FedNow Three Years In: What Changed, and What Late Adopters Should Do Now
Two limit increases during 2025 removed the main reason to prefer RTP, and most published comparisons still describe the old world. What actually changed since launch, what late adoption costs, and how the rail choice looks now that both networks sit at $10M.
Most comparisons of FedNow and RTP still describe the situation as it stood in 2024. That includes, until recently, ours. The rails changed materially during 2025 and the standard advice did not follow.
If you deferred a real-time payments decision on the basis of an analysis written before late 2025, the premises have moved.
What actually changed
The transaction ceiling stopped being a differentiator. FedNow launched with a $100,000 default and a $500,000 network maximum. The Federal Reserve raised the network limit to $1 million in June 2025. The Federal Reserve raised the network limit for customer credit transfers and payment returns to $10 million effective November 12, 2025, the second increase during that year, and raised liquidity management transfers from $2.5M to $10M at the same time. RTP had reached $10M earlier in 2025.
Both networks now sit at $10M, and the order of events is worth getting right because most write-ups do not. RTP went to $10M in February 2025, while FedNow was still at $500,000. For most of that year the gap was wider than it had ever been, which is exactly when the guidance that high-value B2B belongs on RTP hardened into received wisdom. FedNow closed it in November. The guidance is now wrong and it is still repeated widely, by people who formed it during the nine months when it was most obviously true.
One caveat that matters more than the headline: participants set their own limits, and most set them well below the network maximum. The number that constrains your product is what your institution and your counterparty's institution actually configured, not what the Fed permits. Ask for that specific figure.
Coverage broadened substantially. FedNow closed 2025 with nearly 1,600 participating institutions and passed 1,800 by mid-2026, against a Federal Reserve ambition of reaching the large majority of the roughly 9,000 US banks and credit unions. Adoption has skewed heavily toward smaller institutions, which is the segment RTP historically reached least well: more than 96% of participants hold under $10 billion in assets.
The question changed from whether to how. In 2023 the reasonable position for many institutions was to wait. Instant payments are now an expectation in enough contexts that waiting has a cost, and institutions joining today are not early adopters.
What the rail choice looks like now
With ceilings equalised, two dimensions remain.
Coverage. Where do your counterparties bank? FedNow reaches the long tail of community institutions better. RTP retains stronger coverage among the largest banks. For a consumer product with payees scattered across hundreds of institutions, FedNow settles more transactions in real time. For B2B with counterparties concentrated at large banks, RTP still does.
Liquidity model. FedNow settles through your Federal Reserve master account. RTP requires a prefunded position in The Clearing House joint account. That is a real difference in capital treatment and in operational process, and it is now the more consequential of the two dimensions rather than the footnote it used to be.
Everything else that used to fill comparison tables has converged. Both are ISO 20022, both are 24/7/365, both are credit-push only with no chargeback, both support request for payment.
What late adoption actually costs, and what it saves
The cost is competitive rather than technical. If your competitors settle instantly and you settle next day, that is a visible product gap in the contexts where it matters: payroll corrections, marketplace payouts, account funding, loan disbursement.
The savings are real and worth naming. Institutions connecting now benefit from vendor tooling that has matured considerably, from certification processes that providers have run many times rather than a handful, and from an operational playbook that the early cohort paid to discover. The first 100 institutions on FedNow did expensive learning. You get it free.
On balance, joining in 2026 is a better experience than joining in 2023 was, and the case for waiting further is weak.
The sequence that works
Talk to your provider before anything else. Almost no institution under a few billion in assets connects directly. Your realistic path runs through your core or digital banking provider, which means your timeline is largely their queue and certification cycle rather than your engineering capacity. Ask two things: the all-in cost including any fraud tooling that is or is not bundled, and a date rather than a capability claim.
Go receive-only first. Receiving carries no send-side liquidity exposure, and it delivers the benefit customers actually notice, which is money arriving instantly. It is not obligation-free: you still screen inbound against sanctions lists in real time and you still have a mule-account problem, since receiving is exactly where mule accounts are useful. What you avoid is the send-side fraud decision under a sub-second budget, which is the harder build. Treating send as a deliberate second phase is the difference between a six-month project and an 18-month one.
Do the operational readiness work before enabling send. A liquidity position that absorbs outflows overnight and over long weekends. Alerting that reaches a person with authority at 3am. Runbooks for stuck and returned payments. This is the part that gets underestimated, and it is not engineering work.
Then send, with limits set low. Configure well below the $10M network ceiling at launch and raise on evidence rather than on request.
The part nobody budgets for
FedNow is 24/7/365 and your operations are probably not.
Funds leave the account on a Saturday night at the same speed they leave on a Tuesday afternoon. That is a permanent change to how the institution runs, not a project task. For a 20 to 40 person IT organization it means one of three things: a genuine on-call rotation with the compensation structure that implies, a managed arrangement with your provider, or an explicit and documented decision to accept degraded response outside business hours.
All three are defensible. Arriving at the third by accident, and discovering it during the first incident, is not.
Budget for the operational model as well as the integration. The integration is finite work. The operating change is permanent.
For the full rail-by-rail comparison see FedNow vs RTP, and for what joining involves as an institution see the FedNow onboarding guide. For engagement details, RTP and FedNow integration or book a call.