


Meta to invest up to $72 billion by 2025 in cutting-edge AI infrastructure, launching colossal data centers like Prometheus and Hyperion to lead the race for superintelligent technology.
Meta CEO Mark Zuckerberg has made one of the most ambitious commitments in the artificial intelligence (AI) arms race, pledging to invest hundreds of billions of dollars to construct colossal AI data centers aimed at powering superintelligence. In a post shared on Threads, Zuckerberg unveiled Meta’s vision to dominate the future of AI through sheer infrastructure and talent.
The first mega-scale facility, Prometheus, is expected to be operational by 2026, followed by Hyperion, which will eventually scale up to 5 gigawatts. According to Zuckerberg, the scale of just one of these facilities rivals “a significant part of the footprint of Manhattan,” emphasizing the enormity of Meta’s AI ambitions.
“We’re building multiple more titan clusters as well,” Zuckerberg noted, referencing a report by SemiAnalysis that claims Meta may soon be the first AI lab to bring a gigawatt-plus supercluster online—a significant milestone in computing power.
This aggressive expansion comes as Meta ramps up efforts in its newly reorganized Superintelligence Labs, a division launched last month following internal challenges with its Llama 4 open-source model and the departure of key engineers. Leading this division are Alexandr Wang, former CEO of Scale AI, and Nat Friedman, former GitHub chief. Meta recently invested $14.3 billion in Scale AI to strengthen this unit’s foundation.
While some within Meta are considering abandoning its largest open-source model, Behemoth, in favor of a proprietary alternative, the overall strategy remains bold: build the most powerful AI systems on Earth and enable new cash-generating tools, from image-to-video ad products to smart glasses powered by Meta AI.
Zuckerberg defended the massive spend—between $64 billion and $72 billion in capital expenditure for 2025 alone—by pointing to Meta’s robust financial standing. The company posted nearly $165 billion in revenue last year and continues to thrive, with shares rising over 20% year-to-date and up 1% during trading on the day of the announcement.
“We have the capital from our business to do this,” said Zuckerberg, acknowledging investor concerns about the long-term payoff.
Industry analysts are cautiously optimistic. Gil Luria, analyst at D.A. Davidson, noted that AI has already started to enhance Meta’s core ad business by enabling more effective and profitable targeting. However, the current scale of investment is likely more about long-term dominance in AI rather than short-term financial gains.
Zuckerberg is personally leading what’s being described as a “talent raid” to attract top-tier AI researchers and engineers, further intensifying competition with rivals like OpenAI and Google DeepMind. The move underscores Meta’s strategy: combine infrastructure supremacy with elite talent to push the boundaries of what AI can do—particularly superintelligence, where machines can potentially outperform humans across a wide range of cognitive tasks.
With Prometheus and Hyperion just the beginning, Meta’s blueprint for AI dominance is clear—and it comes with a hefty price tag, bold leadership, and a clear bet on the future of intelligent machines.