Google, Circle, Stripe, and others are all spending money on AI; what will be the joys and sorrows of these payment giants in Q1 2026?

Text | Kaori

Edited by | Sleepy.md

The first three months of 2026 were very fulfilling for players in the payment sector.

On January 11, Google released the Universal Commerce Protocol (UCP) at the Retail Federation of America's annual meeting, attempting to define a common language for AI Agent commerce. That same week, Revolut announced it was among the first EU payment methods to be compatible with Google AP2, PayPal announced its acquisition of merchant directory synchronization company Cymbio, and Mastercard launched its Agent Suite.

In February, Coinbase officially launched Agentic Wallets, allowing AI agents to have their own wallets and autonomously spend, earn, and trade crypto assets. The x402 protocol is deeply integrated with the Google ecosystem and has already processed over 50 million transactions.

March saw a surge of activity. Circle launched Nanopayments, Ramp introduced Agent Cards, Mastercard announced its acquisition of stablecoin infrastructure company BVNK for up to $1.8 billion, and Stripe and Paradigm-incubated Tempo Chain launched their mainnets and simultaneously released the Machine Payments Protocol (MPP).

Within three months, more than a dozen major moves have occurred, some positive and some negative. These events may seem scattered, but they all point to the same structural change: when the cost of machine-to-machine transactions is driven close to zero, the real enemy of payment giants is no longer each other, but zero cost itself.

Key Events Review

In the era of zero cost, there is no winner-takes-all.

Six months ago, we were still debating who should legislate for AI agents. Stripe's ACP, Google's AP2, and Mastercard's Agent Pay were three separate paths, each vying for the right to define the same proposition.

The war is now effectively over, not because one side won, but because everyone realized that a winner-takes-all scenario is not going to happen.

Google's newly launched UCP at the beginning of the year is compatible with all major standards and handles commercial transactions within the search and Gemini ecosystem. Stripe's MPP, launched alongside Tempo, also supports integration with Mastercard and Visa systems, handling self-service payments between machines. Mastercard's Agent Pay handles auditable authorization for high-value transactions.

What started as a struggle for territory has now become a struggle for territory. The tentatively established framework at the agreement level means that the decisive competition has shifted elsewhere.

Here's some data released by Enterprise Onchain: In the past nine months, AI Agents completed 140 million payments, totaling $43 million, with 98.6% using USDC. The average transaction was $0.31, and there are now over 400,000 AI Agents with purchasing power.

Let's break down the meaning of these numbers.

First, transactions were conducted autonomously by machines. 140 million payments were processed without human intervention, bank approval, or credit card verification. Processes that previously required human signatures, reconciliation, and settlement between codes and protocols are now completed autonomously by machines.

Second, the individual transaction amounts are extremely small. An average transaction value of $0.31 means that the vast majority of these payments are micro-payments for API calls, computing power purchases, and data access. Under traditional payment systems, such transactions would be impossible; the minimum transaction fee on any card network would exceed the value of the transaction itself.

Third, the cost approaches zero. With the x402 protocol, payments are directly embedded in HTTP requests. Circle Nanopayments reduces the gas fee for developers' single transactions to zero by aggregating tens of thousands of small payments off-chain and periodically packaging them for on-chain settlement. The on-chain settlement cost is borne by Circle at the batch settlement layer.

Transactions between machines, without checkout pages, payment gateways, or intermediaries, are the source of the problem.

Of course, zero cost currently only applies to the specific scenario of inter-machine micropayments. Stablecoins are not truly free; on the Ethereum mainnet, the gas fee for a small stablecoin transaction can reach over 20% of the transaction amount. Stripe created Tempo precisely to address this issue.

At the consumer payment level, card networks still have advantages that stablecoins cannot replicate: unified consumer protection, consistent user experience, and the flexible routing capabilities of cards as an abstraction layer at the underlying level.

