Meta is pouring cash into the bodily and technical infrastructure wanted to scale its AI ambitions. The corporate mentioned Wednesday in its second-quarter earnings report that it plans to greater than double its spend on constructing AI infrastructure, like knowledge facilities and servers.
“We at present anticipate 2025 capital expenditures, together with principal funds on finance leases, to be within the vary of $66-72 billion…up roughly $30 billion year-over-year on the midpoint,” Meta mentioned.
That’s an aggressive capex progress, and one which Meta plans to proceed onwards to 2026. The corporate mentioned it expects a equally massive improve in spend on AI infrastructure subsequent 12 months as the corporate continues to “aggressively [pursue] alternatives to deliver further capability on-line to fulfill the wants of [its] synthetic intelligence efforts and enterprise operations.”
“We anticipate that growing main AI infrastructure can be a core benefit in growing the perfect AI fashions and product experiences, so we anticipate to ramp our investments considerably in 2026 to assist that work,” mentioned Susan Li, Meta CFO, through the firm’s Wednesday earnings name.
Li additionally famous that whereas Meta expects to finance most of its AI spend by itself, the corporate is exploring methods to work with monetary companions to co-develop knowledge facilities.
“We don’t have any finalized transactions to announce, however we typically imagine that there can be fashions right here that can appeal to important exterior financing to assist large-scale knowledge heart tasks which are developed utilizing our means to construct world-class infrastructure, whereas offering us with flexibility ought to our infrastructure necessities change over time,” Li mentioned.
Meta has introduced two main AI “titan clusters.” The primary is Prometheus in Ohio, which is poised to be among the many first AI superclusters to hit 1 gigawatt of compute energy when it comes on-line in 2026. Then there’s Hyperion, a cluster in Louisiana that Meta CEO Mark Zuckerberg has bragged would have a footprint the dimensions of Manhattan and will scale as much as 5 gigawatts over a number of years. On high of these, Meta has a number of different unnamed titan-scale clusters underway.
Meta’s knowledge heart tasks promise to take in sufficient power to energy hundreds of thousands of houses, pulling that electrical energy from close by communities. One of many firm’s tasks in Newton County, Georgia, has already caused the water taps to run dry in some residents’ homes.
Meta additionally famous in its earnings report that it expects its second-largest driver of progress to be worker compensation as the corporate spends hundreds of thousands, and probably even billions, to poach gifted AI engineers and researchers to work for Meta’s newly shaped enterprise unit, Superintelligence Labs.
Earlier than earnings, Zuckerberg shared his imaginative and prescient for “personal superintelligence,” the concept that AI ought to assist particular person individuals dwell their greatest lives, primarily by way of the medium of Meta’s sensible glasses and digital actuality headsets.
Meta’s inventory surged 10% in after-hours as buyers responded to Meta’s general efficiency within the quarter and better-than-expected outlook for the third quarter. Meta reported income of $47.5 billion within the second quarter, with expectations to hit between $47.5 billion and $50.5 billion in Q3. Promoting drove Meta’s income good points, fueled by AI instruments — like AI-powered translations and video era — to assist advertisers create extra significant and focused campaigns.
The corporate’s Actuality Labs section, nonetheless, noticed a $4.5 billion loss.
This text was up to date to replicate Meta’s further paths to funding its AI infrastructure builds.