In fiscal 2025 Meta Platforms allocated $25 billion to AI-related capital expenditures, a 50% increase over the prior year. This spending is focused on expanding its data center footprint—adding three new hyperscale facilities in Virginia, Texas and Denmark—and on custom chip development through its in-house AI accelerator program. The company’s total capex for the year reached $40 billion, up from $32 billion in fiscal 2024, signaling management’s commitment to supporting next-generation generative AI services and large-language-model training at scale.

Meta reports that its AI-powered News Feed ranking algorithms boosted daily active user engagement by 5% year-over-year, reaching 2.76 billion in Q4 2025. The deployment of real-time recommendation models has also lifted ad click-through rates by 12% and increased average revenue per user in North America by 8%. Meanwhile, the company’s open-source Llama models are now handling roughly 25% of all external AI token volume, creating a new revenue stream via enterprise licensing and reinforcing Meta’s position as a key supplier to developers building AI applications on its platform.