After Microsoft and Uber, Meta is also dealing with soaring AI bills, caps tokenmaxxing
Meta spent months encouraging employees to use AI for coding, research and daily work. Now, with internal AI costs projected to reach billions of dollars, the company is reportedly introducing spending controls, usage limits and monitoring tools to keep its growing AI bill in check.
AI may be making employees more productive, but it is also becoming a costly affair for big tech. Following reports that Microsoft and Uber are grappling with soaring AI bills, Meta is also now looking for ways to rein in its own spending. According to The Information, the company is preparing to introduce new limits on how employees use AI tools internally.
According to the report, Meta had spent months encouraging employees to embrace "tokenmaxxing", essentially pushing staff to use AI as much as possible across coding, research and day-to-day work. However, that strategy now appears to be getting expensive. Meta reportedly estimates that employee AI usage alone could cost the company billions of dollars in 2026. As a result, the Facebook and Instagram parent is preparing new controls and spending limits to keep those costs from spiralling further.
Meta has reportedly outlined plans to introduce spending controls, budgets and usage limits. In a memo shared with around 6,000 employees earlier this week, the company said it would begin monitoring AI consumption more closely and give teams better visibility into their usage, according to The Information.
At the centre of the effort is Meta's new internal platform, called AI Gateway. According to the report, the system is being built to track AI usage and spending across teams in real time. The platform will include a central dashboard that brings AI activity and costs into one place, allowing teams to understand how many tokens they are consuming and what that usage is costing the company.
Meta is also planning to introduce automated alerts that flag unusual spending spikes. The report notes that employees currently have limited visibility into their AI consumption, making it difficult to understand the financial impact of heavy usage. By 2027, Meta reportedly expects to manage AI resources through a more structured framework that includes token budgets, allocation decisions and dedicated management tools.
The company is also encouraging employees to reduce their reliance on third-party AI tools and instead use internal alternatives. One of those products is MetaCode, Meta's in-house coding assistant, previously known as Devmate. Steering employees towards internal systems could help Meta gain greater control over both costs and usage patterns.
Meta's plans to limit AI usage come at a time when more companies are discovering the hidden costs of widespread AI adoption. While the cost of generating individual AI responses has fallen over time, overall spending continues to rise because usage is growing even faster. The emergence of AI agents and coding assistants has further increased demand for computing resources, with complex tasks consuming far more tokens than a simple chatbot interaction.
Meta is not alone in facing this issue. Reports suggest that Uber exhausted its planned 2026 AI coding budget within just the first four months of the year because of soaring token consumption. Microsoft has also reportedly been exploring ways to optimise internal AI spending as usage expands across its workforce.
AI may be making employees more productive, but it is also becoming a costly affair for big tech. Following reports that Microsoft and Uber are grappling with soaring AI bills, Meta is also now looking for ways to rein in its own spending. According to The Information, the company is preparing to introduce new limits on how employees use AI tools internally.
According to the report, Meta had spent months encouraging employees to embrace "tokenmaxxing", essentially pushing staff to use AI as much as possible across coding, research and day-to-day work. However, that strategy now appears to be getting expensive. Meta reportedly estimates that employee AI usage alone could cost the company billions of dollars in 2026. As a result, the Facebook and Instagram parent is preparing new controls and spending limits to keep those costs from spiralling further.
Meta has reportedly outlined plans to introduce spending controls, budgets and usage limits. In a memo shared with around 6,000 employees earlier this week, the company said it would begin monitoring AI consumption more closely and give teams better visibility into their usage, according to The Information.
At the centre of the effort is Meta's new internal platform, called AI Gateway. According to the report, the system is being built to track AI usage and spending across teams in real time. The platform will include a central dashboard that brings AI activity and costs into one place, allowing teams to understand how many tokens they are consuming and what that usage is costing the company.
Meta is also planning to introduce automated alerts that flag unusual spending spikes. The report notes that employees currently have limited visibility into their AI consumption, making it difficult to understand the financial impact of heavy usage. By 2027, Meta reportedly expects to manage AI resources through a more structured framework that includes token budgets, allocation decisions and dedicated management tools.
The company is also encouraging employees to reduce their reliance on third-party AI tools and instead use internal alternatives. One of those products is MetaCode, Meta's in-house coding assistant, previously known as Devmate. Steering employees towards internal systems could help Meta gain greater control over both costs and usage patterns.
Meta's plans to limit AI usage come at a time when more companies are discovering the hidden costs of widespread AI adoption. While the cost of generating individual AI responses has fallen over time, overall spending continues to rise because usage is growing even faster. The emergence of AI agents and coding assistants has further increased demand for computing resources, with complex tasks consuming far more tokens than a simple chatbot interaction.
Meta is not alone in facing this issue. Reports suggest that Uber exhausted its planned 2026 AI coding budget within just the first four months of the year because of soaring token consumption. Microsoft has also reportedly been exploring ways to optimise internal AI spending as usage expands across its workforce.
