PerceptAI Beating AI News, SemiAnalysis founder Dylan Patel predicted in a recent podcast that by 2028, OpenAI and Anthropic could command 70% to 80% of the global AI supply, with a combined computing power exceeding 100GW. Patel stated that the two companies currently contribute around 30% of the world's annual new AI supply, a percentage that is rapidly increasing.
Patel pointed out that the business model of leading AI labs is undergoing a transformation, with a significant increase in AI computing efficiency. Anthropic is currently generating approximately $50 million in revenue per megawatt of computing power and may see this figure rise to $100 million. This enables OpenAI and Anthropic to acquire or lease computing power at a premium of $25 million to $50 million per megawatt.
Of particular note, Patel projected that global AI-related capital expenditure from 2024 to 2029 will reach around $11 trillion, with over $5 trillion requiring financing through debt. Due to the significantly higher potential returns of AI infrastructure compared to traditional industries, tech giants may accept higher financing costs, leading to an increase in overall credit rates and putting pressure on the valuation of traditional assets and highly leveraged economies.
Furthermore, Patel believes that future new computing power may not necessarily be primarily used for external model inference services, but may be more directed towards internal R&D within AI labs and model self-improvement.

