Four competing protocols now vie to become the settlement standard for AI agent commerce: Stripe/Paradigm's Machine Payments Protocol (MPP), Coinbase's x402, Google's AP2, and Visa's Trusted Agent Protocol (TAP). The contest intensified on April 14 when Visa launched an anchor validator node on T...
"Agentic payments is very early, and we still are figuring out the best way to structure these." — Matt Huang, Co-founder of Tempo and Managing Partner at Paradigm
Four competing protocols now vie to become the settlement standard for AI agent commerce: Stripe/Paradigm's Machine Payments Protocol (MPP), Coinbase's x402, Google's AP2, and Visa's Trusted Agent Protocol (TAP). The contest intensified on April 14 when Visa launched an anchor validator node on Tempo, the $5 billion payments blockchain co-founded by Stripe and Paradigm that went live on mainnet March 18. Zodia Custody, majority-owned by Standard Chartered, joined simultaneously.
The stakes are large. Gartner projects 40% of enterprise applications will embed task-specific AI agents by end of 2026, up from under 5% in 2025. By 2028, 90% of B2B buying — over $15 trillion in annual spend — could be agent-intermediated. The protocol that captures the settlement layer stands to extract fees from a market that dwarfs current DeFi volumes by orders of magnitude.
Yet actual transaction data tells a sobering story. Coinbase's x402, the most mature entrant, processes roughly $28,000 in daily volume, much of it test traffic. MPP launched with 100-plus integrated services but has no publicly reported volume figures. The protocol war is being fought over infrastructure positioning, not present-day revenue.
Tempo raised $500 million in its October 2025 Series A at a $5 billion valuation, backed by Thrive Capital (Joshua Kushner) and Greenoaks Capital alongside lead investors Stripe and Paradigm. The chain went through 3.5 months of testnet before launching mainnet on March 18, 2026.
The architecture departs from general-purpose chains in three ways:
No native gas token. Transaction fees — targeted at under $0.001 per TIP-20 transfer — are paid directly in stablecoins (USDC, USDT). Validator compensation is denominated in stablecoins rather than a volatile native asset. This eliminates the circular dependency where users must acquire a speculative token to use a payments network.
Dedicated payment lanes. Non-payment activity cannot congest payment transaction processing. The chain claims throughput above 100,000 TPS with sub-second deterministic finality (approximately 0.6 seconds) and no chain reorganizations.
Permissioned-to-permissionless transition. Tempo currently operates with a defined validator set. The roadmap calls for permissionless validation, though no firm date has been announced.
Engineering is led by Liam Horne, former CEO of Optimism, and Paradigm CTO Georgios Konstantopoulos. Matt Huang, Paradigm's co-founder and a Stripe board member, co-founded the project.
Design partners that worked with Tempo ahead of mainnet included Visa, Mastercard, Deutsche Bank, Standard Chartered, Revolut, Nubank, Shopify, OpenAI, Anthropic, Ramp, DoorDash, Klarna, Kalshi, and UBS — a coalition spanning traditional banking, fintech, and AI.
On April 3, Tempo integrated with Dune Analytics, opening queryable data on throughput, stablecoin transfers, and fee dynamics. This marks an early step toward the transparency expected of public blockchains, though comprehensive volume data remains sparse.
On April 14, Visa announced it had launched a validator node on Tempo following six months of collaborative engineering integration, enabling direct connection between Visa's secure internal infrastructure and the blockchain network. Cuy Sheffield, Visa's Head of Crypto, stated: "By operating a validator on Tempo, we're extending Visa's commitment to reliability, security, and trust into blockchain networks — supporting the development of stablecoin payment systems that meet the high operating standards our clients and partners expect."
Zodia Custody, the crypto custodian majority-owned by Standard Chartered, joined as validator on the same date. Stripe itself operates as validator. These three constitute Tempo's first external validators beyond the founding team.
