BlockBeats News, July 1st, the use of Enterprise AI is shifting from "maximizing usage" to "budgeted usage." According to a Token Budgeting report released by SemiAnalysis on July 1st, the earlier trend of token maxing, which encouraged employees to consume AI tokens as much as possible to boost productivity, is now being replaced by a more realistic budgeting system. However, the organization believes that media narratives about enterprises cutting AI spending have been exaggerated, and OpenAI and Anthropic's API business did not face substantial budget risks in the latter half of this year.

The SemiAnalysis team stated that after engaging with over 50 enterprise clients through Slack, phone calls, and the Databricks AI Summit, they found that most companies have indeed started to set limits on AI usage, but there is no uniform standard. Lower-end budgets may be as low as $250 to $500 per person per month, while higher-end budgets can reach up to $2,000 per month or even tens of thousands of dollars. A major U.S. aerospace and defense manufacturer set some employees' monthly limits at $250, a large pharmaceutical company set theirs at $500; tech-forward companies like Workday and Stripe allocated a portion of employees a budget of around $2,000 per month.

This contrasts with the earlier concept of "token maximization." The report mentioned that companies like Meta and Salesforce had encouraged employees to heavily use AI tools. Meta even had an internal dashboard named "Claudeconomics" that ranked the top 250 heavy users in the company. Data showed that Meta employees consumed over 60 trillion tokens within 30 days, with the highest individual user consuming approximately 280 billion tokens. The dashboard was shut down two days after related reports emerged. Uber was also reported to have depleted its annual budget for Claude Code and Codex within four months, then proceeded to set a limit of $1,500 per person per month, with excess requests requiring individual approval.

However, SemiAnalysis believes that these extreme cases more reflect incentive mechanisms and loose management rather than an overall peak in enterprise AI spending. The report stated that the top 10% of high-consumption customers contributed most of the AI lab's revenue, and these customers have a low risk of cutting API expenses for the rest of the year. Even though Meta consumed around 700 trillion tokens per month in February and estimated that each employee cost nearly $50,000 per year at list price, SemiAnalysis estimates that this still accounts for only 3% to 5% of Anthropic's revenue.

Corporate expenditure distribution is also highly uneven. SemiAnalysis cited Ramp data stating that the top 1% of customers spend nearly $90,000 per employee per year on AI, the top 10% of customers around $7,300, while the median customer spends only $136. The organization also mentioned that many tech-leading Fortune 500 companies still have an average AI spend per employee of less than $2,000, with large expenditures mainly concentrated in the engineering and data science departments. This implies that there is still significant room for growth along the enterprise AI adoption S-curve.

The rise of budget governance is transforming employee usage patterns. Some companies are shifting from the default model from Opus to Sonnet, turning off advanced models or quick modes; while some employees first draft and summarize using Microsoft 365 Copilot, then use the more expensive Claude or Codex tokens for critical tasks. A global travel technology company spends close to $10 million annually on AI, recently switching the default Claude model from Opus to Sonnet but still allowing employees to manually switch back to Opus. Certain roles have a default budget of only $200 per month, but engineers or senior staff can request higher limits.

SemiAnalysis's conclusion is that budget management will persist in the long term, but it does not equate to diminished demand. Instead, companies are transitioning AI from an experimental tool to formal cost management. Coding is currently the strongest demand vertical, with SemiAnalysis estimating that over 70% of OpenAI and Anthropic's ARR can currently be attributed to coding scenarios. In the future, cybersecurity, white-collar knowledge work, enterprise collaboration, and automated office workflows may replicate the growth trajectory of Claude Code, Codex, and Copilot in the developer marketplace.

This signifies that the AI market is entering a new phase. Early adopter enterprises may have been willing to foot a vague bill for "AI experimentation"; now, finance departments are beginning to demand budgets, quotas, and ROI. However, as long as gains in employee efficiency offset costs, companies will not cease token purchases. For AI model companies, the risk is not clients suddenly abandoning AI but rather having to prove that every dollar of token consumption translates into faster code, shorter recruitment processes, improved sales efficiency, or reduced human effort.