Tesla caps AI spending at $200/week per employee, starting July 6

Tesla told employees it will cap weekly spending on artificial intelligence tools at $200 per person, starting July 6, according to an internal memo first reported by The Information. The policy reverses the company's earlier push to get staff to use more AI, and it marks one of the most concrete signs yet that the cost of running generative AI models is forcing companies to set hard limits.

Employees who need to spend more than $200 a week on AI tokens – the usage-based units that determine how much it costs to query a model – must get a manager to approve the overage. Before the cap, some software engineers were spending thousands of dollars a week on tokens, the memo showed.

The restriction is a sharp turn from Tesla's stance over the past six months. The company had consolidated AI access through an internal platform called Bottle Rocket, which gave employees access to models from OpenAI, Anthropic, xAI and Cursor. Some teams even built dashboards that tracked token consumption, effectively encouraging staff to use more AI.

That push worked better than expected. Heavy users ran up big bills, and the new cap is an attempt to bring those costs under control.

One carveout stands out. The $200 cap does not apply to beta versions of xAI's products, the AI company founded by Tesla CEO Elon Musk. That exemption nudges employees who need more AI capacity toward Grok and Cursor's Composer model instead of competing platforms.

Musk has been steering Tesla toward his own AI ventures. He previously encouraged Tesla employees to use Composer after xAI partnered with Cursor. Separately, SpaceX is in talks to buy Cursor's parent company, Anysphere, in a deal that could value it at roughly $60 billion, people familiar with the matter said.

The strategy has hit internal resistance. Many Tesla engineers still prefer Anthropic's Claude over Grok, according to people familiar with the matter, despite the company's efforts to push its own ecosystem.

The spending cap arrives at a sensitive moment for Tesla. Musk has repeatedly argued that the company's future depends more on autonomous Robotaxis and the Optimus humanoid robot than on its traditional electric vehicle business. Automotive revenue has been roughly flat for two years. Tightening controls on AI operating costs – even relatively small ones – raises questions about the economics of running AI at scale across a fleet of autonomous vehicles and humanoid robots.

Tesla is not alone. The broader tech industry is moving away from the "tokenmaxxing" era, where employees were measured by how much AI they used, toward stricter cost discipline. Uber imposed a monthly AI spending cap of $1,500 after burning through its entire 2026 AI budget by April, according to a person familiar with the matter. Meta, Amazon and Walmart have also introduced limits or directed employees toward cheaper AI models as token-based pricing has made AI expenses harder to ignore.

Alongside the spending cap, Tesla has tightened its AI security policies. The company now restricts access to AI models outside the Bottle Rocket platform on company devices and reminded employees not to upload confidential company information into unapproved AI systems.