Meta’s head of Instagram, Adam Mosseri, thinks the company will need to start capping how much engineers can spend on AI within the next year or two. Speaking on Lenny’s Podcast, he said AI token costs could soon rival what the company pays those same engineers in salary.
‘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,’ Mosseri said.
The comments come at a moment when tech companies are waking up to just how expensive AI experimentation has become. Token spend, which refers to the cost of processing AI prompts and generating responses, has quietly ballooned into a major budget line item for some of the biggest names in tech.
Meta is already dealing with the fallout. The company recently shut down an internal AI token spend leaderboard after it became clear that unchecked usage was putting the company on track to rack up billions of dollars in AI costs in 2026. As reported by TechCrunch, Mosseri acknowledged that some of what was happening simply did not make business sense.
‘It’s not that hard to build a token incinerator, and that doesn’t create a lot of value,’ he said.
Meta is not alone in hitting this wall. A pattern is forming across the industry:
- Uber blew through its entire 2026 AI coding budget by April
- Microsoft cancelled Claude Code licenses for its engineers and moved them onto its own Copilot CLI tool instead
- Meta shut down its token spend leaderboard after costs spiraled out of control
Mosseri’s view is that token budgets should be treated like any other resource the company manages. He compared it to how Meta already allocates GPUs, storage, labeling budgets, and headcount across different teams. Token spend, he argues, is simply the next thing that needs the same kind of oversight.
‘I think of it like any other resource,’ he said. ‘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. 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.’
Right now, Meta does not have token caps for any employee. But Mosseri expects that to change, with any future limits tied to how much the company trusts a given engineer to spend the budget in a way that actually produces results. The idea is not to punish experimentation, but to make sure AI usage is generating real value rather than just burning through money.
Longer term, Mosseri expects token costs to come down as AI model providers compete harder for customers and start cutting prices. That pricing pressure could eventually make the current concerns less urgent. But in the short term, companies like Meta are learning that giving engineers unlimited access to AI tools without any guardrails is a fast way to blow a budget, and the industry is now figuring out how to manage that responsibly.
