Archive — US — techcrunch
Meta’s Adam Mosseri says AI token budgets could soon be capped per engineer
Meta's Adam Mosseri suggests AI token budgets for engineers could soon be capped to control rising costs, similar to payroll. With companies like Uber and Microsoft facing overruns, he emphasizes managing AI resources as they do with other operational expenses, anticipating lower future token costs as competition increases.
Key points
- Mosseri anticipates the need for caps on AI token spending within 1-2 years due to rising costs, equating them to salaries.
- Meta's internal AI spending led to a planned budget overrunning into billions by 2026, prompting a reevaluation of expenditures.
- Other companies like Uber and Microsoft also faced AI budget overruns, indicating a broader industry issue.
- Mosseri sees AI token costs needing management like other resources (GPUs, payroll) to ensure ROI-positive outcomes.
- Current token spending caps do not exist, but he believes they will be beneficial in the future as AI pricing becomes more competitive.
Why it matters
- The shift in AI budgeting reflects a broader trend in tech companies reevaluating resource allocation amidst rising costs.
- Historically, tech firms have encountered similar challenges with new technologies leading to unsustainable spending.
- As of 2025, controlling operational costs, particularly in AI, continues to be crucial amid innovations and competitive pressures.
Mosseri's insights highlight the necessity for practical AI resource management as firms face escalating costs, aligning budgeting practices for AI with traditional financial strategies to ensure sustainable growth.