I spent the last week mapping the new payment rails. It’s the kind of deep-dive work that doesn’t feel urgent until it suddenly becomes critical. My partner bank newsletters from Q1 are full of mentions of FedNow, Plaid partnerships for A2A payments, and even new crypto on-ramps. The ground is definitely shifting.
For a founder building a neobank, “instant payments” is a feature. But which instant payment? That’s not a feature, that’s a fundamental architectural decision with lasting consequences on unit economics, user experience, and reliability. Choosing your rails is like choosing the foundation for your house. You don’t want to get it wrong.
This feels like the next chapter in the story I started exploring in What Visa Taught Me About Dental Banking. Dee Hock’s genius was creating one standardized network. Today, we’re facing a move toward a multi-rail world. It’s less of a single superhighway and more of a complex national park system with multiple routes to the same destination.
The Two Main Roads: FedNow vs. RTP
For the last few years, The Clearing House’s RTP® network was the only real-time option. Now, the Federal Reserve’s FedNow® service is live and growing fast. They’re not exactly competitors; they’re two different roads built by two different entities. Based on my research, here’s the tale of the tape:
| Feature | FedNow Service | RTP (The Clearing House) | |---|---|---| | Operator | Federal Reserve (Public) | The Clearing House (Private) | | Launch | July 2023 | 2017 | | Txn. Limit | $100k (up to $500k) | $1M | | Settlement | Direct in Fed master accounts | Pre-funded joint account at Fed |
Source: Comparisons from Cross River and industry analysis.
The key difference for a builder is the settlement model. RTP requires its member banks to pre-fund a joint account, which guarantees liquidity but can be a barrier for smaller institutions. FedNow settles directly in a bank’s own account at the Fed, a more flexible model that has led to rapid adoption by over 1,400 institutions.
Don’t Forget the Third Road: Account-to-Account (A2A)
While FedNow and RTP are the rails, Open Banking-powered A2A payments are the vehicles that can ride on top of them, often initiated by third-party providers like Plaid. This is where it gets really interesting for disrupting card payments.
The growth here is undeniable. The US real-time payments market is projected to have a 39% CAGR through 2030. Why? Because it’s cheaper. A lot cheaper.
- A FedNow transaction: ~4-5 cents.
- A typical credit card transaction: 2-3.5% of the total.
For a dental practice processing hundreds of thousands of dollars a year, that difference is not trivial. It’s thousands of dollars that go straight to their bottom line. That’s a feature that sells itself.
So What? The Founder’s Dilemma
As a founder, this multi-rail world presents a strategic choice. Do you bet on one rail? Do you integrate with both for resiliency? How do you abstract away the complexity for the end user so it just feels instant?
There is no single right answer, but here’s my current thinking:
- Interoperability is a Myth (for now): The two networks don’t talk to each other. To ensure a payment can be sent from anyone to anyone, a bank (and by extension, my future neobank) will likely need to connect to both. This adds complexity.
- A2A is the UX Play: For a dental practice paying a supplier or a patient paying for a crown, the dream is to do it right from their bank account. No card numbers, no interchange fees. This is a huge opportunity for a vertical-specific neobank to build a killer feature.
- This is Free R&D: While Dentplicity is helping practices make sense of their current cash flow, all this research into the plumbing is free R&D for CLIN. Understanding these rails today is what will allow CLIN to fundamentally change that cash flow tomorrow. 🧠
It’s a lot to track. But if you’re building a bank, you have to be obsessed with how money moves. The plumbing is everything.
Data sources: Cross River Bank Q1 2025 Newsletters, FedNow.org, TheClearingHouse.org, Mordor Intelligence, CUInsight.