Epoch AI Report: Anthropic Generates $9 Million Per Employee, Surpassing OpenAI and Nvidia
A recent report by AI research firm Epoch AI shows that Anthropic generates approximately $9 million per employee, while OpenAI generates about $5.6 million, surpassing all tech giants on the Forbes Global 2000 list.
In comparison, Nvidia generates about $5.1 million per employee; these estimates are based on media reports and average employee growth rates.
Leading AI companies maintain high per-employee revenue growth even as annual revenues reach $10 billion, reflecting extremely high talent density and computing power leverage, breaking through the capital efficiency limits of traditional software companies.
Source: Public Information
ABAB AI Insight
Epoch AI's previous tracking shows that Anthropic's revenue has grown more than ninefold since 2023, while OpenAI's has grown about threefold since 2024; Anthropic has approximately 2,500-5,000 employees, and OpenAI has about 4,000-4,500 employees, both maintaining a highly streamlined structure before reaching the $10 billion revenue threshold.
In terms of capital strategy, both companies prioritize massive funding into top talent recruitment and their own computing infrastructure rather than large-scale sales team expansion, leveraging high-margin subscriptions and API revenues from Claude and GPT products to create a capital efficiency loop of "few people, high output," reducing reliance on traditional sales and back-office personnel.
Similar to the efficiency peaks of early Google and Meta during the search and social media explosion, as well as Nvidia's peak per-employee output during the GPU cycle, current leading AI companies are in the mid-expansion phase of transitioning from model training to product commercialization, where talent density becomes a core competitive barrier.
Structural judgment: This essentially represents a technological replacement. AI models directly replace a large amount of traditional software engineering and operational manpower, with the mechanism being the self-iterative capability of large models and the scale effect of computing power, transforming human capital into extreme leverage, driving the entire tech industry from a "human wave tactic" to a structural reconstruction of concentrated pricing power in "talent + computing power."
ABAB News · Cognitive Law
High per-employee revenue even at $10 billion is the true measure of talent density, not a scale illusion. Traditional software sells manpower, while AI companies sell leverage; ultimately, the market only recognizes unit labor output. Computing power amplifies one person to the output of a hundred, and historical turning points are always rewritten by a few individuals redefining the efficiency formula.