This week reminded me of the early cloud days. Everything, including my cleaning service, was "moving to the cloud" and no one could explain the ROI yet. Same pattern. Different decade.
The Accountability Era Has Begun
For two years, engineering set AI spend. This week, finance took it back.
One enterprise customer racked up $500 million in Claude charges in 30 days. The FT published the chart showing four of five hyperscalers have negative AI ROI through 2030. A Fortune 100 CEO said AI has not displaced any of his 341,000 employees. An academic study said AI gets 20% of health questions wrong. Cisco said no frontier model is safe.
The marketing era is over. The audit era has begun.
This memo covers 3 signals that flipped the AI conversation this week, plus 3 actions before the next board meeting.
What Changed This Week
1. A single enterprise ran up a $500 million Claude bill in 30 days. No governance. No spend caps. Just employees burning tokens. The same week, Ramp data showed AI token spend up 13x across 50,000 companies since January 2025. Microsoft canceled most of its Claude Code licenses on cost grounds. Gary Marcus published the FT chart proving only Amazon clears positive AI ROI through 2030. The era when engineering could spend whatever it wanted on tokens just ended.
2. The vendor market restructured in seven days. Amazon began selling its e-commerce AI to competitors via AWS, with Kate Spade as the first customer. Anthropic closed $65B at a $965B valuation on $47B annualized revenue, passing OpenAI. Microsoft will unveil in-house AI coding models at Build June 2-3 to compete with Claude Code and Cursor. GPU rental prices fell 30%. A year ago, Fortune 500 buyers had one realistic frontier model decision. Today they have four wholesalers. CFOs and CIOs who renegotiate in the next 90 days will get terms that customers in the next 180 days will not.
3. Three independent audits landed in the same week. Costco CEO Ron Vachris told the Economic Club of Chicago that AI has not displaced any of Costco's 341,000 employees. Penn State scored AI chatbots at 76.2% accuracy on health questions, with error rates double those of human physicians. Cisco showed every major frontier model fails multi-turn iterative attacks at success rates up to 88.3%. The marketing claims have a measurement problem. The measurement just landed.
Bottom Line
If your board is still asking how much to invest in AI, you are asking last quarter's question. The right question is whether you can defend what you have already spent, prove what it returned, and explain what it broke.
The Accountability Era is here. The companies that survive it will be the ones whose finance, security, and HR functions get the same seat at the AI table that engineering has had for two years.
On My Desk This Week
Few other reads I found useful this week (no overlap with the signals above):
Brian Merchant on Anthropic and the Vatican: The contrarian voice on the $965B story.
Pope Leo XIV, "Magnifica Humanitas": The first papal encyclical on AI.
CodeRabbit on AI code quality: AI-generated code introduces 1.7x more issues than human code.
NextEra-Dominion $67B merger: The energy layer is now part of the AI stack.
White House scraps AI executive order: Tech CEOs called the President. The order was shelved.
IBM 2026 CEO Study: 79% decentralizing. 76% have a CAIO (up from 26% in 2025).
Alibaba Qwen3.7-Max ranks #4 on Code Arena: Only Anthropic ranks higher.
Rohit Prabhakar CMO. CDO. Transformation Leader. Building growth engines where commercial instinct meets AI.