Meta Platforms has announced a sweeping investment plan of approximately $600 billion to be spent in the United States over the next three years, with a large portion of this aimed at AI-driven data-centre build-out. The company says the scale and speed of these facilities are central to its vision of “superintelligence” infrastructure and supporting its next-generation AI ambitions.
A key component of the plan includes building “titan” AI clusters, such as the data‐campuses known as Prometheus and Hyperion. For example, Prometheus is scheduled to go online in 2026, while Hyperion is expected to scale to around 5 gigawatts of computing capacity—comparable in footprint to a significant part of Manhattan.
Meta’s strategy appears to be front-loading infrastructure investment in the belief that compute capacity will become a bottleneck in the AI arms race. CEO Mark Zuckerberg has framed this as a “preparedness” measure—building more than is needed today so the company isn’t caught off guard tomorrow.
However, the scale of investment also prompts questions about demand, payback timelines and environmental impact. With such massive infrastructure bets, the risk is that actual revenue and productivity gains will lag expectations, or that power/energy constraints could become a limiting factor. Observers of the industry note that while the infrastructure is real, whether the returns will match the hype—and speed—remains uncertain.
