Decoder with Nilay Patel The AI industry's existential race for profits
258 snips
Apr 9, 2026 
Hayden Field, a senior AI reporter covering OpenAI and Anthropic, digs into AI’s scramble for real profits. She gets into IPO pressure, why OpenAI pulled back on Sora, Anthropic’s tighter Claude strategy, the shift toward enterprise coding, soaring compute costs, executive turmoil, and how public skepticism could shape what survives.
AI Snips
Chapters
Transcript
Episode notes
AI Compute Scarcity Moved From Training To Inference
- The bottleneck shifted from training bigger models to serving heavy inference demand from agentic products, especially coding tools.
- Nilay Patel contrasts the old AGI-through-scale narrative with today's scramble to supply enough compute for customers actively using models.
Agents Broke Anthropic's Old Subscription Economics
- Agents changed the economics because they consume vastly more tokens than normal chat use, forcing Anthropic to tighten pricing around Claude.
- Hayden Field says OpenClaw can prompt Claude far beyond human rates, so Anthropic both protects scarce compute and pushes users into its own co-work moat.
The Profit Path Runs Through Enterprise And Government
- Hayden Field says any real path to profit likely runs through enterprise, military, and government contracts rather than consumer subscriptions.
- Even $200 monthly power-user plans cannot match the revenue scale of back-office software and public-sector deals.