One year ago, machine-to-machine payments were just a concept. Today, four competing payment architectures are live, backed by Coinbase, Stripe, Google, Visa, and American Express. AI Agents have settled over $730 million across 176 million transactions, while traditional incumbents have invested more than $8 billion in acquisitions to capture this emerging payment stack.

This report is produced in collaboration between Keyrock, Coinbase, and Tempo, examining how this payment stack is being assembled, whether its economic model holds, and what obstacles remain.

Protocols Are Not Competing—They Are Stacking

In September 2024, if you wanted to enable an AI Agent to pay, there was essentially only one unsafe option. Just 12 months later, four distinct architectures now exist, each backed by some of the largest tech companies in the world.

Coinbase built x402, a crypto-native protocol that transforms stablecoin wallets into universal API keys. Stripe and Tempo launched MPP, a payment method-agnostic standard that processes cards, cryptocurrencies, and the Lightning Network through a single HTTP flow. Google assembled AP2, an authorization layer enabling users to delegate payment permissions to Agents via cryptographic credentials. Visa expanded its existing card rails by introducing AI-ready tokenized credentials.

What most coverage overlooks is that these four solutions are not purely competitive. While there is overlap at the protocol layer, the more significant dynamic is their integration into a layered payment stack. The right question is not “Which protocol will win?” but rather “Which companies control the most layers—and thus capture the most value?”

The $0.30 Wall

Among the 176 million x402 transactions to date, the median transaction size lies between $0.01 and $0.10, with 76% of activity falling below the $0.30 threshold of traditional card processing fees. This number almost entirely explains why legacy payment rails cannot serve this market. A fixed ~$0.30 per-transaction fee makes microtransactions unprofitable. An Agent paying just 3 cents for a weather API call cannot route through Visa.

Layer 2 stablecoin settlement costs just $0.0001. For Agents, this means blockchain rails are not optional—they are essential.

Single Stablecoin Dominance

Of those 176 million transactions, 98.6% were settled in USDC. Stablecoins have effectively won the settlement layer for machine commerce—they are the only tools capable of handling microtransactions without collapsing their economic model.

This concentration serves as both validation and vulnerability. It validates Circle’s role as the default settlement asset, but also implies that the entire Agent payment ecosystem depends on a single stablecoin issuer’s reserve management, regulatory standing, and technical infrastructure. No one in the industry openly discusses this risk. We believe they should.

The Vertical Integration Race

Coinbase and Stripe each cover five of the six layers in the emerging payment stack. Coinbase controls the settlement layer (Base), wallets (Agentic Wallets), routing (internal infrastructure), payment protocol (x402), and governance (as a partner in AP2). Stripe mirrors this with Tempo (settlement), Privy (wallet), Bridge (routing, acquired for $1.1B), MPP (protocol), and its compliance infrastructure.

Over the past 12 months, traditional incumbents have invested over $8 billion in acquisitions to close gaps in their payment stack coverage. Capital One acquired Brex for $5.15B, Mastercard spent $1.8B to buy BVNK, and Stripe acquired Bridge. These are all infrastructure consolidation moves from firms treating machine payments as a core business imperative requiring vertical expansion.

From Bot Activity to Agent Commerce

The machine economy has arrived—but it hasn’t yet begun commercializing. Yet the signals are clear: AI Agents account for 37% of all Safe transactions on Gnosis Chain, peaking above 75%. Coinbase has deployed tens of thousands of Agents with built-in guardrails. Over 104,000 Agents are registered across 15 or more directories and registries.

The shift from extractive bot activity to productive Agent commerce is underway. The payment infrastructure studied in this report is precisely what enables this transformation.

Regulation as Constraint

MiCA, the GENIUS Act, and the EU AI Act will all reach enforcement stages within weeks of each other in mid-2026. None address autonomous machine-to-machine transactions. This is not a future problem—it is a present one, unfolding in real time against actual capital commitments.

What Comes Next

The market is moving toward greater Agent autonomy, but we believe the pace will not be set by technology—technology is already largely ready. Instead, the pace will be determined by the trust infrastructure that makes everything safe. The vision of fully permissionless systems is theoretically appealing, but it assumes an AI reliability level that does not yet exist. Before Agents stop hallucinating, they should not be granted unsupervised access to user funds.

We find the bottom-up argument the most compelling framework for what comes next. Crypto rails have already won the micro-payment game by default. As transaction volume grows and trust infrastructure matures, larger-value transactions will increasingly migrate on-chain. The question is no longer whether machine-native payments can scale—but how quickly the trust layer can catch up to the settlement layer.

This article is a summary of core findings. The full report dives deeper into data, including protocol architecture analysis, interviews with Coinbase and Tempo, transaction economics modeling, and the regulatory landscape.

Authors: Ben Harvey, Translated by DeepTide TechFlow