Meta’s Adam Mosseri predicts per-engineer AI token spending caps within two years

In an interview on Lenny’s Podcast, Meta’s Instagram head Adam Mosseri said he expects companies to impose per-engineer caps on AI token spending within one to two years. He argued that as the cost of AI processing approaches the total compensation of strong engineers, firms will need to budget token use like any other constrained resource—comparing it to allocations for GPUs, labeling budgets, and payroll.

Mosseri did not state that Meta currently enforces token caps for employees, but he described token budgets as inevitable once costs rise to parity with engineer salaries. He suggested that caps would be proportional to an engineer’s demonstrated ability to use tokens in an ROI-positive manner.

Meta has already taken steps to curb AI spending by removing low-value experiments, including shutting down an internal leaderboard that tracked token consumption. According to Mosseri, such measures reduced wasteful spending without eliminating productive use of AI tools.

He also predicted that token costs will eventually decline as model providers enter pricing wars to attract users, but emphasized that near-term management of token budgets will be necessary for cost discipline.

Mosseri’s remarks follow reports of other companies reassessing AI spending. Uber reportedly exceeded its 2026 AI coding budget by April, and Microsoft canceled Claude Code licenses in favor of consolidating engineers around its own Copilot CLI tool amid soaring token costs.