Tokenminimizing: Meta Moves to Curb Employee AI Usage as AI Costs Reach Billions

Meta Platforms plans to clamp down on skyrocketing AI costs inside the company by imposing limits on employees’ token usage, the company told staff in a memo on Tuesday, just weeks after it pushed them to adopt AI tools in their work.

The company is building an internal platform to track employee AI usage and spending in real time, set budgets and establish limits on employees’ token spend, according to an internal memo reviewed by The Information, which Meta shared with about 6,000 staffers earlier this week. The effort is part of a broader efficiency program aimed at cutting costs.

“We’ve seen an exponential increase in AI usage and [we] are tracking to spend billions on internal use alone in 2026,” the memo said. “At the same time, individuals and teams have limited visibility into and control over how they use AI. In 2027, we expect Meta will move toward managing AI tokens in a more structured way—with budgets, allocation decisions, and supporting tools.”

As part of the spending crackdown, a team of product developers and engineers has created a central dashboard called AI Gateway to monitor usage and spending in one place. According to the memo, Meta will also roll out automated alerts for unusual spending spikes. The group is also keeping track of current costs to forecast future spending so it can plan computing capacity and negotiate with vendors. The company plans to announce the new controls and tools to a broader set of employees in the coming weeks.

Meta is also preparing to encourage staff to shift away from third-party AI tools for things like coding toward more in-house solutions, such as its internally developed coding assistant, MetaCode (previously called Devmate), according to the memo. At the same time, the company also intends to let employees continue to have access to new third-party models.

A Meta spokesperson said, in a statement previously shared with The Information: “It’s well known that this is a priority, and we’re focused on using AI to help employees with their day-to-day work.”

In recent weeks, the company has tasked engineers in its newly formed division, Applied AI Engineering, to improve MetaCode in order to reduce its reliance on external tools like Anthropic’s Claude, which engineers use heavily for coding work, according to two people familiar with the efforts. Leaders in the AAI group have instructed engineers to create high-quality reinforcement-learning data by generating programming challenges for Metacode to solve, training it to respond better, the people said.

A Reversal

Meta’s efforts to curb AI spending come after months of aggressively encouraging employees to adopt AI tools across the company. The company has given staff access to its own internally developed models, as well as tools from OpenAI, Anthropic and Google. In November, Meta told employees that demonstrating “AI-driven impact” would become a “core expectation” this year, with top performers rewarded for delivering the greatest results.

That push to embed AI more deeply into daily work has in some cases produced unexpected results. Earlier in the spring, some employees began competing in a practice known as tokenmaxxing to signal their status as AI power users. For a time, workers attempted to climb an internal leaderboard called Claudeonomics, which ranked the top 250 employees based on AI token usage.

Some staffers even attempted to inflate their usage by instructing AI agents to run multiple tasks at once and maximize token consumption. A copy of the dashboard reviewed by The Information in April showed that employees had burned through 60.2 trillion tokens in a 30-day period. Usage later climbed to 73.7 trillion tokens before the leaderboard was taken down.

Meta’s chief technology officer, Andrew Bosworth, has sought to discourage wasteful AI token usage. In an April memo, Bosworth told employees to use AI tools only when they meaningfully improve productivity.

“Nobody should be using AI tools just for the sake of using them,” he wrote. “All motion is not progress and token usage alone is not a measure of impact of any kind. We will be using these tools because they’ll genuinely allow us to do better work, faster.”

In the same memo, Bosworth also revealed that Meta was reorganizing internal work around AI agents through an initiative called the Agent Transformation Accelerator. He said its current AI tools had become fragmented and the company is now working toward a more unified approach.

Meta isn’t the only business seeking to impose limits on employee AI usage to bring down costs. Uber and ServiceNow used up their entire annual budgets for Anthropic’s tools in just the first few months of 2026. ServiceNow is also monitoring employees’ daily usage in a bid to track and reduce its costs. Venture capitalists are also capping their own employees’ AI usage after finding they were racking up thousands of dollars a day in token costs.

Meta’s move to reduce spending on internal AI tools comes as the company prepares to invest as much as $145 billion this year, in part to expand its AI infrastructure. The spending is expected to support the development of data centers, AI chips and talent acquisition.

At the same time, Meta faces growing pressure from investors to generate returns from its substantial AI investments, and the company has started to find ways to grow its revenue beyond advertising. As part of that effort, it has introduced paid subscription tiers across Facebook, Instagram and WhatsApp, while also signaling it plans to charge businesses that use its AI business agents.