The significance is structural. Visa processes approximately $14.8 trillion in annual payment volume. Its decision to run infrastructure — not merely integrate via API — represents operational commitment to blockchain-based settlement. The company now operates validator or settlement infrastructure across three blockchain networks: Solana (USDC settlement with Cross River Bank and Lead Bank), Circle's Arc (design partner), and Tempo.
Visa's stablecoin-linked card issuing programs now exceed 130 across 40-plus countries. Its international pilot programs collectively surpassed an annualized $3.5 billion in stablecoin volumes as of November 2025, while Visa-linked crypto card spending surged 525% year-over-year to monthly transaction volumes above $1.5 billion by late 2025.
The agentic payments space has fragmented into four competing protocol standards. Each occupies a different layer of the emerging stack.
Backers: Stripe, Tempo, Paradigm Launched: March 18, 2026 Mechanism: Sessions-based model where an agent pre-authorizes a spending limit and streams granular micropayments continuously. Thousands of small transactions aggregate into single settlement transactions. Supports stablecoins natively and fiat via Stripe Shared Payment Tokens. Visa contributed specifications for card-based agent payments. Lightspark provides Bitcoin Lightning integration. Adoption: 100-plus integrated services at launch; no public volume data.
Backers: Coinbase, Cloudflare, Circle, AWS Launched: V2 December 2025 Mechanism: Embeds stablecoin payments directly into HTTP using the 402 status code. Stateless, no accounts or API keys required. Multi-chain: Base, Ethereum, Polygon, Solana, Avalanche, Sui. Zero protocol fees — only blockchain gas on L2s. Adoption: Claims 100 million payments across APIs, apps, and AI agents in six months. However, according to CoinDesk onchain data, daily volume stands at approximately $28,000, described as predominantly testing and promotional transactions. Google, OpenAI, and Circle joined x402's nonprofit governance under the Linux Foundation on April 2, 2026.
Backers: Google, 60-plus partners (Adyen, American Express, Mastercard, PayPal, Coinbase, Revolut, Worldpay) Mechanism: Two-step cryptographically signed mandate system (Intent Mandate, Cart Mandate) extending Google's Agent2Agent (A2A) protocol. Payment-method agnostic. Jointly created A2A x402 extension with Coinbase, Ethereum Foundation, and MetaMask. Position: Defines the authorization and trust layer rather than the settlement layer.
Backers: Visa, 100-plus global partners, 30-plus in sandbox, 20-plus agent platforms Mechanism: Cryptographically signed HTTP messages verifying agent identity and payment authorization. Works within existing card network infrastructure. Early adopters include Nuvei, Adyen, and Stripe. Limitation: Inherits card-rail economics — interchange fees, settlement delays, and chargeback mechanics are poorly suited for high-frequency, low-value agent transactions.
The protocols are not strictly competitive. Industry analysis suggests the likely outcome is a layered stack: AP2 or TAP handles trust/authorization, x402 or MPP handles settlement/execution. The question is which layer captures the most economic value.
Tempo is not building in isolation. Three purpose-built stablecoin Layer 1 chains are now in various stages of deployment:
Tempo (Stripe/Paradigm): Mainnet live March 2026. $500M raised. Stablecoin-denominated fees. Payments-first design. Permissioned validator set expanding toward permissionless.
Arc (Circle): Public testnet live. Mainnet planned for 2026. Built as the native chain for USDC, integrated into Circle's full-stack platform (CCTP, Gateway, institutional on/off-ramps). CEO Jeremy Allaire disclosed on April 14, 2026 in Seoul that Circle is exploring a native token. Arc will debut with quantum-resistant cryptographic features. Visa is a design partner.
Stable (Tether): Under development. Dedicated L1 for USDT. Limited public technical details.
Each represents a stablecoin issuer or payments company building dedicated blockchain infrastructure rather than relying on general-purpose chains. The economic logic: if stablecoins constitute the primary use case, a chain optimized for that use case can deliver lower fees, faster finality, and purpose-built compliance tooling.
