Frontier AI labs advocate for regulatory pacing on safety grounds. They also gain financially by prolonging premiums on existing models. Rules that slow new model releases protect their investments and defer costly spending. Without such rules, a lab slowing down risks rivals advancing past it. Newer models tend to be both smarter and cheaper, driving rapid customer switching. The price of intelligence halves roughly every 46 days. Usage data shows demand split between cost-conscious and quality-conscious customers. Quality-conscious labs like OpenAI and Anthropic dominate the premium market. Cost-conscious providers target buyers prioritizing affordability. The competitive cycle forces labs to invest repeatedly to maintain leads. Each cycle risks widening the gap between development costs and earnings. The essay argues labs’ safety claims warrant independent scrutiny due to financial incentives.

• price of intelligence halves every 46 days
• $1.72 vs $8.75 per task (GPT-6 Astra vs Fable 5)
• $60.1M usage in 40–45 score band
• $55.9M usage in 50–55 score band