Meta is turning AI ambition into a major infrastructure buildout. In its second-quarter earnings report, the company said it expects 2025 capital expenditures, including principal payments on finance leases, to reach $66-72 billion as it expands the data centers, servers and compute capacity behind its artificial intelligence plans.
A sharp rise in AI infrastructure spending
The company framed the spending increase as part of a broader push to bring more capacity online for artificial intelligence efforts and business operations. At the midpoint of the range, Meta said the increase would be up approximately $30 billion year-over-year.
That figure points to a major shift in how Meta is allocating resources. The company is not only adding infrastructure for current products; it is preparing for AI systems and product experiences that require far more physical and technical support.
Meta also signaled that the spending surge is not limited to 2025. The company said it expects a similarly large increase in spend on AI infrastructure in 2026 as it continues to aggressively pursue opportunities to add capacity.
Susan Li, Meta CFO, described leading AI infrastructure as a core advantage for building the best AI models and product experiences. That statement makes the strategic logic clear: Meta sees infrastructure not as a background cost, but as a competitive asset.
Data centers become the center of the strategy
The buildout is tied to large AI clusters designed to support Meta’s artificial intelligence work. The company has announced two major AI “titan clusters”: Prometheus in Ohio and Hyperion in Louisiana.
Prometheus is expected to be among the first AI superclusters to hit 1 gigawatt of compute power when it comes online in 2026. Hyperion, according to Meta CEO Mark Zuckerberg, could have a footprint the size of Manhattan and could scale up to 5 gigawatts over several years.
Meta also has several other unnamed titan-scale clusters underway. Together, these projects show how the AI infrastructure race is moving beyond software teams and model development into land, power, construction and long-term capacity planning.
The source article notes that Meta’s data center projects promise to soak up enough energy to power millions of homes, pulling that electricity from nearby communities. It also reports that one project in Newton County, Georgia, has already caused water taps to run dry in some residents’ homes.
Meta is weighing outside financing
Meta expects to finance most of its AI spending on its own. Even so, Li said the company is exploring ways to work with financial partners to co-develop data centers.
There were no finalized transactions to announce. But Meta said it believes models could attract significant external financing for large-scale data center projects, while giving the company flexibility if its infrastructure requirements change over time.
That flexibility matters because AI infrastructure needs can change quickly. A company building at this scale may want access to more capacity without locking every future decision into the same balance-sheet structure.
The financing discussion also shows how expensive the AI infrastructure race has become. Data centers, servers and clusters are not short-term experiments; they are long-lived assets that require substantial capital before they can support product growth.
Talent costs are rising alongside compute
Infrastructure is not the only major expense tied to Meta’s AI plans. The company said its second-largest driver of growth is expected to be employee compensation.
That spending is connected to Meta’s efforts to recruit AI engineers and researchers for its newly formed business unit, Superintelligence Labs. The source article says the company is spending millions, and possibly even billions, to poach talented people for that work.
Before the earnings report, Zuckerberg shared his vision for “personal superintelligence.” In that vision, AI helps individual people live their best lives, mainly through Meta’s smart glasses and virtual reality headsets.
This links Meta’s infrastructure spending to its product strategy. The company is building compute capacity, hiring specialized talent and tying both to AI experiences that could reach users through hardware and software products.
Investors focused on performance and outlook
Meta’s stock surged 10% in after-hours trading as investors reacted to the company’s overall performance in the quarter and its better-than-expected outlook for the third quarter.
Meta reported revenue of $47.5 billion in the second quarter. For Q3, the company said it expects revenue between $47.5 billion and $50.5 billion.
Advertising drove the revenue gains, helped by AI tools such as AI-powered translations and video generation. Those tools are designed to help advertisers create more meaningful and targeted campaigns.
At the same time, Meta’s Reality Labs segment posted a $4.5 billion loss. That result sits beside the company’s broader AI push, underscoring how much Meta is investing across future-facing products while its advertising business continues to generate the revenue gains highlighted in the earnings report.