Tesla caps its employees' AI spending at $200/week, but leaves out Musk's Grok

Tesla has told its employees internally that, starting July 6, it will impose a $200 weekly cap on spending on AI tools, according to an internal memo cited by The Information. Curiously, this restriction comes just months after Tesla's own management aggressively pushed its workforce to use more AI, even creating internal dashboards that ranked employees by their token consumption to encourage usage.

According to the Electrek article, that incentive worked too well: some software engineers ended up consuming "thousands of dollars" in tokens each week. Faced with runaway spending, Tesla has decided to slam on the brakes, now requiring express authorization to exceed $200 per week per employee.

The most telling detail, and the one at the center of the outlet's criticism, is the exception included in the policy: the cap does not apply to beta versions of products from xAI, Elon Musk's artificial intelligence company. In practice, this pushes Tesla employees to use Grok and Composer (Cursor's coding model) instead of competitors' tools, while alternatives from other providers are subject to the spending limit.

The article recalls that Musk has spent months trying to get Tesla staff to adopt tools tied to his web of companies: after his AI lab's approach to Cursor in April, he sent a company-wide email encouraging people to try Composer. In addition, SpaceX is about to acquire Anysphere, Cursor's parent company, in a deal valued at $60 billion and expected to close this quarter. Tesla engineers also took part as early testers of unreleased versions of Grok and Composer, with direct participation from xAI product leads in internal feedback channels.

However, according to sources cited by Electrek, the maneuver isn't working: Grok is not popular among Tesla staff, who mostly prefer to use Anthropic's Claude. This connects with already-known precedents: the integration of Grok in Tesla's cars didn't even interact with the vehicle's functions, and Musk himself admitted that xAI "was not well built" just weeks after Tesla invested $2 billion in the company.

The backdrop to this story is that Tesla's valuation narrative increasingly depends on AI: Musk has said the company's future value will hinge on the large-scale deployment of AI across its Robotaxi network and the Optimus humanoid robot, rather than on car sales, in a context of revenue that has stagnated over the past two years. Tesla has also launched Nova, an AI tool trained on internal data to standardize processes, and is integrating AI into engineering and quality control on its production lines, alongside a tightening of security policies that restricts the use of unapproved external models on corporate devices.

Electrek frames this episode within a broader pattern in the sector: Uber, Meta, Amazon and Walmart have also imposed AI spending limits after seeing their costs soar under the token-based billing model. What's striking about the Tesla case is the speed of the about-face, going from encouraging consumption through gamification to cutting it drastically within a matter of months.

The outlet's final analysis is critical: if Tesla can't rein in spending of just a few thousand dollars per week per engineer on AI tools, it's legitimate to question its ability to scale this technology to a massive fleet of Robotaxis and millions of Optimus units. Furthermore, it interprets the xAI exception as a way to forcibly gain internal market share for Musk's products, rather than having them prevail on their own merits, in a context where its own engineers prefer to use the competition.