Across this module we built up the machinery of card economics: the four-party model, interchange, scheme fees, the merchant's effective rate, routing. Every one of those line items is a cost to someone, which makes each a target. This lesson is about who is aiming at them, what actually lands, and what the networks have quietly co-opted.
The framing that matters: "pressure on the model" is three separate stories wearing one headline. Account-to-account rails attack the rails themselves. Network bypass attacks interchange and acceptance habits. Stablecoin settlement, despite the crypto framing, mostly attacks the back-end settlement leg and so far reinforces the networks rather than displacing them. Conflate them and you will misprice the threat.
Account-to-account rails: a real second network
An account-to-account (A2A) payment moves funds directly between two bank accounts over a clearing rail, with no card and no four-party interchange in the path. The instant-payment versions of these rails are now large. Brazil's Pix is used by the large majority of adults and, in 2025, overtook credit cards as the most-used e-commerce payment method by purchase value. India's UPI processed over 220 billion transactions in 2025. The US has FedNow and the privately run RTP network, both still early on the adoption curve relative to those two.
The cost difference is the whole pitch. A FedNow transfer costs the bank a few cents. US credit card processing averaged roughly 2.35 percent in 2025, on unregulated interchange (see Module 3) plus scheme fees (Module 4). For a merchant doing $10 million a year, moving even part of that volume to an A2A rail at a sub-1 percent all-in cost is six figures of saved acceptance cost.
Why the cost gap has not collapsed card volume
Three things the card model bundles do not come free on a raw bank transfer. First, consumer protection: a card chargeback (Module 5) lets a buyer reverse a disputed charge, while a standard A2A push payment is irrevocable once sent, so redress depends on local banking law and the merchant's own terms. Second, rewards: US card rewards are funded out of interchange (Module 7), and a cheaper rail with no rewards is a worse deal for the cardholder even when it is a better deal for the merchant. Third, recurring authorization and credit. A single Pix push does not, by itself, handle a subscription or extend the buyer credit.
The market is closing two of those gaps. Brazil launched Pix Automatico in 2025 to authorize recurring A2A payments, the use case cards previously owned. In the UK, the open-banking push toward variable recurring payments (VRP) targets the same recurring-debit territory. These are the developments to watch, because recurring billing is where A2A economics bite hardest and where the irrevocability problem is most manageable.
Network bypass: real at the rail, contested at the front end
"Bypass" is precise: a QR-code or pay-by-bank checkout that settles over FedNow, RTP, or ACH carries no Visa or Mastercard interchange, no scheme assessment, no network in the loop. At the rail level the bypass is total. The contested question is the front end, where the buyer chooses, and that is exactly where the networks are defending.
Watch what the incumbents are buying and building rather than what they say. Visa acquired the open-banking firm Tink and launched Visa A2A, a framework that puts pay-by-bank and VRP under a common rulebook, API standard, dispute-management process, and commercial model. In June 2025 Visa added card-style buyer protections onto pay-by-bank transfers in the UK. Read that move plainly: the networks would rather operate the bypass rail, and charge for governance and disputes on it, than let an unbranded transfer route around them entirely. The interchange line may shrink while the network's role does not.
A worked comparison
Take a $200 online purchase from a recurring biller, US context.
- Card: interchange plus scheme fees plus acquirer markup land near 2.4 percent, so roughly $4.80 of acceptance cost, with chargeback rights and rewards attached.
- Raw A2A push over RTP: a few cents of rail cost, no interchange, no scheme fee, but no built-in chargeback and no rewards. Saving to the merchant is real and large.
- A2A under a network-operated framework (Visa A2A style): rail cost stays low, the network reintroduces a dispute and protection layer and charges for it. Cheaper than the card, but the network keeps a seat and a fee.
The merchant's saving is largest in the middle row and smallest in the third. The bottom row is the one the networks are working to make the default.
Stablecoin settlement: mostly back-end, mostly co-opted
This is where careful reading matters most, because the crypto framing oversells the disruption. Most stablecoin activity on card rails is on the settlement leg, the back-end money movement between issuer, network, and acquirer, not on the consumer's tap at checkout.
Visa launched US domestic USDC settlement in December 2025 after a pilot, letting issuer and acquirer partners settle with Visa in Circle's USDC, with early participants settling over the Solana blockchain. By April 2026 Visa reported the settlement program at a roughly $7 billion annualized run rate and supporting nine blockchains, having added chains including Base, Polygon, Canton, Arc, and Tempo. Visa is a design partner on Circle's Arc chain and a validator on Tempo and Canton. Mastercard, since April 2025, has let merchants opt into USDC settlement and built out wallet, issuance, and settlement pieces, and as of June 2026 supports intraday and weekend settlement in regulated stablecoins.
Notice what this is and is not. The cardholder still presents a Visa or Mastercard card. The four-party model, the network, the interchange, all still stand. What changes is that the settlement leg can clear faster, including on weekends, in a stablecoin instead of a same-day or next-day bank transfer. That is an efficiency upgrade to the existing model, captured by the incumbents, not a route around them.
The genuine displacement risk is one layer out
The settlement story leaves the model intact. The thing that could actually bypass it is a stablecoin payment that never touches a card at all: a wallet-to-wallet or agent-to-agent transfer at the point of purchase. Tempo, which Visa is now a validator on, is explicitly designed for agentic and machine-to-machine payments. If buying agents settle directly in stablecoins for a meaningful share of commerce, that is true network bypass. It is also early, thin on consumer protection, and uncertain on regulation. Treat it as a watch item, not a current threat, and do not confuse it with the settlement-leg pilots that are live today.
Takeaway
Sort every "death of cards" claim into one of three buckets before you react. A2A rails are a real second network and already dominant in some markets, but cards survive on protection, rewards, and credit, and the fight has moved to recurring payments. Network bypass is total at the rail and contested at the front end, where the incumbents are buying and standardizing their way back into the flow. Stablecoin settlement on card rails is, for now, a back-end efficiency the networks have absorbed, not a bypass. The pressure is real. Most of it, so far, the model has bent to fit rather than broken under.