Adam Mosseri, head of Instagram at Meta, told Lenny’s Podcast that he can imagine a scenario — possibly within a year or two — in which Meta will need to cap employees’ AI token spending.
Mosseri illustrated the concern with an example: “I think that you can imagine, at least in a year or two … that the burn rate of a strong engineer might be the same as their salary, or their cost of employment. And in that world, you’re going to probably need to put in some caps.”
What are AI tokens?
References to AI token spend denote the costs associated with using AI models — essentially fees for processing prompts and generating responses. The subject has gained attention because token-driven usage can translate into substantial operational expenses.
Steps Meta has taken
Meta shut down an internal AI token-spend leaderboard after rising AI costs pushed the company toward projections of billions of dollars in AI expenses for 2026. Mosseri said Meta currently does not enforce per-employee token caps, but he thinks introducing such limits could be healthy in the future.
Industry-wide reactions
Meta is not alone in reassessing AI experimentation budgets. Uber reportedly exhausted its 2026 AI coding budget by April. Microsoft cancelled certain Claude Code licenses and consolidated engineers onto its own Copilot CLI tool in response to soaring token costs.
Mosseri’s rationale and outlook
Mosseri argued that token budgets should be treated like any other resource — similar to GPUs, CPUs, storage, RAM, operational expenditure (OpEx) for labeling, or payroll. He added that any per-engineer cap should be proportional to the company’s trust that the engineer will use the budget in an ROI-positive way.
Looking further ahead, Mosseri expects token prices to decline as AI model providers engage in pricing competition to attract users to their tools over competitors.
Short-term measures
In the short term, Meta has curtailed low-value uses that primarily burned tokens — Mosseri referred to some activities as “silly things,” pointing to the token leaderboard as an example. “It’s not that hard to build a token incinerator, and that doesn’t create a lot of value,” he said.
Summary
Mosseri’s comments indicate that large tech firms may soon treat AI token spend as a standard operational decision comparable to payroll or OpEx. If token burn rates approach the cost of employment for engineers, companies like Meta may impose per-engineer caps to manage budgetary risk.
