Big Tech companies are putting limits on how much each employee can spend on artificial intelligence (AI), with Apple being the latest company to do so. It cut the daily AI allowance for some engineers to $200 from $300 in October and plans a further cut to $100 for some teams, PYMNTS reported. For some teams outside engineering, Apple is considering a cut to $25 a day from $100.
Apple joined a list that has been growing since spring. Uber used up its AI budget for the year by April, and Chief Technology Officer Praveen Neppalli Naga said the company was “back to the drawing board,” PYMNTS reported. Uber now limits each employee to $1,500 in monthly token spending per AI coding tool, such as Anthropic’s Claude Code or Cursor, an Uber spokesperson told Bloomberg. Employees can seek permission to exceed the limit.
Tesla set a $200 weekly limit per employee starting July 6, according to an internal memo, The Information reported in July. Tesla software engineers had been using thousands of dollars’ worth of tokens each week. Spending above $200 now needs approval. The limit does not apply to beta versions of products from xAI.
Meta and Microsoft limited an outside vendor instead of setting a dollar cap per person. Meta cut the number of employees using Claude Code to 30,000 from about 60,000, largely by pushing them to its own Muse Code and MetaCode tools, PYMNTS reported. Microsoft had been on track to spend $1 billion this year on internal use of Anthropic’s technology and lowered that figure by one-third, The Information reported.
Heavy Spenders Move Routine Work to Open Models
The limits are a departure from August, when the Ramp AI Index showed the top 1% of U.S. businesses spending a median $7,400 per employee a month on AI, PYMNTS reported.
Even then, heavy spenders held back on the priciest model. Fable 5, Anthropic’s most expensive model, made up 6% of the tokens businesses bought from Anthropic and 11.4% of the dollars in its first month, the index showed. Advanced spenders were increasingly buying open-source models instead.
Open models are now a main way companies cut the bill. A growing number of software companies are turning to open-weight models because of the cost of proprietary models from U.S. AI developers, PYMNTS reported. Some of those companies ran into higher upfront costs and a need for specialized talent, and some still use models from the big AI labs when they need the best capabilities.
AT&T shows the savings. AT&T cut the cost of coding and some other advanced AI tasks by as much as 56% by routing employees’ queries to cheaper models when appropriate, and quality fell 2%, PYMNTS reported. AT&T aims to raise the share of queries handled by open-source models to between 60% and 70% from 40%.
62% of Large Companies Spend Over $10 Million on AI
The limits apply to what employees spend on tokens. Company AI budgets are still large. Among U.S. enterprises with at least $1 billion in annual revenue, 62% spent more than $10 million on new forms of AI in the prior 12 months, PYMNTS Intelligence report “AI at Work: Why Deeper Enterprise Use Produces Stronger Returns” found in September.
The companies that use AI most spend more and expect to wait. AI leaders spent an average $14.5 million on new forms of AI over the prior year, compared with $9.1 million among companies in the bottom third, another PYMNTS Intelligence report, “The Agentic Edge Index: How Leading Companies Put AI to Work,” found this month. Among those leaders, 66% expect full returns to take at least three years.
