AI Is Slowing Down

Source: Where’s Your Ed At Date: 2026-06-08 URL: https://www.wheresyoured.at/ai-is-slowing-down/

Summary

Ed Zitron’s essay arguing that AI adoption and revenue growth are decelerating precisely when the industry’s infrastructure bets require acceleration. The core tension: data center buildouts estimated at $9.5–$15 trillion require demand that doesn’t exist — projected combined revenues for OpenAI and Anthropic reach only ~$358 billion by 2029. After major platforms shifted to token-based billing in Q1 2026, companies quickly hit ROI visibility problems; KPMG data shows only 26% of firms have a comprehensive view of AI costs. Specific cases cited: Microsoft’s AI team expressing intent to reduce Anthropic usage due to cost, Uber capping AI spend at $1,500/month per user, Brex at $500/week for engineers.

Implications

• Capital markets / demand thread: The infrastructure-to-revenue gap Zitron names is the structurally important argument, not the behavioral slowing. If $9–15T in compute investment needs to be justified by $358B in projected AI revenue, the math requires either a step-change in enterprise AI value creation or a significant infrastructure writedown cycle.
• Enterprise adoption thread: Token billing exposing ROI opacity is a real and underreported friction point. Enterprises that can’t measure AI spend against outcomes will throttle usage — the spending caps at Uber and Brex are early data points of a broader pattern worth watching in Q2/Q3 earnings calls.
• Counterweight to the “AI everywhere” narrative: Pair this signal with the OpenAI S-1 filing — the IPO will expose actual revenue figures and growth rate, providing a public reality check against both bull and bear projections.
• Coding agents specifically: If enterprise ROI pressure is tightening AI spend, developer-facing tools (coding agents, IDE integrations) face the most direct pressure to demonstrate measurable productivity lift on a per-seat basis.