Today's AI Daily Brief explores the end of AI's subsidy era, where venture-backed companies can no longer afford to serve AI below cost. The host examines implications for markets, job displacement, and enterprise AI adoption as usage-based pricing becomes the new norm.

The discussion covers how the shift to agentic AI has dramatically increased token consumption, forcing companies like GitHub and Anthropic to abandon flat-fee models. Major themes include compute constraints, pricing restructures, and strategic recommendations for enterprises navigating rising AI costs.

The Agentic Era Drives Massive Token Consumption

Semi-Analysis identified Claude Code as 'the inflection point for AI agents' in February, predicting exceptional revenue growth for Anthropic

Individual token usage has exploded - the host alone consumed a billion tokens last month, equivalent to about 7,500 books worth of words

Between November and April, never-AI-users dropped from 26% to 17%, while frequent users grew from 17% to 24%

Anthropic Struggles Under Compute Constraints

Stratechery's Ben Thompson questioned whether Anthropic's reluctance to release Mythos was due to security concerns or simply lacking enough compute

Technology author Tay Kim wrote: "It's obvious Anthropic vastly underestimated compute growth needs, which is expanding faster than expected"

Anthropic admitted last week to making moves that decreased Claude's performance after investigating user complaints

A Reddit user was charged $200 in usage fees just for having 'Hermes.md' in their git commit history, later refunded as a bug

GitHub Copilot's 6x Price Hike Signals Industry Shift

GitHub announced consumption-based pricing starting June 1st, replacing their unsustainable $39/month flat-fee model

Chief Product Officer Mario Rodriguez explained: "Agentic usage is becoming the default, and it brings significantly higher compute and inference demands"

New multipliers show dramatic increases: Claude Opus from 7.5x to 27x, Gemini 3.1 Pro and GPT 5.3 Codex both from 1x to 6x

Developer Peter Didene called the new multipliers "absolutely ridiculous," predicting users will lock into single vendors to manage costs

OpenAI Capitalizes on Compute Advantage

Sam Altman emphasized OpenAI's inference capabilities: "To a significant degree, we have become an AI inference company now"

OpenAI's Codex app users grew 20x this year, from 200,000 on January 1st to 4 million before GPT-5.5 launched

Greg Brockman highlighted compute efficiency with GPT Images 2.0: "really incredible what you're now able to create with a little bit of compute"

AI Costs Approach Human Labor Parity

Goldman Sachs reports companies are "blowing past their AI inference budgets by orders of magnitude"

Inference costs in engineering are approaching 10% of total headcount costs and potentially reaching salary parity within quarters

Abacus AI's Bindu Reddy wrote: "our AI bill will overtake payroll in six months. We now have limits on how much employees can use our product"

Meta and Microsoft show the transition pattern: headcounts down 10% and 7% respectively, while AI capex up 400%

Enterprise Strategy for the Post-Subsidy Era

Cost savings ranked nowhere in top AI benefits, while new capabilities grew from 21.9% to 29.3% as the primary benefit

Companies should conduct "AI spending leak" audits to identify where premium models do tasks suitable for cheaper alternatives

Recommended approach includes "cheap model bake-offs" to test smaller models against frontier models for specific tasks

"Escape hatch architecture" allows systems to escalate from cheap models to premium ones based on confidence, ambiguity, or case sensitivity

Resources Mentioned

on sophisticated AI collaboration is worth your time

ehaviors are teachable at scale. If you're trying to move from AI access to real capability, KPMG's research on sophisticated AI collaboration is worth your time. Learn more at kpmg.com slash us slash