Tesla limits its employees’ AI spending to $200 per week – just a few months after the company had actively encouraged the workforce to intensify AI usage. According to an internal memo first reported by The Information, the new cap has been in effect since July 6, 2026.
What the internal memo stipulates
As Electrek reports citing The Information, employees will now need approval from supervisors once their weekly spending on AI tools exceeds the $200 mark. Individual software engineers had reportedly incurred several thousand dollars per week in token costs, according to two people familiar with the usage.
From encouragement to brake: how it came to this
The new cap marks a significant turnaround. Over the past six months, Tesla’s leadership had been working to consolidate scattered AI usage company-wide – with approved models, formal safety guidelines, and a central internal platform called “Bottle Rocket,” through which employees can access models from OpenAI, Anthropic, xAI, and Cursor. To boost usage, individual teams even introduced internal leaderboards that ranked employees based on their token consumption. This encouragement apparently worked too well: costs rose so sharply that Tesla is now counteracting.
The notable exception: xAI beta products
Particularly striking about the new regulation is an exception: Spending on beta versions of xAI products does not count towards the $200 cap, according to the memo. xAI is the AI company also led by Tesla CEO Elon Musk, which develops the chatbot Grok and the programming tool “Composer,” among others. Several media outlets interpret this exception as an attempt to steer employees specifically towards Musk’s own AI ecosystem while restricting the use of competitors’ tools.
Not an isolated case: Other tech companies are also pulling the cost brake
Tesla is not alone in this step. According to a report by Inc.com, Uber had already exhausted its entire AI budget for programming tools intended for 2026 within four months and has since limited employees to $1,500 per month per tool for so-called agentic programming tools like Cursor or Claude Code. Uber COO Andrew Macdonald spoke unusually openly about the actual effect: The connection between AI spending and measurable benefits for the company is “not yet apparent,” he said, according to a report by TheStreet. Similar restrictions or guidelines for using cheaper models have now also been introduced at Meta, Amazon, and Walmart, according to several media reports.
Capital expenditures remain untouched
The cost brake on ongoing tool expenses stands in stark contrast to Tesla’s other investment plans: When presenting quarterly figures in April, the company raised its investment forecast for 2026 to over $25 billion, mostly for the expansion of data centers and robotics. The restriction specifically targets ongoing, usage-based billed expenses for external AI tools – not the company’s fundamental willingness to invest in AI infrastructure.
Context
The Tesla case fits into a pattern that can be observed industry-wide in 2026: Companies that initially actively encouraged their workforce to use AI are now encountering the downside of usage-based pricing models, where each individual request incurs immediate costs. Notably, Tesla’s connection of cost discipline and corporate structure remains significant: While the use of external AI providers is regulated, beta offerings from the company’s own AI firm xAI are exempt from the cap.
