Adam Mosseri, the head of Instagram at Meta, has made a bold prediction about the future of artificial intelligence spending in tech companies. He believes that companies will soon need to manage AI token budgets the same way they handle payroll or other operating expenses. This could mean engineers will face limits on how much they can spend using AI tools.
Mosseri's Prediction on AI Token Budgets
Speaking on the Lenny Rachitsky podcast, Mosseri explained that managing AI resource allocation is becoming a standard part of company budgets. According to HyperAI, Mosseri emphasized that companies will eventually need to cap AI token spending per engineer, treating it like any other operational cost.
Mosseri described some AI projects as "token incinerators" — initiatives that waste processing tokens without delivering real value. He said Meta has already cut its AI spending by shutting down these inefficient projects.
Meta's Approach to Cutting AI Costs
Mosseri revealed that Instagram has successfully reduced its artificial intelligence spending by discontinuing projects that were unnecessarily consuming processing tokens. He described these as "silly things" that helped burn through tokens without meaningful results.
According to Threads, Mosseri also dismissed the idea of token-consumption leaderboards, calling them a "terrible idea." Instead, he believes companies should focus on efficient use of AI resources.
What This Means for Engineers
The prediction suggests that engineers working with AI tools may soon have to work within strict budgets for token usage. This is a shift from the current environment where many developers have relatively free access to AI models for experimentation and development.
Mosseri expects AI costs to remain volatile as usage rises, but he also noted that pricing for frontier AI models is becoming more competitive. This could help balance some of the cost pressures companies face.
Our Take: A Necessary Reality Check
In our view, Mosseri's prediction is a realistic and necessary step for the tech industry. AI tools are powerful, but they are not free. Treating token spending like any other business expense — with budgets, limits, and accountability — is common sense. The era of unlimited AI experimentation without oversight was never sustainable. Companies that implement smart token budgets early will likely have a competitive advantage. The key is to balance innovation with cost control, not to stifle creativity with rigid caps.
