Programmers are increasingly using artificial intelligence, and for some companies it has become a real competition

AI tokens as a new KPI: companies track “code consumption”
In several leading tech firms an unusual way of measuring employee effectiveness has emerged — counting the tokens they “eat” while working with artificial intelligence. A token is a small unit of text (roughly a word) that models process and generate. The more tokens an employee uses, the higher their rating in corporate tables.

Record figures
- OpenAI: one engineer interacted with AI bots for 210 billion tokens in a week — enough to “fill Wikipedia” 33 times. This is a new record among all employees.

- Anthropic: a Claude Code user spent over $150 k per month using tokens to generate code.

How companies use the data
At Meta✴ and Shopify leaders include “token consumption” in productivity assessments. Programmers who actively use AI tools receive bonuses, while those who avoid them are noticed by management.

> “This is a new reality for programmers: neural networks promised to boost productivity and cut costs, but now they have turned into a status game — tokenmaxxing.”
> — *Max Linder*, developer from Stockholm

Token maximization
Initially advanced users could consume thousands of tokens per day: a student wrote an essay with 10 k tokens (≈7,500 words). Millions of tokens would require many hours of work, and billions seemed unattainable.

The emergence of AI agents for code generation changed the situation. Such agents can operate autonomously, analyzing and editing large codebases, creating new programs based on user requests. They generate thousands of tokens per iteration, and systems like OpenClaw run 24/7.

> “If you have several constantly running agents, one full-fledged agent can consume 700 million tokens a week. That’s not that much.”
> — *Ege Erdel*, co-founder of Mechanize

He himself estimated his consumption from 1 to 10 billion tokens per week.

Financial implications
Companies selling tokens see revenue growth. Anthropic doubled its revenue forecast in two months thanks to the rise of AI code agents. OpenAI reported that Codex tripled the number of weekly active users since the start of the year, and total token consumption grew fivefold.

Google announced its own plans to scale its models, but exact figures have not yet been disclosed.

Bottom line:
AI tokens are becoming a new KPI. Employees compete for “code consumption,” and companies use this data as a productivity metric and an additional revenue source. This changes the culture of programmers’ work and opens new financial opportunities for AI startups.