In the history of technology, protocol wars are not unusual. TCP/IP vs OSI. HTTP vs Gopher. WiFi vs Bluetooth. In each case, a new communications or commerce layer required a standard, multiple parties proposed competing standards, and the market eventually converged — usually on the standard with the largest installed base, not the technically superior one.
The agentic payments protocol war is now live. As of April 2026, four major protocols are competing for the standard that will govern how AI agents discover products, initiate transactions, and settle payments. The outcome will determine the infrastructure layer of the next decade of commerce.
Google Universal Commerce Protocol (UCP)
Google's UCP is the most merchant-friendly of the competing protocols, according to early feedback from retailers. It gives merchants more control over how their products are presented to agents — pricing, availability, promotions — than OpenAI's ACP. Google announced UCP integration with Stripe and Salesforce in March 2026, giving it immediate reach into a large fraction of e-commerce infrastructure.
The distribution advantage is significant. Google's Gemini AI has 750 million monthly active users, making it the second-largest consumer AI by usage. Every Gemini user is a potential UCP participant. Gap announced a partnership with Google to support shopping within the Gemini app in March 2026, the first major fashion retailer to commit to the protocol.
UCP's weakness is that it is a Google protocol. Merchants who have experienced Google's historical leverage in search — the ability to change ranking algorithms and effectively regulate visibility — are cautious about ceding similar leverage in agentic commerce.
OpenAI Agentic Commerce Protocol (ACP)
OpenAI's ACP launched alongside the broader ChatGPT commerce push in late 2025. With 900 million weekly active users as of February 2026, ChatGPT has the largest installed base of any consumer AI. ACP's distribution advantage is therefore significant.
The Instant Checkout initiative, which attempted to implement ACP-based purchasing directly within ChatGPT, was shuttered in March 2026 due to low adoption. OpenAI has since refocused its commerce strategy on ChatGPT Apps — allowing third-party developers to build commerce experiences within the ChatGPT interface — rather than direct checkout. This represents a strategic pivot from platform-as-merchant to platform-as-marketplace.
ACP's technical architecture is more permissive than UCP, giving agents more autonomy to navigate and transact across merchant sites. Merchants find this less controllable; consumer advocates find it more useful. The tension between these perspectives will shape ACP's adoption trajectory.
Visa Machine Payments Protocol (MPP)
Visa's Machine Payments Protocol takes a different approach from Google and OpenAI: rather than building a new discovery and commerce layer, it focuses on the payment execution layer. MPP provides the cryptographic infrastructure for agents to initiate card-based transactions — using tokenised credentials, verifiable agent identity, and programmable spending limits.
Visa announced extended support for MPP in March 2026, with Stripe and Tempo confirming implementation and a full SDK release. As part of the Visa Intelligent Commerce infrastructure, MPP benefits from Visa's Trusted Agent Protocol — a framework for verifying agent identity and delegation chains that is more mature than anything Google or OpenAI has proposed.
MPP's limitation is scope. It solves the execution problem — how an agent presents a payment — but not the discovery and commerce problems that UCP and ACP address. A complete agentic commerce stack will likely require both an application-layer protocol (UCP or ACP) and an execution-layer protocol (MPP or equivalent).
Alibaba Alipay Protocol
Alibaba's protocol, operating through the Qwen AI app with 300 million monthly active users, is primarily a China-first solution. Its cross-border integration scope remains unclear, and the geopolitical constraints on Alibaba's Western expansion limit its near-term relevance outside Asia. In the medium term, it will be the dominant standard for agentic commerce in China and potentially in Belt and Road markets.
What the outcome determines
The protocol that wins will not just determine how agents pay — it will determine who can see what agents are buying, who can influence what agents choose, and who captures the margin in agentic commerce. The infrastructure stakes are significant: the winning protocol operator will have structural insight into a $1.5 trillion market by 2030.
The most likely outcome is not a single winner but a layered standard: application protocols (UCP and ACP) sitting above execution protocols (MPP), with interoperability requirements imposed by regulators as the market scales. The firms investing in the infrastructure layer now — the identity, authorisation, and settlement primitives that all four protocols depend on — are the ones that will be positioned regardless of which application protocol prevails.