Regulatory Proposals from Frontier AI Labs: The Economic Incentives of 'Pacing' Hidden Behind Safety Rationales

Behind the 'safe pacing' proposals advocated by frontier AI labs lie strong economic incentives of investment protection and premium maintenance. In a structure where the market share of existing models plummets upon the release of new models, regulatory adjustments function as a means to suppress competitors and defend commercial positions.

Rapid Collapse in Model Prices and Disappearance of Premiums

The rate of price decline for benchmark-standard intelligence is extremely fast. A trend is observed where prices drop to 3–7% of the initial price within 98–294 days after the launch of flagship models, rapidly accelerating customer churn. For example, when comparing GPT-6 Astra and Fable 5, the score difference is only 0.56 points, yet the cost differs by 2.3 times.

Financial Inefficiency of the 'Racing' Scenario

Simulating a 6-month development cycle scenario without pacing reveals that cumulative development spending over 36 months reaches $112.6B, while customer contributions amount to only $33.4B. Maximum unrecovered spending reaches $79.2B, and contribution margins plummet from 40% to 10%. In contrast, a pacing scenario that extends the development cycle to 15 months allows for a turnaround to profitability by the end of 36 months.

Dependency Revealed by Actual Financial Data

The actual financial status of companies starkly reveals their dependency on external funding. OpenAI covered only about 29% of its 2025 R&D costs ($19.18B) with its own revenue. The R&D coverage ratios for MiniMax and Z.ai are merely 7% and 11.8%, respectively. In this structure, pacing through government intervention acts as a rational choice for leading companies to escape a 'dilemma' and secure massive financial benefits.