Instagram chief Adam Mosseri predicts that within a year or two, companies will need to treat AI token spend like payroll, imposing caps on engineers. He warns that unchecked token burn could equal an employee's salary, forcing budget limits.
Key Takeaways
- Meta is considering caps on AI token usage to control costs.
- Engineers' token burn could match their salary, making limits necessary.
- Future AI pricing wars may lower token costs, but ROI‑focused governance will be essential.
Instagram CEO Adam Mosseri told listeners of Lenny’s Podcast that within the next one to two years, Meta may need to impose strict caps on how much AI token budget each employee can consume. He explained that as AI tools become more powerful, the “burn rate” of a high‑performing engineer could soon equal the cost of their salary or total employment expense.
Understanding AI Token Spend
AI token spend refers to the monetary cost of processing prompts and receiving responses from large language models. The issue has surged into the spotlight after Meta shut down an internal token‑spend leaderboard, fearing that unchecked usage could push its AI expenses into the billions by 2026.
Industry‑wide Reassessment
Meta is not alone in confronting soaring AI costs. Uber blew through its 2026 AI coding budget by early April, prompting a strategic reset. Microsoft, facing similar pressures, cancelled its Claude Code licenses and consolidated its engineering teams around its own Copilot CLI tool. These moves illustrate a broader trend: tech giants are now treating AI experimentation as a line‑item expense rather than a free‑for‑all sandbox.
Mosseri’s Analogy: AI as Operating Expenditure
Drawing a parallel with traditional operating expenditures (OpEx), Mosseri said, “I think of it like any other resource—limited GPUs, CPUs, storage, RAM. I must allocate OpEx for labeling budgets across teams, and likewise allocate payroll for headcount.” In his view, AI token budgets will become another resource that needs disciplined distribution based on ROI‑positive usage.
Potential Caps and Future Outlook
Currently, Meta does not enforce token caps for any employee, but Mosseri believes caps could be “healthy” once the company establishes clear ROI metrics. He also anticipates that as AI model providers enter a pricing war to attract developers, token costs will gradually decline, easing the pressure on corporate budgets.
For now, Meta has curbed some wasteful spending by shutting down “silly things” like the token spend leaderboard. “It’s not that hard to build a token incinerator, but it doesn’t create much value,” Mosseri remarked, underscoring the need for purposeful AI investment.