Internet of Payments
TLDR
- EQT Ventures maps agentic payments as a three-phase shift unlocking ~$2.5 trillion in global payments revenue, requiring new infrastructure from instruments to identity.
Key Takeaways
- Three phases: humans paying on AI interfaces (chat/voice), agents paying on human interfaces (RPA), and agents paying on agent-native interfaces (MCP/x402).
- Existing card rails fail agents: traditional fraud systems block rapid micro-transactions; e.g., reading a DoorDash ordering page via HTML already costs $1.87 in compute.
- New instrument categories needed: scoped virtual cards (Stripe, Payman, Nekuda), agent wallets (Coinbase AgentKit, Catena), and agent identity registries (Cheqd, MIT Media Labs Nanda).
- x402 protocol uses HTTP 402 status to let servers request micro-payments before serving responses, with a facilitator verifying and settling the transaction inline.
- Honey ($4B PayPal acquisition) is cited as the first agentic consumer payments product; loyalty and rewards logic may shift from card issuers to agents.
Why It Matters
- Agentic payment volume maps to all card-not-present transaction revenue globally, a market EQT sizes at ~$2.5 trillion, not a niche fintech wedge.
- Existing checkout flows do not accept scoped tokens from Visa or Mastercard, meaning Phase 2 adoption requires new checkout UI or entirely new rails.
- KYA (Know Your Agent) compliance registries are nascent but will compound as agent transaction volume scales, creating a new AML/KYC surface.
· 2026-04-28 · Read the original
Next →