Meta's Mosseri: AI token costs may soon rival engineer salaries
Curated by the Inblix editorial team
Instagram chief Adam Mosseri dropped a sobering prediction on Lenny’s Podcast: within a year or two, an engineer’s AI token burn rate could match their salary. That’s not hyperbole from an alarmist. It’s the head of one of the world’s largest social platforms saying out loud what many tech leaders are quietly realizing — the bill for AI-assisted coding is coming due, and it’s enormous.
The context here is a quiet panic spreading through big tech. Meta recently killed an internal leaderboard that gamified AI token usage after realizing it was on a trajectory to spend billions in 2026. Uber blew through its entire 2026 AI coding budget by April. Microsoft pulled the plug on Claude Code licenses, herding engineers toward its homegrown Copilot CLI. These aren’t isolated incidents. They’re early warning signs of a resource that’s been treated like an all-you-can-eat buffet suddenly looking more like a prix fixe menu.
Mosseri framed token costs as just another operational resource to be rationed — no different from GPU clusters, labeling budgets, or headcount. His exact words: “I think of it like…any other resource.” The logical endpoint is token caps, doled out based on how much the company trusts an engineer to generate a positive return on that spend. It’s a startling concept. We’re moving from unlimited AI access as a productivity perk to AI as a managed line item, with managers deciding who gets how much compute. The phrase “token incinerator” — Mosseri’s term for tools that burn cash without creating value — might become part of the management lexicon.
He does see relief on the horizon, betting that model providers will eventually engage in a pricing war that drives costs down. But that’s a future state. Right now, the industry is waking up to the uncomfortable math that pairing every developer with an omnipotent AI assistant has a price tag that can spiral shockingly fast. The leaderboard shutdown was a small fix. The real reckoning — actual per-engineer budgets — hasn’t even started yet.
💡 Key Takeaways
- Meta's Instagram head predicts per-engineer AI token costs could equal salary expenses within two years, forcing hard conversations about usage caps.
- The company recently axed an internal token spend leaderboard that was accelerating a path toward billions in AI costs by 2026.
- Uber and Microsoft have already made similar moves, with Uber exhausting its 2026 coding AI budget by April and Microsoft canceling third-party AI coding licenses.
- Mosseri believes token budgets will eventually be tied to an engineer's perceived ability to generate positive ROI, turning AI access into a managed resource rather than an open spigot.
Keep reading: See related articles below for more coverage on this topic.
Get smarter about AI
The sharpest AI news, curated daily. Delivered free to your inbox.