → instant rails (TIPS/FedNow): finality in seconds, 24/7
- High-volume, low-value, non-urgent payments → clearing rails (STEP2/FedACH): cheapest per item, and deferred netting is perfectly acceptable when nothing is time-sensitive.
The second decision is whether to join that layer directly or ride someone else. This one isn’t set by the payment type, it’s set by the volume and by the type of organisation.
Because the cost of direct access has a substantial fixed component (the operational burden), it only amortises above a certain throughput. Below that, indirect is simply cheaper, whatever the layer. This same fixed cost is why specialist infrastructure providers exist at all: they pay it once and rent the access out to everyone sitting below the threshold.
But even for an institution with very high volume, going direct doesn’t guarantee a lower cost of liquidity: for a bank it does, because direct access brings intraday credit; but for an EMI, it doesn’t.
In general, a fintech doesn’t move from indirect to direct as it matures: it should make the call flow by flow, and rationally evaluate what makes more sense based on type of flows, volume and type of institution.
One caveat, though. All of this assumed the hard part is the euro or dollar leg, but usually the euro and the dollar are the easy currencies to reach. In emerging-market corridors the bottleneck is the local rail, and there the choice of rails and type of access matters even more.
Resources
- BIS Triennial Central Bank Survey 2025 – link
- BIS Triennial Central Bank Survey 2022 – link
- The Long Road to €1 onchain – link
- The Long Road to $1 onchain – link
- T2 – Service Registration and onboarding guide
- TIPS – ECB TIPS facts page
- TIPS – Instant Payments Regulation (EU) 2024/886
- STEP2 – EBA Clearing STEP2-T admission criteria
- Fedwire – Operating access – Circular 6
- FedNow – Operating acces – Circular 8
- FedNow – $10m transaction-limit increase
- FedACH – Nacha Operating Rules and Third Parties in the ACH Network
