according to a recent report by the information, meta is comprehensively tightening its internal policies on the use of ai resources to address mounting cost pressures stemming from surging ai computing demands. data shows that in just the past month, the total volume of ai tokens consumed by its employees has exceeded 60 trillion, and ai-related expenditures are projected to soar into the tens of billions of dollars by 2026—already a significant financial burden even for a tech giant like meta.
this shift stands in stark contrast to meta’s earlier strategy: back then, the company vigorously promoted an “ai-first” culture, providing employees with various generative ai tools and incorporating ai usage frequency and token consumption into performance evaluations for certain roles. such incentive structures created gray‑area operational practices; platforms like reddit even hosted tutorial posts teaching employees how to rapidly inflate their ai call counts to meet assessment targets.
now, meta has launched a systematic cost‑control initiative. key measures include setting hard caps on ai budgets for each team, implementing token‑quota systems, prioritizing the adoption of in‑house tools such as meta code to improve resource efficiency, and accelerating the development of a unified ai‑usage monitoring platform. this platform will roll out real-time tracking capabilities by 2026, initially enabling visibility into usage patterns and threshold‑based alerts; by 2027, it will evolve into a structured governance framework covering budget allocation, cost attribution, and tool coordination.
notably, meta is far from alone. leading tech firms including uber, microsoft, and salesforce have recently introduced similar restrictive policies, signaling the industry’s transition from “uncontrolled expansion” to “lean operations.” as ai investments no longer automatically translate into efficiency gains but instead manifest primarily as quantifiable costs, rational constraints have become essential for the sustainable deployment of ai technologies.
