Meta is stepping up its AI game, with Mark Zuckerberg revealing plans for a massive expansion of AI infrastructure over the next few years. According to Zuckerberg, Meta aims to add over 5GW of AI compute power, a move that signals the company’s aggressive push to dominate the AI industry. Meta has already invested billions not only in its infrastructure but also in creating a world-class AI team under the Superintelligence Labs banner, featuring top-tier talents poached from companies like Scale AI and OpenAI.
As part of this expansion, Meta plans to roll out ‘multi-GW’ AI clusters, with a projected $65 billion to be spent by the end of the year on AI CapEX. 
The company’s first 1GW lab, named Prometheus, will be online by 2026, while the ambitious Hyperion project will scale the AI computing power to a whopping 5GW in the coming years. These developments reflect Meta’s broader strategy of going all-in on AI, with multiple additional clusters planned to further enhance their AI capabilities.
In comparison to some of the world’s largest supercomputers, Meta’s upcoming AI clusters will demand an unprecedented amount of power. For context, the El Capitan supercomputer requires 30 megawatts, while Meta’s planned clusters will demand around 160 times that amount. This is not just a technical feat-it’s a statement of intent, positioning Meta as a leader in AI development.
As Meta continues to push the envelope on AI, the company is also positioning itself to lead the race toward AGI (Artificial General Intelligence), potentially far ahead of other Big Tech players. With this colossal investment in infrastructure and talent, Zuckerberg’s vision seems clear: Meta plans to dominate AI, using its Llama models to fuel the next generation of intelligent systems.