Meta Boosts 2025 CapEx to $72B as AI Infrastructure Drives Next Wave of Investment
Meta Platforms reported third-quarter 2025 revenue of $51.2 billion, up 26% year-over-year, while income from operations reached $20.5 billion. Results were overshadowed by an $15.9 billion one-time, non-cash tax charge tied to the new “One Big Beautiful Bill Act,” reducing reported net income to $2.7 billion. Excluding this charge, adjusted net income would have been $18.6 billion, or $7.25 per share. Family of Apps revenue grew to $50.8 billion, offsetting Reality Labs’ $4.4 billion operating loss. Meta ended the quarter with $44.5 billion in cash and marketable securities.
Capital expenditures surged to $19.4 billion for the quarter, bringing 2025 year-to-date spending to $50.1 billion. CFO Susan Li said Meta expects total 2025 CapEx between $70 billion and $72 billion—up from the previous $66–72 billion forecast—and signaled even stronger infrastructure investment ahead. The company cited escalating compute demand from its AI model training, Meta Superintelligence Labs, and next-generation AI glasses as the main drivers. Meta also plans to expand both owned and third-party data center capacity, creating “further upward pressure” on CapEx in 2026.
Meta’s management noted that 2026 will bring “notably larger” CapEx growth, led by AI infrastructure and depreciation costs. Employee compensation—especially for new AI and technical talent—will be the second-largest expense growth factor. While advertising growth remains strong, the company expects lower Reality Labs revenue next quarter as it laps the Quest 3S launch.
“We are at an exciting point for our company, where we have continued runway to improve our core services and the opportunity to build new AI-powered experiences,” said Mark Zuckerberg, Meta founder and CEO.
- On compute capacity strategy: CEO Mark Zuckerberg noted the company is “front-loading” infrastructure buildout to be ready if superintelligence arrives sooner than some expect, and in the meantime use that capacity to accelerate core business.
- The call reiterated that Meta will pursue both own infrastructure build-out and contract third-party cloud providers to meet compute-capacity growth.
🌐 Analysis: Meta is firmly positioning itself as a hyperscale AI infrastructure builder alongside cloud peers like Microsoft, Google, and Amazon. The $70–72 billion CapEx forecast underscores an aggressive buildout of AI datacenter capacity, likely tied to custom silicon, optical interconnects, and large GPU clusters. With Meta Superintelligence Labs now operational, 2026 could see Meta rival top hyperscalers in AI infrastructure scale and power consumption as it expands its “AI factory” footprint.