Tesla puts the brakes on AI costs, caps staff spending at ₹19,000 weekly, starting July 6

Tesla has instructed employees to rein in spending on artificial intelligence tools, imposing a weekly limit of $200 starting July 6, according to an internal memo first reported by The Information. The move marks a sharp departure from the company's recent push encouraging employees to make extensive use of AI, signalling a broader shift in how businesses are managing the rising costs of generative AI.

Under the new policy, employees seeking to exceed the $200 [ ₹19,046 (at roughly ₹95.37 per USD] weekly cap must obtain managerial approval. Before the restriction was introduced, some software engineers were reportedly spending thousands of dollars each week on AI tokens—the usage-based units that determine the cost of interacting with AI models.

From AI adoption drive to spending restraint

The new limits represent a dramatic reversal. Over the past six months, Tesla had consolidated employees' AI access through an internal platform known as Bottle Rocket, which offered models from OpenAI, Anthropic, xAI and Cursor. Some teams even introduced dashboards tracking token consumption, effectively encouraging employees to increase AI usage.

That strategy appears to have succeeded beyond expectations, with heavy users generating substantial AI expenses. The latest directive now reflects an effort to curb those rapidly escalating costs.

xAI receives a notable exemption

One notable aspect of the policy is that the $200 spending cap does not apply to beta versions of xAI's products, the artificial intelligence company founded by Tesla CEO Elon Musk. The exemption effectively encourages employees requiring higher AI usage to rely on Grok and Cursor's Composer model instead of competing platforms.

Elon Musk has increasingly promoted AI products connected to his other ventures. He previously encouraged Tesla employees to adopt Composer after xAI partnered with Cursor. Separately, SpaceX is reportedly preparing to acquire Cursor's parent company, Anysphere, in a deal valued at about $60 billion.

However, the strategy has reportedly faced resistance internally. According to people familiar with the matter, many Tesla engineers continue to favour Anthropic's Claude over Grok despite the company's efforts to promote its in-house ecosystem.

AI ambitions meet cost realities

The spending restrictions come at a significant moment for Tesla, whose long-term growth narrative is increasingly centred on artificial intelligence. 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, while automotive revenue has remained broadly flat over the past two years.

Against that backdrop, tighter controls over relatively modest AI operating costs could raise broader questions about the economics of deploying AI at scale across autonomous vehicle fleets and large numbers of robots.

Part of a broader industry trend

Tesla's move also reflects a wider shift across the technology sector as companies reassess soaring AI expenses. The industry has begun moving away from measuring employee productivity through maximum AI usage—a trend sometimes referred to as "tokenmaxxing"—towards greater cost discipline.

Several large companies have introduced similar controls. Uber reportedly imposed a monthly AI spending cap of $1,500 after exhausting its entire 2026 AI budget by April. Meta, Amazon and Walmart have also introduced spending limits or directed employees towards lower-cost AI models as token-based pricing has made AI usage expenses increasingly visible.

Alongside tighter cost controls, Tesla has strengthened its AI security policies by limiting access to AI models outside the Bottle Rocket platform on company devices and reminding employees not to upload confidential company information into unapproved AI systems.