Meta cuts 8,000 jobs while lifting 2026 AI capex to $135-145B

Meta is shrinking headcount and growing infrastructure at once. The company eliminated about 8,000 roles, roughly 10% of staff, and froze 6,000 planned hires. Reporting says support functions and teams outside AI absorbed most of the cuts. At the same time, AI capex for 2026 is set at $135-145B, nearly double the $72B spent in 2025.

The reallocation reflects how Meta now funds its AI push: savings in operating costs offset some of the massive capital outlay for data centers and accelerators needed to train Muse models. In regulatory filings Meta warned that rising civic and environmental opposition to data center construction, siting delays and power-grid constraints pose real risks to matching competitors' capacity, and that each new frontier model needs far more compute than the last.

Meta is not alone in this pattern. Microsoft reports a $677B commercial backlog with demand outrunning supply, and SpaceX filed plans for a $24B Starbase complex with its own gas plant for xAI. Compute is the binding constraint across the industry, and labor budgets are where companies find cash.

The filings come the same day Meta shipped Muse Wonder and Muse Ember, so investors can weigh spending against model output. Watch for detail on which product teams were cut and whether further data center projects slip because of local opposition.