Instagram head Adam Mosseri thinks companies will need to cap how much each engineer can spend on AI tools. He says it could happen within a year or two.
“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,” Mosseri said on Lenny’s Podcast. “And in that world, you’re going to probably need to put in some caps.”
This is not a distant concern for a few tech giants. AI token spending is becoming a real cost for any business that uses AI tools regularly. The pricing model for most AI services is pay-per-token, and those costs add up fast when teams use AI throughout the day.
Mosseri compared managing AI costs to any other business resource. “I have to decide how to deploy payroll for headcount across my teams,” he said. Token budgets will work the same way.
The cost of letting AI run free
Meta is not the only company dealing with runaway AI costs. Uber blew through its entire 2026 AI coding budget by April, just four months into the year. The company had to cap employee AI spending after the early overspend.
Microsoft cancelled Claude Code licenses across the company, consolidating its engineers around its own Copilot CLI tool instead. The message from all three companies is consistent: letting everyone use AI tools without limits leads to a surprisingly large bill.
Mosseri acknowledged that Meta contributed to the problem internally. The company shut down an internal AI token spend leaderboard after realising it was encouraging employees to use AI tools competitively rather than productively. “It’s not that hard to build a token incinerator, and that doesn’t create a lot of value,” he said.
How AI pricing works
Every time you ask an AI model a question, you pay for the tokens used in both your prompt and the model’s response. A token is roughly a word or part of a word. A single complex task can use thousands of tokens.
A developer running AI-assisted code completion all day might use millions of tokens per month. When a whole team does this, the monthly bill can reach thousands of dollars per person. Mosseri’s point is that this cost eventually rivals what that person costs in salary and benefits.
The good news is that costs are coming down. Model providers are cutting prices rapidly, and Mosseri expects this to continue as a pricing war develops. But the fundamental problem remains: cheaper tokens do not matter if usage grows even faster.
What businesses should do now
The practical lesson for any business is to start tracking AI costs before they become a problem. Know which teams are using which tools. Set soft limits before you need hard caps.
Treat AI tokens like you would any subscription or utility. Monitor usage monthly. Ask whether the spending is producing results. If a team is burning tokens without clear ROI, that is not a technical problem — it is a management problem.
Mosseri suggested caps should be proportional to trust. Teams that use AI wisely and produce results should get more budget. Teams that waste it on low-value experiments should get less. That is not different from how you manage any other business expense.
Bottom line
AI token costs are real, they are growing, and they need to be managed like any other operating expense. Meta, Uber, and Microsoft are already adapting. Smaller businesses should pay attention now, before the bill becomes a shock.
Set a budget. Track usage. Cut what does not work. The companies that do this well will get the benefits of AI without the financial headache.