A small cluster of companies is gobbling up an outsized share of the world's artificial-intelligence computing power. According to Crypto Briefing, OpenAI, Anthropic, and xAI together now consume 21% of global AI compute — more than a fifth of the processing capacity dedicated to running and training AI systems worldwide.
That concentration matters because AI compute — the specialized chips and data centers that power chatbots and model training — is finite and expensive. When three firms account for so much of it, the economics of the entire industry start to bend around them.
The race among those leaders isn't only about who burns the most silicon, though. It's also about reach versus revenue. According to Benzinga, citing figures from research firm IDC, OpenAI's ChatGPT towers over rival Claude on sheer scale, boasting 900 million users. Anthropic, the maker of Claude, has a far smaller audience by that measure.
Yet size of audience and size of business are not the same thing. Benzinga reports that, per IDC, Anthropic may actually be winning the AI revenue race despite ChatGPT's commanding user lead. In other words, the company with fewer users could be generating money more effectively — a sign that enterprise contracts and paying customers may matter more than raw headcount.
Taken together, the two reports sketch a picture of an industry consolidating around a handful of players who command both the infrastructure and the market. OpenAI leads on users and is part of the compute-heavy trio; Anthropic competes on revenue while also ranking among the biggest compute consumers.
Why it matters: when a few firms control both the scarce hardware that powers AI and the largest pools of users and revenue, the direction of one of the most consequential technologies of the decade increasingly rests in very few hands.