AI's Economics Don't Make Sense

4 hours ago

GitHub Copilot will switch to usage-based pricing on June 1, 2026, ending its flat-rate subscription model.
Microsoft claims the change is due to higher compute costs from AI advancements, but critics argue it was always economically unsustainable.
Historically, AI services like Copilot have been heavily subsidized, with users costing companies more per month than they paid.
The shift to token-based billing reveals the true expense of AI, where tasks like coding can cost $11 per request.
AI subscriptions are criticized as deceptive, hiding real costs and creating unsustainable user habits.
OpenAI and Anthropic face enormous financial commitments, needing hundreds of billions in revenue to cover compute contracts.
Data center economics are precarious, with projects like Oracle's Stargate Abilene relying on unproven AI demand.
Many AI companies are unprofitable, with questionable ROI, and may collapse if subsidies end and token billing becomes standard.
The AI industry is accused of being a con, built on obfuscated costs and exaggerated capabilities.
The future of AI depends on unrealistic growth projections, risking widespread financial fallout if demand doesn't materialize.

Hasty Briefsbeta