Meta’s Adam Mosseri Warns AI Token Budgets May Be Capped Per Engineer Within Two Years

Instagram head Adam Mosseri has predicted that Meta may need to impose caps on AI token spending per engineer within the next one to two years, as the company faces ballooning costs from internal AI experimentation. Speaking on Lenny’s Podcast, Mosseri said the burn rate of a strong engineer could soon rival their salary, forcing the company to treat AI tokens like any other finite resource — such as payroll or operating expenses.

Why AI Token Budgets Are Becoming a Concern

Mosseri explained that the cost of processing AI prompts and responses — measured in tokens — has become a significant operational expense for Meta. The company recently shut down an internal AI token spend leaderboard after realizing its AI costs were on track to hit billions of dollars by 2026. This move came as part of a broader effort to rein in what Mosseri described as “silly things” that burn tokens without creating value. “It’s not that hard to build a token incinerator, and that doesn’t create a lot of value,” he said.

Industry-Wide Trend: AI Cost Reckoning

Meta is not alone in reassessing its AI spending. Uber reportedly blew through its 2026 AI coding budget by April of this year, prompting an internal review. Microsoft canceled Claude Code licenses and consolidated engineers around its own Copilot CLI tool after soaring token costs. These examples highlight a growing industry challenge: balancing the promise of AI-assisted productivity with the reality of rapidly escalating compute costs.

How Token Caps Would Work

Mosseri drew a direct analogy between token budgets and other operational resources. “I have to decide how to deploy capacity to my different teams because I have a limited number of GPUs and CPUs and storage and RAM,” he said. “I have to decide how to deploy OpEx for labeling budgets across my teams. I have to decide how to deploy payroll for headcount across my teams. Token budgets will be the same.” He added that any future cap per engineer would be proportional to the company’s trust in that engineer’s ability to use the budget in an “ROI-positive” way. Currently, Meta does not have token caps for any employee, but Mosseri believes their use could become healthy as costs grow.

Conclusion

Mosseri’s comments signal that even the largest tech companies are struggling to manage the financial realities of widespread AI adoption. While he expects token costs to eventually decline as AI model makers enter a pricing war, the immediate future may require tighter controls. For now, Meta has managed to reduce its token burn by eliminating wasteful projects — but the era of unlimited AI experimentation may be drawing to a close.

FAQs

Q1: What is an AI token budget?
A token budget refers to the allocated cost or limit on the number of AI prompts and responses an employee or team can generate. Each interaction with an AI model consumes tokens, which have a real cost based on compute resources.

Q2: Why is Meta considering token caps now?
Meta’s internal AI costs were on track to reach billions of dollars by 2026, driven by uncontrolled experimentation. Shutting down a token spend leaderboard and curbing wasteful projects were initial steps, but Mosseri sees caps as a necessary future measure to prevent runaway spending.

Q3: How do other companies handle AI token costs?
Uber exhausted its 2026 AI coding budget by April, while Microsoft canceled some third-party AI tool licenses and consolidated around its own Copilot CLI to control costs. These moves reflect a broader industry trend toward stricter AI resource management.