Meta is doubling down on AI infrastructure spending. The company plans to more than double its AI-related capital expenditures in 2025, aiming to pump $66-72 billion into data centers, servers, and tech upgrades. That’s a jump of about $30 billion compared to last year.
Meta says this big ramp-up will continue into 2026 to meet AI demands and power its business. CFO Susan Li called AI infrastructure “a core advantage” for better models and products.
Susan Li said,
“We expect that developing leading AI infrastructure will be a core advantage in developing the best AI models and product experiences, so we expect to ramp our investments significantly in 2026 to support that work.”
Meta might team up with financial partners to co-develop some of these massive data centers but hasn’t sealed any deals yet.
The company’s AI “titan clusters” are massive undertakings. Prometheus in Ohio is set to hit 1 gigawatt of compute power in 2026. Hyperion in Louisiana could scale to 5 gigawatts — that’s a datacenter footprint as big as Manhattan. More unnamed titan-scale clusters are in the works.
These projects are giant energy hogs that strain local resources. A data center in Newton County, Georgia, already drained residents’ water supplies.
Employee pay is Meta’s next big cost spike. The company is shelling out millions — maybe billions — to hire top AI engineers for its new Superintelligence Labs.
Zuckerberg pitched “personal superintelligence” ahead of earnings — AI tools that help people via smart glasses and VR headsets.
Meta’s Q2 results boosted its stock 10% after hours. Revenue hit $47.5 billion, with Q3 expected between $47.5 and $50.5 billion. Ad revenue gains leaned on AI-driven tools for translations and video creation.
Reality Labs still bleeding — it lost $4.5 billion in the quarter.
This article was updated to note Meta’s new options for funding AI infrastructure beyond self-financing.
