On June 10, 2026, the two largest card networks simultaneously launched production infrastructure for AI agent payments. Mastercard debuted Agent Pay for Machines (AP4M), an open protocol enabling autonomous machine-to-machine transactions across 31 launch partners, with agent credentials stored ...
"Agent Pay for Machines will create the conditions for a superbloom of AI business models." — Jorn Lambert, Chief Product Officer, Mastercard
On June 10, 2026, the two largest card networks simultaneously launched production infrastructure for AI agent payments. Mastercard debuted Agent Pay for Machines (AP4M), an open protocol enabling autonomous machine-to-machine transactions across 31 launch partners, with agent credentials stored on Polygon, Solana, and Base blockchains and settlement in six regulated stablecoins. The same day, Visa announced a strategic integration with OpenAI, embedding Visa's payment rails, tokenization, and anti-fraud systems directly into ChatGPT's agentic commerce workflow.
The moves come as Cloudflare Radar data shows automated bot traffic now accounts for 57.4% of global HTTP requests, surpassing human traffic for the first time. Juniper Research projects agentic commerce spending will grow from $8 billion in 2026 to $1.5 trillion by 2030. Mastercard, Visa, and Stripe rank as the top three infrastructure providers in Juniper's 2026 Competitor Leaderboard for agentic commerce payments.
The economic question is straightforward: existing payment infrastructure was built for humans clicking buttons. AI agents transact at sub-cent values, hundreds of times per minute, 24/7. Legacy checkout flows, session-based authentication, and batch settlement cannot handle this. Both card networks are rebuilding their authorization and settlement stacks to accommodate a transaction model that does not yet generate meaningful revenue but could represent $1.5 trillion in annual flow within four years.
Two announcements landed on June 10, 2026, at different venues:
Mastercard — Agent Pay for Machines (AP4M). Launched with 31 named partners spanning crypto infrastructure (Coinbase, OKX, Aave Labs, Alchemy, Anchorage Digital, MoonPay, Solana Foundation, Polygon, RippleX), traditional payments (Stripe, Adyen, Checkout.com, Global Payments, Getnet by Santander), and technology firms (Cloudflare, Ant International). The protocol enables AI agents to transact autonomously across cards, bank accounts, and stablecoins at sub-cent values.
Visa — OpenAI Integration. Announced at the Visa Payments Forum in San Francisco, the partnership embeds Visa's network, tokenization, and fraud-detection capabilities directly into OpenAI's platform. Online retailers can accept transactions initiated by AI agents rather than humans, with commerce, authorization, and payment occurring within a single conversational flow.
The simultaneity was not coincidental. Both firms are positioning for what Visa predicts will be millions of consumers using AI agents to complete purchases by the 2026 holiday season.
AP4M operates through a four-layer system:
1. Credentialing. Each AI agent receives a verifiable identity through what Mastercard calls "Verifiable Intent" — a programmable set of spending limits and authorization rules recorded on public blockchains. Chain provides governance controls; t54 Labs delivers KYA (Know Your Agent) verification, extending traditional KYC/AML frameworks to autonomous software.
2. Permissioning. Authorization rules and spending limits are stored on Polygon, Solana, and Base, with broader blockchain access planned for later in 2026. Downstream parties can query chain verification rather than relying on centralized attestation. This is the first time a tier-one payments network has used public blockchain infrastructure as an authorization and credentialing ledger rather than solely for settlement.
3. Transacting. The system supports high-frequency, low-latency execution — transactions worth fractions of a cent, hundreds per minute, without human approval loops. A single user prompt could trigger a chain of agent actions: domain registration, hosting setup, stock image licensing, and checkout completion, each involving a separate payment, all executed autonomously.
4. Settlement. Multi-rail settlement supports traditional fiat alongside stablecoins including USDC, RLUSD, PYUSD, USDG, USDP, and SoFiUSD across eight blockchain networks — Arbitrum, Base, Ethereum, Polygon, Solana, XRPL, and others.
Raj Dhamodharan, Mastercard's Blockchain and Digital Assets leader, framed the approach in terms of institutional trust infrastructure: "We're bringing the same level of trust and ability to find the right set of agents," citing Mastercard's decades of experience solving trust problems in card and B2B payments.
Visa's approach is architecturally different. Rather than building an open protocol with on-chain credentialing, Visa is embedding its existing infrastructure into AI platforms directly.
The integration with OpenAI allows AI agents to move beyond information retrieval into transaction execution — booking services, purchasing software subscriptions, ordering supplies — all from within a chat or agent workflow. Transactions operate within user-defined parameters: spending limits, merchant category restrictions, and approval requirements.
Visa provides its network, tokenization capabilities, and security infrastructure. OpenAI gains access to Visa's anti-fraud mechanisms and merchant acceptance network. The user remains in control via permission boundaries but does not need to interact with a traditional checkout interface.
Alongside the OpenAI partnership, Visa launched two additional tools:
Agent Score, developed with New Generation, evaluates merchant websites for agentic commerce readiness — specifically whether AI agents can navigate, understand, and complete tasks on a given site.
Agentic Directory, a registry of agents and merchants that Visa has verified as legitimate participants in agentic commerce. This addresses a fundamental challenge: how does a merchant know which AI agent is authorized to spend real money on behalf of a real customer?
Visa predicts millions of consumers will use AI agents for purchases by the 2026 holiday season. Pilot programs are underway in Asia Pacific and Europe, with Latin America and Caribbean readiness expected within the next year.
The infrastructure investment makes more sense in the context of internet traffic composition. Cloudflare CEO Matthew Prince initially predicted at SXSW in March 2026 that automated bot traffic would exceed human traffic by the end of 2027. The timeline accelerated. According to Cloudflare Radar data, bots now account for 57.4% of global HTTP requests versus 42.6% from humans.
