OpenAI’s native checkout collapsed in March after Walmart reported conversion rates three times worse than standard e-commerce flows, and fewer than 30 Shopify merchants ever went live. The failure didn’t slow agentic commerce investment, it just showed that the problem was never AI capability but the lack of infrastructure.

Adyen, Stripe, Visa, and Mastercard are now racing to build the plumbing that OpenAI's experiment exposed as missing. The market they're targeting is large: eMarketer projects $144 billion in U.S. e-commerce originating from AI platforms by 2030. The companies that solve the infrastructure gap will capture the most durable toll position in that stack.

The Infrastructure Gap Investors Are Betting On

Adyen’s research team spent several months interviewing more than 200 global enterprise merchants and identified five structural constraints blocking agentic commerce at scale: protocol fragmentation across AI commerce interfaces, product data systems machines cannot reliably query, enterprise checkout stacks built for linear human flows, fraud frameworks calibrated for human-initiated transactions, and merchant onboarding models that do not scale across platforms.

The VC thesis emerging around this gap treats agentic payments infrastructure the way early cloud investors treated API abstraction layers: whoever becomes the universal translator captures fees at every transaction, regardless of which AI platform wins consumer distribution. That logic explains why Adyen launched Adyen Agentic on June 16, a modular three-layer API suite; Agentic Feed, Agentic Cart, and Agentic Payments. All of these products are designed to connect enterprise merchants' existing systems to any AI commerce surface through a single integration.

A Market Map: Incumbents, High-Growth Players, and Challengers

Adyen. Adyen's positioning is as a universal translator: merchants integrate once, Adyen abstracts every protocol layer. Adyen Agentic is compatible with Meta AI checkout, Google's Universal Commerce Protocol (UCP), the Agent Payments Protocol (AP2), and OpenAI's Agentic Commerce Protocol (ACP). Early partners include American Express, Mastercard, Salesforce, and Visa, and enterprise retailers ESW, Scheels, Sézane, and SharkNinja. The product launched in limited U.S. availability with global expansion planned. Adyen's advantage is its existing enterprise-grade tokenization and fraud infrastructure, which processes trillions of dollars annually. Its watch-out: value narrows significantly if UCP consolidates the protocol layer, because the translation premium disappears.

Stripe. Stripe has been shipping agentic tooling longer than any other incumbent: shared payment tokens, an open-source agent toolkit, live integrations inside ChatGPT and Gemini, and stablecoin wallet infrastructure for machine-to-machine payments through AWS AgentCore. Stripe co-developed the Agentic Commerce Protocol with OpenAI, then pivoted in April to join Google's UCP Tech Council alongside Amazon, Meta, Microsoft, and Salesforce -a tacit acknowledgment that ACP lost the protocol race. Stripe sells to the developer. Adyen sells to the CTO. Neither can yet point to meaningful agentic transaction volume.

Visa, Visa's Intelligent Commerce initiative, including its partnership with OpenAI, positions the network as the credential and trust layer for AI agents. Visa's TAP (Token Authentication Platform) and its AP2 and UCP compatibility give it network-wide issuer trust. The constraint: Visa needs PSP and acquirer partners to complete end-to-end agentic flows. It controls authentication but not checkout orchestration.

Mastercard. Mastercard launched Agent Pay in June 2026, introducing Agentic Tokens and Payment Passkeys to enable trusted agent credentialing, permissioning, and multi-rail settlement. AP4M (Agent Pay for Machines) supports consumer rules and tokenized credentials at the issuer level. Like Visa, Mastercard's execution gap is the merchant-side integration it cannot complete alone.

Google AP2 / UCP . Google's Universal Commerce Protocol has effectively won the standards race. Stripe joined in April. Adyen endorsed UCP from day one. Amazon, Meta, Microsoft, and Salesforce are all inside the council. UCP extends A2A and MCP, with Google Pay and Gemini Search providing consumer-side distribution and merchant-side retail integration. Google's risk is the merchant concern that it owns too much of the discovery and monetization layer simultaneously.

PayPal. PayPal's agentic commerce strategy centers on Store Sync, which connects product inventory and fulfillment directly into the PayPal trust wallet. The combination of consumer trust, buyer and seller protection infrastructure, and dispute resolution gives PayPal a distinct position in flows where the consumer, not the enterprise CTO, is the relevant decision-maker.

Coinbase / x402 Foundation. x402 revives HTTP 402 as an HTTP-native payment standard for machine-to-machine and micropayment flows. With stablecoin rails, ERC-20 and USDC/EURC support, and Coinbase's Developer Platform facilitating settlement, x402 is the strongest challenger for programmatic, agent-to-tool monetization use cases. Its structural limitation is limited merchant-of-record infrastructure and consumer-protection coverage at the scale of Visa or Mastercard.

The Interoperability Parallel That Matters

The last time commerce infrastructure faced this level of fragmentation was the transition from in-store to online retail in the late 1990s. At that point, each merchant ran proprietary checkout, card networks operated on separate rails, and there was no standard for cross-platform identity or session continuity. The resolution wasn't a single winner -it was a set of interoperability standards, primarily SSL/TLS for trust and the card network protocols for settlement, that let the entire ecosystem scale without requiring every player to rebuild from scratch.

Agentic commerce faces an identical structural problem . Each AI platform today runs different protocols, requires different product data formats, and has different checkout and authentication requirements. McKinsey’s analysis of AI’s impact on European e-commerce found that the companies best positioned for AI-driven commerce are those investing now in API-first infrastructure and machine-readable product data -not those waiting for a single standard to win. The parallel is the following: the companies that built clean, interoperable infrastructure in 1999 captured durable positions regardless of which browser or platform dominated distribution.

The luxury sector surfaces the stakes most clearly. McKinsey research on agentic commerce in luxury finds that high-consideration purchase flows: where brand experience, curation, and consumer trust are the actual product. These are the most exposed to infrastructure fragmentation. An agent that cannot reliably query real-time pricing and inventory, or that cannot preserve brand logic through a checkout flow, destroys the value that luxury merchants have spent decades building. The infrastructure problem goes beyond the payments problem and is also a brand and CX problem.

What Investors and Founders Should Watch

The protocol race is largely decided. UCP has the most council members, Stripe's defection from ACP removed its only credible challenger, and Google's distribution through Search and Gemini gives UCP a consumer surface no other standard can match. The remaining competition is at the execution layer: who builds the best enterprise connector between existing merchant backends and a UCP-dominant world.

Adyen’s "build once, transact everywhere" framing is the clearest articulation of that bet. But Adyen itself acknowledges the limits of what a single infrastructure provider can solve. The five constraints it identified; protocol fragmentation, machine-unreadable product data, linear checkout stacks, human-calibrated fraud systems, and non-scalable onboarding require changes across the entire ecosystem, not just at the payment layer. No single company closes all five gaps.

For founders, the whitespace is in the middle layers: machine-readable product data infrastructure, agent identity and credentialing, and fraud systems that can distinguish legitimate AI agents from malicious automation at transaction speed. For investors, the durable position belongs to whoever becomes the trusted settlement and authentication layer for agent-initiated transactions -the role Visa and Mastercard built for human-initiated ones over the past five decades. The incumbents have the trust infrastructure. Whether they move fast enough to own the agentic layer before challengers like x402 or new entrants define it is the central question of the next two years.