Major AI labs subsidize consumer and team plans at rates far below API-equivalent compute costs, locking organizations into workflows vulnerable to abrupt repricing. Anthropic Claude Pro usage at $20 monthly equates to $200-400 in API token charges for typical knowledge work, with similar gaps reported for Microsoft GitHub Copilot at over $20 loss per user. OpenAI VP statements and WSJ coverage confirm the subsidy model targets adoption over sustainable unit economics. Agentic capabilities have widened the compute gap, accelerating pressure on providers to shift enterprise focus. Historical cloud vendor patterns show initial low pricing followed by unpredictable increases once workflows depend on the service. Primary analysis from thestateofbrand.com and cross-referenced OpenAI financial disclosures indicate corrections are already under discussion rather than hypothetical.