Inference got 430x cheaper per token, but agents burn 100x more of them. Real savings on agentic workloads is 4.3x, not 430x. Why the curves diverge.

Apple just handed Google the intelligence layer of 2.2 billion devices, and the contract clause protecting users costs Google almost nothing.

Mastercard and Visa announced agent credentialing systems on the same day. The claim underneath: the right to decide which AI agents can act in commerce.

Every open-weight AI model runs on CUDA. NVIDIA owns the platform and the chip beneath it, so the substrate captures the economics the model gives away.

Alibaba ran 28.8 million queries through Claude to distill it. The API that monetizes a frontier model is also the vector that copies it.

Cross River and Stripe mint a single-use card for every agent purchase. The autonomous agent economy is really delegation under bank supervision.

Thursday, September 17, 2026 at 11:00 AM Eastern Daylight Time.

S&P cut Oracle to one notch above junk over OpenAI counterparty risk. The AI dependency now runs up the capital stack to lessors and bondholders.

CME compute futures let traders set GPU rental prices. From October 5, labs and neoclouds pay whatever the pit decides.

Everyone needs a personal brand - from the company’s newest intern, to the most established CEO, and in the AI era the rules of engagement are changing fast. In the second in a series of films, FT Working It editor Isabel Berwick explains how to harness the…