Total stablecoin transfer volume hit $33 trillion in 2025, expanding 72% year-over-year. The three dedicated chains are competing to capture settlement share of that flow.
The gap between projected market size and current adoption is stark.
Projections:
Current reality:
The infrastructure is being built for a market that does not yet exist at scale. This is consistent with historical patterns — TCP/IP infrastructure preceded the commercial internet by years. However, it also means current valuations (Tempo at $5 billion, x402 ecosystem at roughly $7 billion) rest entirely on forward projections rather than current revenue.
The economic question is where value accrues in the emerging stack.
Validator/Settlement layer: Tempo's model pays validators in stablecoins, eliminating token speculation from the consensus incentive. This is economically cleaner but potentially caps validator compensation to thin transaction margins. At $0.001 per transaction, even 1 billion annual transactions would generate only $1 million in total fees — requiring massive volume to justify infrastructure costs.
Protocol layer: x402 charges zero protocol fees, relying on underlying blockchain gas. MPP aggregates micropayments into batch settlements, reducing per-transaction costs but extracting value through Stripe's fiat processing rails (typically 2.9% + $0.30 for card transactions).
Authorization layer: AP2 and TAP define permissions and trust frameworks but do not directly process payments. Their value capture depends on becoming required middleware — extracting licensing or integration fees from downstream settlement providers.
Application layer: The actual agent platforms (OpenAI, Anthropic, Google) that generate payment demand have the strongest structural position. They control traffic regardless of which protocol settles the transaction.
The pattern mirrors early internet economics: the most valuable positions were held by those who controlled demand (Google, Amazon) rather than those who built protocol infrastructure (Cisco, ISPs). Current capital allocation — $500 million into settlement infrastructure at Tempo — may be mispriced if the value capture layer sits elsewhere in the stack.
Visa's April 14 validator launch on Tempo marks the first time a $14.8 trillion-volume payment network has operated blockchain consensus infrastructure for a startup chain, moving beyond API integration to direct operational involvement.
Four competing agentic payment protocols (MPP, x402, AP2, TAP) are positioning across different layers of a payment stack that does not yet process meaningful volume. x402's $28,000 daily volume contrasts with projections of $15 trillion in agent-intermediated B2B spend by 2028.
Three purpose-built stablecoin L1 chains (Tempo, Arc, Stable) from Stripe, Circle, and Tether respectively are fragmenting what was previously a general-purpose chain use case. Combined stablecoin transfer volume of $33 trillion in 2025 provides the addressable market.
Tempo's stablecoin-denominated validator compensation and fee structure eliminates native token speculation from the economic model but caps fee revenue to thin margins requiring enormous volume to justify the $5 billion valuation.
The protocol war's outcome likely produces a layered stack rather than a single winner. The highest-value position may sit at the application layer (AI platforms controlling demand) rather than the settlement layer currently attracting the most capital.
The entry of Visa, Standard Chartered, Mastercard, Deutsche Bank, and UBS into blockchain validator and design partner roles represents a structural shift from experimentation to operational integration. These institutions are not buying tokens or funding accelerators. They are running infrastructure.
The timing coincides with a protocol standardization battle reminiscent of early internet networking wars. Four standards are competing, none has achieved dominant adoption, and the installed base of actual agent-initiated transactions remains negligible relative to projections.
For the economic value to materialize, two conditions must be met. First, AI agent commerce must transition from developer sandboxes to production-scale deployment — the 40% enterprise adoption figure Gartner projects for year-end 2026 will serve as the first measurable checkpoint. Second, the protocol stack must consolidate enough for merchants and service providers to integrate without supporting four parallel standards.
Until those conditions are met, the $500 million deployed into Tempo and billions in aggregate protocol ecosystem valuations represent positioning bets on a market that, as Matt Huang himself acknowledged, participants are "still figuring out."