AI BriefingFrontier AI labs still don't use most of the world's AI compute resources
Key point
Frontier AI labs currently use less than half of the world's AI compute resources, but their growth rate significantly outpaces the industry average.
Details
OpenAI sparked the AI boom, but that doesn't mean they monopolize the world's AI compute resources. As of the end of 2025, OpenAI's compute share is estimated at around 10-15% of global supply, and even combining Anthropic, xAI, Google, and Meta together is estimated to fall short of half of the total.
The world's AI compute resources currently amount to roughly 20 million H100e. A significant portion of this is being consumed not by frontier model development, but across diverse fields such as cloud services, recommendation systems, audio/visual generation, biology, and robotics.
However, the market trend is shifting. OpenAI and Anthropic are securing compute resources at a pace far faster than the industry average. OpenAI is increasing its data center power capacity roughly 4x annually, and Anthropic is also expanding its compute scale on the back of rapid revenue growth.
If this trend continues, it is likely that frontier labs will capture a larger share of global compute resources within the next few years, driving consolidation of the market.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Anthropic Pacing to $100B ARR; Rumors of OpenAI Solving Navier-Stokes Spark ControversyOpenAI and Anthropic's RSI Strategy and the Uncertainty of AI Profitability