Zuckerberg pledges massive Meta investment in AI superintelligence centres

Reuters

Meta CEO Mark Zuckerberg announced plans to invest hundreds of billions of dollars into building next-generation AI data centres, signalling an aggressive long-term bet on superintelligence and reaffirming Meta’s leadership ambitions in the global AI race.

Meta Platforms will pour hundreds of billions of dollars into constructing several multi-gigawatt data centres to power artificial superintelligence, CEO Mark Zuckerberg said on Monday, as the company intensifies its push to develop advanced AI systems capable of surpassing human performance in key tasks.

The first of these facilities, named Prometheus, is set to go online in 2026, while another, Hyperion, is designed to scale up to 5 gigawatts — a capacity that Zuckerberg says rivals the physical footprint of parts of Manhattan. "We're building multiple more titan clusters as well," he posted on Threads, calling the project a turning point in Meta’s AI journey.

Meta, which earned nearly $165 billion in revenue last year, is leveraging its robust advertising business to fund the vast infrastructure. The investment, Zuckerberg argued, is justified by strong returns in AI-driven ad tools and emerging platforms such as Meta AI, video generation technologies, and smart glasses.

Despite investor scepticism over the scale of spending, Meta shares rose by 1% following the announcement and have gained more than 20% this year. Analyst Gil Luria of D.A. Davidson noted the short-term benefits to Meta’s ad business but described the superintelligence push as a longer-term play aimed at securing dominance in the next generation of AI models.

Zuckerberg’s reorganisation of the company’s AI efforts under Superintelligence Labs last month follows the uneven rollout of its open-source Llama 4 model and key personnel exits. The new unit will be co-led by Alexandr Wang (ex-Scale AI) and Nat Friedman (former GitHub CEO), after Meta invested $14.3 billion into Scale AI.

According to a New York Times report, top engineers are even weighing a shift from Meta’s open-source Behemoth model to a closed alternative, as competition heats up across the AI landscape.

Tags