Meta is pouring money into the physical and technical infrastructure needed to scale its AI ambitions. The company said Wednesday in its second-quarter earnings report that it plans to more than double its spend on building AI infrastructure, like data centers and servers. 

“We currently expect 2025 capital expenditures, including principal payments on finance leases, to be in the range of $66-72 billion…up approximately $30 billion year-over-year at the midpoint,” Meta said. 

That’s an aggressive capex growth, and one that Meta plans to continue onwards to 2026. The company said it expects a similarly large increase in spend on AI infrastructure next year as the company continues to “aggressively [pursue] opportunities to bring additional capacity online to meet the needs of [its] artificial intelligence efforts and business operations.”

“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,” said Susan Li, Meta CFO, during the company’s Wednesday earnings call.

Li also noted that while Meta expects to finance most of its AI spend on its own, the company is exploring ways to work with financial partners to co-develop data centers.

“We don’t have any finalized transactions to announce, but we generally believe that there will be models here that will attract significant external financing to support large-scale data center projects that are developed using our ability to build world-class infrastructure, while providing us with flexibility should our infrastructure requirements change over time,” Li said.

Meta has announced two major AI “titan clusters.” The first is Prometheus in Ohio, which is poised to be among the first AI superclusters to hit 1 gigawatt of compute power when it comes online in 2026. Then there’s Hyperion, a cluster in Louisiana that Meta CEO Mark Zuckerberg has bragged would have a footprint the size of Manhattan and could scale up to 5 gigawatts over several years. On top of those, Meta has several other unnamed titan-scale clusters underway. 

Meta’s data center projects promise to soak up enough energy to power millions of homes, pulling that electricity from nearby communities. One of the company’s projects in Newton County, Georgia, has already caused the water taps to run dry in some residents’ homes.

Meta also noted in its earnings report that it expects its second-largest driver of growth to be employee compensation as the company spends millions, and possibly even billions, to poach talented AI engineers and researchers to work for Meta’s newly formed business unit, Superintelligence Labs. 

Before earnings, Zuckerberg shared his vision for “personal superintelligence,” the idea that AI should help individual people live their best lives, mainly through the medium of Meta’s smart glasses and virtual reality headsets.

Meta’s stock surged 10% in after-hours as investors responded to Meta’s overall performance in the quarter and better-than-expected outlook for the third quarter. Meta reported revenue of $47.5 billion in the second quarter, with expectations to hit between $47.5 billion and $50.5 billion in Q3. Advertising drove Meta’s revenue gains, fueled by AI tools — like AI-powered translations and video generation — to help advertisers create more meaningful and targeted campaigns.

The company’s Reality Labs segment, however, saw a $4.5 billion loss.

This article was updated to reflect Meta’s additional paths to funding its AI infrastructure builds.

View Bio