Behind Mark Zuckerberg's Major Meta Supercluster Push
In a sweeping announcement on Facebook, Meta CEO Mark Zuckerberg outlines a bold strategy to invest "hundreds of billions of dollars" into developing artificial superintelligence — AI systems with capabilities surpassing human cognition.
Central to this grand vision is the deployment of advanced data centres.
Meta is spearheading two innovative initiatives — Prometheus and Hyperion — crafted to deliver industrial-scale computational power via superclusters.
These data centres act as the physical backbone of Meta’s AI drive, supplying the necessary bandwidth, processing capability and technical framework to train its next-generation models.
“Meta Superintelligence Labs will have industry-leading levels of compute and by far the greatest compute per researcher,” Mark says.
This undertaking not only requires recruiting the top researchers but also involves developing infrastructures that can fulfil the ambitious scale and computation power present in these visionary projects.
Constructing superintelligence
The cornerstone of this endeavour is Prometheus, introducing a 1GW data cluster. Expected to be operational by 2026, this cluster will advance beyond any current AI infrastructure.
According to SemiAnalysis, Meta is poised to be the pioneer in launching a cluster exceeding 1GW.
Costing an estimated US$30bn per GW, Meta's computational ambitions are transformative.Hyperion, a subsequent project, is intended to be scalable to 5GW through multiple phases over the years.
These superclusters are designed to support Meta’s burgeoning model suite, including forthcoming iterations of Llama 4 and beyond.
Together with Prometheus and Hyperion, Meta is planning additional "titan" clusters to further augment overall capacity.
Meta’s pivot in data infrastructure mirrors its competitors like Microsoft and Amazon, who are already collaborating with prominent AI developers and significantly scaling their own computational resources.
In his message, Mark emphasizes Meta’s unique advantage in scale and autonomy, stating: “We have the capital from our business to do this.”
Recruiting elite talent
To effectively utilise these clusters, Meta has established Meta Superintelligence Labs (MSL), a fresh division steering its AI pursuits.
The division is guided by industry leaders Alexandr Wang, previously of Scale AI, Daniel Gross, CEO and Co-Founder of Safe Superintelligence and Nat Friedman, former CEO of GitHub.
MSL already gains from Meta’s US$14.3bn acquisition of Scale AI, renowned for its data labelling infrastructure and the HLE (Humanity’s Last Exam) benchmark, addressing previous Llama 4 quality issues.Meta is undertaking an extensive recruitment campaign, pulling talent from top technology firms to push ahead.
New inductions include individuals from Google DeepMind, OpenAI and Anthropic, with names such as Hongyu Ren (GPT-4o and o-series models co-creator), Jiahui Yu (GPT-4o co-creator), Jack Rae (Gemini pre-training lead), Pei Sun (Gemini reasoning expert) and Shengjia Zhao (ChatGPT co-creator).
Reportedly, Meta has presented packages as lucrative as US$300m over four years, igniting competition across the industry.
Mark’s personal engagement in the recruitment process — hosting meetings at his Lake Tahoe and Palo Alto residences — demonstrates how central MSL is to the future trajectory of Meta.
In a prior memo distributed internally to Meta staff, Mark said: “As the pace of AI progress accelerates, developing superintelligence is coming into sight.
“I believe this will be the beginning of a new era for humanity and I am fully committed to doing what it takes for Meta to lead the way.”
In the recent declaration, Mark elaborates on his strategy to realise this vision: “For our superintelligence effort, I'm focused on building the most elite and talent-dense team in the industry.”
In response to the announcement, Karl Freund, Principal Analyst at Cambrian AI Research, told the BBC, "Clearly, Zuckerberg intends to spend his way to the top of the AI heap."
"The talent he is hiring will have access to some of the best AI Hardware in the world," he added.
