The Future of Payments Infrastructure

Where the rails are heading, and what the next generation of payments architecture looks like.

Over the past five modules, we have mapped the architecture of card payments as it exists today: the transaction lifecycle, the acquiring stack, card networks, chargebacks, and security layers. This architecture has been remarkably durable. The four-party model is decades old. ISO 8583 dates from the 1980s. Interchange has been a feature of the system since the beginning.

But the infrastructure is changing. Real-time payment networks are live in over 70 countries. Open banking is turning bank accounts into payment rails. Stablecoins are being tested for settlement. And AI agents are beginning to initiate payments autonomously, without a human in the loop.

This module examines the five shifts that are reshaping payments infrastructure. Not predictions about what might happen. An assessment of what is already being built, where the investment is flowing, and which problems remain unsolved.

The question is not whether payments infrastructure will change. It is whether the existing rails will adapt fast enough to absorb these shifts, or whether new rails will emerge alongside them. The answer is probably both.

Shift 1: Real-Time Payments

Real-time payment (RTP) networks settle transactions in seconds, 24/7, 365 days a year. They bypass the card networks entirely. Money moves directly from the payer's bank account to the payee's bank account through a national or regional clearing system.

The major real-time payment systems include FedNow (US, launched 2023), Faster Payments (UK, operational since 2008), UPI (India, processing over 10 billion transactions per month), SEPA Instant (EU), and PIX (Brazil). Each operates differently, but the core proposition is the same: instant, low-cost, account-to-account transfers.

The threat to card networks is real but nuanced. RTP networks are excellent for person-to-person transfers, bill payments, and payroll disbursements. They are less mature for e-commerce checkout, where the consumer protection mechanisms (chargebacks, fraud liability shifts) that cards provide remain valuable. The irrevocable nature of real-time payments, once the money is sent it cannot be pulled back, is a feature for merchants but a risk for consumers.

India's UPI is the clearest example of what full-scale RTP adoption looks like. Card payments account for a small fraction of Indian digital commerce. UPI dominates because it is free for consumers, nearly free for merchants, and integrated into every major app. Whether this model translates to markets with deeply entrenched card networks is the open question.

Shift 2: Open Banking and Account-to-Account Payments

Open banking gives regulated third parties access to bank account data and the ability to initiate payments directly from a customer's bank account. It is enabled by regulation (PSD2 in Europe, Consumer Data Right in Australia) and increasingly by market forces (banks opening APIs voluntarily in the US).

For payments, open banking enables a new flow: the customer authorises a payment directly from their bank account to the merchant's bank account, bypassing the card networks entirely. No interchange. No scheme fees. No chargeback mechanism.

Open banking payment initiation is already live in the UK and EU through Payment Initiation Service Providers (PISPs). Companies like TrueLayer, GoCardless, Volt, and Token.io provide the infrastructure. The card networks are not standing still: Visa acquired Tink and Mastercard acquired Finicity specifically to participate in open banking flows.

The adoption challenge is consumer experience. Card payments are effortless: tap, done. Open banking payments require the customer to authenticate with their bank, which currently involves app switching and additional friction. Until that experience is as smooth as a card tap, open banking will complement card payments rather than replace them for everyday transactions.

Shift 3: Embedded Finance

Embedded finance is the integration of financial services, including payments, lending, insurance, and accounts, into non-financial software platforms. Instead of the merchant or consumer interacting with a bank or payment provider directly, the financial service is invisible, woven into the software they already use.

We touched on this in Module 2 when we discussed ISVs embedding payments into their software. Embedded finance extends the concept further. A logistics platform that offers instant supplier payments. A marketplace that provides seller lending based on transaction history. A SaaS platform that holds customer deposits in an embedded account.

The payments infrastructure implications are significant. As financial services become features of software platforms, the traditional distribution model (bank sells to merchant) gives way to a platform model (software platform bundles financial services and controls the relationship). The companies providing the infrastructure behind the scenes, the Banking-as-a-Service providers and payment facilitators, become critical but invisible plumbing.

Shift 4: Stablecoins and On-Chain Settlement

Stablecoins are cryptocurrencies pegged to a fiat currency (typically the US dollar). Tether (USDT) and Circle's USDC are the largest, with a combined market capitalisation exceeding $200 billion. They are increasingly being used for real payment flows, not just crypto trading.

For payments infrastructure, stablecoins offer something genuinely new: programmable, 24/7 settlement on rails that are not owned by any single institution. A cross-border payment that would take two days and cost $25 through correspondent banking can settle in minutes on-chain for a fraction of the cost.

Mastercard's $1.8 billion acquisition of stablecoin platform BVNK in March 2026 is the clearest signal that the card networks view on-chain rails as complementary to their existing infrastructure, not a threat to be ignored. Visa has been running stablecoin settlement pilots since 2021 and now supports USDC settlement for select issuers.

The challenges are regulatory (stablecoin frameworks are still being finalised in most jurisdictions), user experience (bridging between fiat and stablecoin is still friction-heavy for consumers), and consumer protection (no chargeback mechanism, no deposit insurance). But for B2B payments, cross-border settlement, and treasury operations, stablecoins are already competitive with traditional rails.

The card networks' stablecoin investments are a hedge. If on-chain rails displace card rails for certain flows (particularly cross-border), the networks want to be the ones operating the infrastructure. If they do not, they want to be connected to whoever does.

Shift 5: Agentic Commerce

This is the most speculative shift, but it is already being tested in production. Agentic commerce is the use of AI agents to discover, evaluate, negotiate, and purchase goods and services autonomously on behalf of a human principal. The agent handles the entire transaction without the human being present at the point of sale.

Both Visa and Mastercard are actively running trials. Visa is testing agent-initiated payments with Santander in Latin America and has recruited banks across the UK and Europe for issuer trials. Mastercard is pursuing similar pilots. The core infrastructure question is straightforward: how does an AI agent authenticate, authorise, and pay?

The current card payment architecture was designed for a human at a checkout. Authentication assumes a person who can enter a password, approve a push notification, or provide a biometric. Card-present transactions assume someone physically holding a card. None of these assumptions hold when the buyer is software.

The infrastructure being built for agentic commerce includes:

• Agent identity frameworks. How do you verify that an AI agent is authorised to act on behalf of a specific human, with defined spending limits and merchant category restrictions?
• Delegated authentication. Pre-authorisation schemes where the human sets rules (spend up to $200 at these merchant categories) and the agent operates within those boundaries without requiring per-transaction approval.
• Machine-readable commerce protocols. Standards like Coinbase's x402 (HTTP 402 payment-required protocol) that allow agents to discover prices and make payments natively within web interactions.
• Settlement for autonomous transactions. Whether card rails, stablecoin rails, or new purpose-built rails are the right infrastructure for transactions initiated by software that may operate at speeds and volumes incompatible with current batch settlement.

Agentic commerce does not need a new payments system. It needs the existing system to support a new type of buyer. The companies that solve agent identity, delegated authentication, and machine-readable pricing will define the next layer of payments infrastructure. Whether those companies are the card networks, the tech platforms, or someone new is not yet determined.

What Persists and What Changes

With five structural shifts underway, it is tempting to predict the end of card payments. That prediction would be wrong. Here is a more grounded assessment of what persists and what changes.

Glossary