Meta (META, Financials) is diving headfirst into the AI arms race; on Monday, CEO Mark Zuckerberg announced that the company will invest “hundreds of billions of dollars” into AI computing infrastructure—starting with its first data supercluster, Prometheus, expected to go live in 2026.
The new AI compute hub—designed to train and scale large language models—is just the beginning. Zuckerberg says Meta is also building additional clusters, including one named Hyperion, which could scale up to five gigawatts over the next few years; for perspective, that's more power than some small countries consume.
In a Facebook post, Zuckerberg said the company's Meta Superintelligence Labs would offer “industry-leading levels of compute” and “by far the greatest compute per researcher”; the goal is to attract top-tier AI talent and rebuild Meta's momentum after a sluggish reception to its Llama 4 model in April.
Meta's AI pivot is fueled by urgency; Zuckerberg has reportedly grown frustrated with the pace of AI innovation inside the company. Since June, Meta has launched a full-blown hiring spree—highlighted by its $14 billion investment in Scale AI—and is now focused on building the most elite, densely packed AI team in the industry.
With OpenAI, Google, Amazon, and Microsoft all ramping up efforts in model development and infrastructure, Meta's bet on compute firepower is bold—but it might be exactly what it needs to catch up.