Prince posted on X in early June 2026: "Thought it would be end of 2027, then early 2027, but agentic traffic growing so fast that bots have now passed human traffic online for the first time in the Internet's history."
Before generative AI, bots accounted for roughly 20% of internet traffic. The tripling of that share in under two years reflects the rapid deployment of AI agents across enterprise and consumer applications. Gartner projects that 40% of enterprise applications will integrate task-specific AI agents by the end of 2026, up from less than 5% in 2025.
Cloudflare's Stephanie Cohen, Chief Strategy Officer, noted the infrastructure implications: "The internet was built for human interactions, but the infrastructure of the future must be built for autonomous ones." Cloudflare is an AP4M launch partner.
Juniper Research published its agentic commerce forecast in April 2026, analyzing 38,000 data points across 61 countries. The projections: $8 billion in total agentic commerce transaction value in 2026, growing to $1.5 trillion by 2030.
The research firm's 2026 Competitor Leaderboard for agentic commerce payments infrastructure ranked 14 providers:
The evaluation criteria included "specific capabilities enabling agentic flows and participation in agentic commerce protocols." The top three reflect what Juniper characterizes as decisive early-mover advantage.
Trust remains the primary barrier to deployment, according to the report. Agentic commerce will develop as "an important access channel" but "will not replace traditional eCommerce checkouts for the foreseeable future."
For scale context: the global AI agents market was estimated at $7.63 billion in 2025 and is projected to reach $182.97 billion by 2033, growing at a 49.6% CAGR, according to Grand View Research. The AI-in-payments market specifically is projected to grow from $7 billion to $93 billion by 2032.
The most structurally significant aspect of AP4M is not the stablecoin settlement — Mastercard already announced that capability on June 3, 2026 — but the use of public blockchains as an authorization and identity layer.
When an AI agent's spending permissions are recorded on Polygon, any counterparty can independently verify those permissions without querying Mastercard's centralized systems. This is a meaningful architectural choice. Centralized permission databases create bottlenecks when thousands of agents attempt simultaneous verification. On-chain credentialing distributes that verification load across the blockchain's node network.
The KYA (Know Your Agent) framework developed by t54 Labs extends regulatory compliance concepts to autonomous software. Only stablecoins with explicit regulatory licenses qualify for settlement — Circle (USDC), Ripple (RLUSD, NYDFS trust license), Paxos (USDG, USDP). This creates a permissioned layer on top of permissionless infrastructure: anyone can read agent credentials on-chain, but only licensed entities can issue the stablecoins those agents settle in.
Visa's approach bypasses blockchain for authorization entirely, relying instead on its existing tokenization infrastructure and direct API integration with OpenAI. The competitive dynamic will test whether open, blockchain-based credentialing (Mastercard) or closed, API-based integration (Visa) proves more scalable as agent counts grow from thousands to millions.
Mastercard's stablecoin buildout provides the financial plumbing for AP4M. The timeline:
The $1.8 billion BVNK acquisition closed the gap between Mastercard's card-centric infrastructure and the on-chain settlement rails required for sub-cent micropayments. Traditional card interchange fees of 1.5-3% are economically non-viable when the underlying transaction is $0.001. Stablecoin settlement with near-zero marginal cost per transaction makes micropayment business models feasible.
Visa's stablecoin settlement program, while not directly integrated into the OpenAI announcement, operates at a $7 billion annualized run rate across nine blockchains as of April 2026, growing 50% quarter-over-quarter.
The value distribution in agentic commerce differs fundamentally from traditional card payments. In a standard Visa or Mastercard transaction, interchange fees of 1.5-3% are split among the issuing bank, acquiring bank, and card network. This model assumes human-initiated, relatively high-value transactions.
AI agent transactions invert this model. Values are sub-cent. Volumes are orders of magnitude higher. Traditional per-transaction interchange is not viable. The revenue model likely shifts toward:
Neither Mastercard nor Visa has disclosed specific pricing for agent payment services. The economics remain speculative. What is clear is that the current card interchange model cannot apply to transactions worth fractions of a cent, and both networks are building infrastructure before the revenue model is fully defined.
This is consistent with the broader blockchain ecosystem pattern identified in webthreepedia's economic value research: infrastructure investment frequently precedes sustainable revenue generation, with early movers subsidizing adoption in expectation of future network effects.
The June 10 announcements represent a structural commitment by the two largest card networks to infrastructure that does not yet generate meaningful revenue. Mastercard is building an open protocol with on-chain credentialing. Visa is pursuing closed integration with the dominant AI platform. Both are making multi-billion-dollar bets — Mastercard's $1.8 billion BVNK acquisition, Visa's strategic partnership with OpenAI — on a market that Juniper sizes at $8 billion today.
The underlying thesis is that payment infrastructure built for human-initiated transactions cannot serve an economy where 57.4% of web traffic is already automated. The question is not whether AI agents will transact autonomously — Cloudflare's data confirms they already dominate internet activity — but whether the economics of sub-cent, high-frequency machine payments can sustain the kind of margins that card networks have historically earned on human commerce.
The blockchain component is worth monitoring. Mastercard's decision to use public chains for agent credentialing rather than building a proprietary identity system is an architectural choice that could become an industry standard or a competitive liability. If agent verification requires the throughput and cost profile of public blockchains, then Polygon, Solana, and Base gain a new utility function beyond DeFi and NFTs. If centralized APIs prove faster and cheaper at scale, Visa's approach wins.
At $8 billion in current transaction value, agentic commerce is a rounding error on the $25.5 trillion that Visa and Mastercard processed in 2025. At $1.5 trillion, it would represent 6% of combined volume. The infrastructure being built today will determine who captures that share.