Meta is trying to do two things at once with artificial intelligence: build more and spend less.
According to Reuters, Meta's newly formed AI team delivered its first key models internally this month, the company's chief technology officer said. That marks an early milestone for a reorganized effort as Meta pushes to develop its own advanced AI systems in-house.
At the same time, senior Meta leaders are talking openly about reining in the cost of that work. In comments reported by MSN, Instagram head Adam Mosseri joined a growing debate over rising AI expenses, saying the company shut down what he called "silly things" — wasteful early experiments that were driving up costs without delivering enough value.
Mosseri also pushed back on one common industry habit. Per MSN, he described the practice of tracking massive amounts of AI data consumption as a terrible idea, suggesting that chasing raw data volume is the wrong way to measure progress or justify spending.
Together, the two threads paint a picture of a company moving from an experimental, spend-freely phase into a more disciplined one. Meta wants working models from its new team, but it also wants to trim the kind of open-ended projects that ballooned budgets as the AI race heated up.
Why it matters: How a giant like Meta balances ambitious AI development against ballooning costs will shape which products get built — and signals whether the broader industry's expensive AI spending spree is starting to face harder financial limits.