In short

A map of the infrastructure of the emerging AI agent economy across three contested layers, communication, payments, and identity, where US open-standards approaches diverge from China's vertically integrated, state-linked model. Citing Alibaba's mass agent deployments and Alipay's agent transactions, it argues this economic layer is 'born captured' by private firms and governments, unlike the early internet, and the window for democratic input is closing fast.

The early internet was built, broadly, as a neutral commons before anyone monetised it. The agent economy is being built the other way round, by the firms and states that intend to own it, before most people know it exists.

This essay maps that construction across three layers: how agents talk (communication protocols), how they pay (the most consequential layer), and how they prove who they are (identity). In each, a US model of open, competing standards is running up against a Chinese model of vertical integration that makes open protocols unnecessary. The choices being locked in now, about payments rails and identity credentials, will shape rent extraction, surveillance, and network power for a long time.

It’s a concrete case study in the project’s core claim: infrastructure is politics, and the window to shape it is open only briefly.

Key takeaways

  • The agent economy already operates at machine scale (e.g. an Alibaba campaign: 10 million orders in 9 hours).
  • Communication standards (Google's A2A, Anthropic's MCP) favour the US; China prefers vertical integration that makes protocols unnecessary.
  • Payments are the most consequential layer: rival US protocols (ACP, AP2, x402) vs China's e-CNY and mBridge.
  • Identity choices (DIDs and Verifiable Credentials vs Alipay/WeChat tied to national ID) embed governance and surveillance questions.
  • Unlike the internet, the agent economy is 'born captured', and the democratic governance window is closing.

Read the full piece

This is a summary. Read the complete essay, with all the sources and argument, on Substack.

Read on Substack →

Frequently asked questions

What is the AI agent economy?
It is autonomous economic activity in which AI agents take actions and make transactions at scales and speeds beyond direct human oversight, built on emerging communication, payment, and identity infrastructure.
How do US and Chinese approaches to agent infrastructure differ?
The US pushes open interoperability standards (A2A, MCP, competing payment protocols), while China pursues state-linked vertical integration via Alibaba/Alipay/WeChat and e-CNY, making cross-platform protocols unnecessary.
Why is the agent economy 'born captured'?
Unlike the publicly funded, neutrality-assuming early internet, the agent economy is being deliberately constructed by private firms and governments, leaving a fast-closing window for democratic governance input.

People & ideas in this piece

Adam ToozeJeffrey DingGoogle (A2A)Anthropic (MCP)StripeCoinbaseAlibaba / AlipayTencent / WeChatModel Context ProtocolAgent2Agent protocole-CNY / mBridgeVerifiable Credentials

Topics: AI & Geopolitics , The Political Economy of AI