However, this doesn't change the fundamental concern. In scenarios of high-frequency micro-payments between machines, zero cost is already a reality, and this gap is rapidly widening. Deloitte predicts that the global agent market will reach $45 billion by 2030. This is a completely new world of transactions, tearing a huge gap at the edge of traditional payments.

The giants' response: From collecting tolls to building bridges

Faced with the threat of zero costs, traditional payment giants have adopted different strategies, but they share a common underlying logic: since they cannot collect fees in the scenario of inter-machine micro-payments, they should control the bridge between the old and new systems and charge fees there.

Visa's strategy is absorption rather than confrontation. USDC settlement has officially launched in the United States, and crypto-friendly banks such as Cross River Bank and Lead Bank have begun using it. Visa Direct supports stablecoin pre-deposits and direct payments.

In other words, you can use stablecoins, but please use my network. Visa also participated in the development of the MPP, extending the protocol to card payment scenarios—a typical example of joining in if you can't beat them.

Mastercard's $1.8 billion acquisition of BVNK was essentially buying a bridge between fiat currency and stablecoins. BVNK supports fiat-to-stablecoin conversion across all major blockchain networks in over 130 countries, which is the most critical infrastructure for the AI ​​Agent payment era.

Mastercard's Chief Product Officer, Jorn Lambert, responded directly to the claim that stablecoins threaten the card business: the card business is fine, and the acquisition is for expansion into new areas such as remittances. However, the deeper logic is that as stablecoin trading volume grows rapidly, controlling the clearing bridge between fiat currency and stablecoins means controlling the lifeline of value flow.

Stripe has the biggest ambitions. It owns its own blockchain Tempo, its own protocol MPP, and Open Issuance, a platform that allows companies to issue their own stablecoins and share reserve yields—this is the ultimate in vertical integration.

The combination of Tempo, MPP, and Open Issuance means that Stripe is no longer just a payment processor; it is becoming an infrastructure operator in the era of AI Agent payments.

PayPal took a different approach. Its acquisition of Cymbio wasn't about controlling the payment pipeline, but rather controlling the distribution of merchant catalogs. Cymbio's core capability is its Store Sync technology, which allows merchants to synchronize their product catalogs across multiple AI shopping surfaces with a single click. This means that small and medium-sized merchants no longer need to adapt to each AI platform individually.

When AI agents replace humans in discovering products, whether merchants' product catalogs can be seen by AI becomes a matter of life and death. PayPal is betting that in the era of Agent Commerce, being discovered by an agent is valuable in itself.

An interesting intermediate state is Ramp Agent Cards, which issue virtual cards to AI agents, still using the Visa card network, but dynamically authorizes each transaction and does not expose real card information. Essentially, this turns corporate spending cards into agent wallets.

Whether this is a continuation of traditional payments or a stopgap measure during the transition period remains unclear. If the mainstream form of machine-to-machine transactions ultimately moves towards a stablecoin-native approach, then Agent Cards may be the last chance for traditional card networks to be needed in the new era.

How to make money in the new era?

There's a question that hasn't been answered directly. On the zero-cost track, the transactions themselves don't incur fees. So, who profits?

Circle's Nanopayments earns revenue from infrastructure service fees, Stripe's Open Issuance earns revenue from reserve funds, and Mastercard, after acquiring BVNK, earns revenue from converting between fiat currency and stablecoins.

All three charging methods share a common characteristic: the charging point has shifted from the transaction itself to the conditions under which the transaction can occur. Essentially, they are closer to infrastructure rent than transaction taxes.

This represents a fundamental shift in business models. For the past fifty years, the moat of payment networks has been the network effect. The more merchants there are, the more consumers are willing to use it; the more consumers there are, the more merchants need to connect. This flywheel profits from the commission rights brought by scale.

In the world of machine-to-machine transactions, this flywheel has failed. Machines only need a stable, programmable, and low-cost settlement layer; whoever can provide that becomes the new infrastructure provider.

The survival of payment giants is not a major issue. The real question is: in an industry that relies on commissions to maintain its power, where does that power go once the commissions lose their meaning?