About News Writing Resources Contact
All Stories

Meta Commits $115–$135B to AI Capex in 2026, Nearly Doubling Last Year

Meta confirmed capital expenditures of $115–$135B for 2026, roughly 2x 2025. The company released Muse Spark, a smaller multimodal model it claims matches Llama 4 at ~10% the compute, and accelerated data-center buildout to support agentic workloads across WhatsApp, Instagram, and Ray-Ban Meta devices.

Meta's pitch is "we are the only frontier lab that already owns 3 billion distribution endpoints," and the capex number is the proof-of-seriousness. The real question for every non-FAANG executive is whether you are a customer of this infrastructure or a competitor for its attention — there is no middle ground anymore. If your company's AI roadmap still depends on a startup's API quota, you are underwriting someone else's distribution problem; the winning move this quarter is locking multi-year compute commitments before the next capacity crunch in Q3.
Read Original Source