Visa, Mastercard, x402, MPP, Google UCP, AWS AgentCore — three competing architectures now in production for agent payments. The structural question: they are fighting over plumbing, not commerce.
Google, Mastercard, and Coinbase are collaborating through standards bodies to define the rules of AI agent commerce before governments can. But technical security and consumer trust are not the same thing.
Three competing architectures now live for agent payments. The developer choice between them tells you where the industry thinks agent commerce actually starts.
Bain, McKinsey, and JP Morgan project $300B-$1T in agent commerce by 2030. Current protocol data shows $50M. The gap between projections and evidence is the structural story.
ClickUp deployed 3,000 AI agents and restructured around a model where fewer employees do more work for dramatically higher pay. Industry data suggests most companies can’t make it pay.
A new study of 21,599 U.S. firms finds that companies spending the most on AI hired more people, including at the entry level. But the gains stopped there.
Microsoft launched Frontier Company, a $2.5B program embedding 6,000 AI engineers in customer enterprises. Four days later, it cut 4,800 of its own roles. The juxtaposition raises questions about who benefits from the agent transition.
Oracle cut 21,000 jobs (13% of workforce) and went $23.7 billion negative on free cash flow to fund a $55.7 billion AI infrastructure buildout. A bondholder lawsuit alleges the company concealed the scale of its planned borrowing.
Microsoft eliminated 4,800 roles while investing $190B in AI infrastructure. Analysts say workforce reductions fund AI capex. The ‘not replacing people’ framing examined.
More than half of leaders who cut staff for AI say they were wrong. Three anchor cases show what happens next.