Meta to Invest Up to $72B in AI Infrastructure by 2025
Meta is dramatically expanding its spending on AI infrastructure, aiming to stay competitive in the accelerating global compute race. In its second-quarter earnings report, the company announced plans to invest between $66 billion and $72 billion in capital expenditures in 2025. This figure includes principal payments on finance leases and represents an increase of approximately $30 billion compared to the previous year.
Although this is a sharp jump, Meta does not intend to slow down. The company made it clear that it expects another substantial increase in 2026 as it strives to meet the growing demands of its artificial intelligence initiatives. “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’s CFO.
To support this rapid infrastructure buildout, Meta is largely funding the efforts independently. However, it is also considering partnerships. “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,” said Li. This flexibility could help Meta adapt as its needs evolve.
Titan-Scale Clusters and Energy Strain
To handle the growing AI workload, Meta is constructing massive compute clusters. These include Prometheus in Ohio, which is expected to be one of the first superclusters to reach 1 gigawatt of compute power, and Hyperion in Louisiana, which may eventually scale up to 5 gigawatts. Both are part of a broader initiative to deploy several other titan-scale clusters that remain unnamed.
However, these data centers come with considerable energy demands. In fact, Meta’s projects are expected to draw enough electricity to power millions of homes. As a result, the environmental impact is becoming a concern. One example is the Georgia-based project in Newton County, which has already led to water shortages in local communities. This raises questions about how Meta’s expansion might affect regional resources over time.
Strategic Talent and Investor Confidence
Beyond infrastructure, Meta is investing heavily in talent. The company’s second-largest growth expense is employee compensation. This includes aggressive recruitment of AI engineers and researchers for its new division, Superintelligence Labs. While Meta is willing to spend significantly to attract top talent, this strategy reflects a broader goal: to lead in AI development across platforms, particularly through smart glasses and virtual reality.
Ahead of the earnings report, CEO Mark Zuckerberg reiterated his long-term vision of “personal superintelligence.” He believes AI can empower individuals in their everyday lives through immersive technologies. Investors responded positively to Meta’s performance and outlook. Following the announcement, Meta’s stock rose by 10% in after-hours trading.
The company reported $47.5 billion in Q2 revenue and anticipates earnings of $47.5 to $50.5 billion in the third quarter. While ad revenue boosted by AI tools is the main driver, the Reality Labs unit continues to operate at a loss, posting a $4.5 billion deficit this quarter.
