Barclays' latest research report indicates that OpenAI's safety monitoring requirements for high-capability models, disclosed on August 18, 2026, are expected to increase inference and post-training computing power demand by 20%. This will drive an approximate 18% increase in overall computing costs for the AI industry starting from 2027, corresponding to over $44 billion in new expenditures. The report suggests that AI safety regulation is evolving from an ethical issue into a quantifiable cost variable, potentially reshaping the investment logic for AI infrastructure. Barclays anticipates that the gross margins for inference at AI labs will gradually converge from their current level of over 80% to a long-term average of 65%.