Meta CEO Zuckerberg: We're building several multi-GW clusters, with the first named Prometheus, coming online in 2026.

Meta CEO Zuckerberg: We're building several multi-GW clusters, with the first named Prometheus, coming online in 2026.

Meta, the social media giant, has embarked on an ambitious expansion of its AI infrastructure, with CEO Mark Zuckerberg personally leading the charge. The company is investing heavily in building several multi-gigawatt (GW) clusters, with the first named Prometheus, set to come online in 2026. This move is part of Meta's broader strategy to catch up to rival AI labs and reassert its leadership in the field of generative AI.

Meta's current AI infrastructure, while substantial, has been criticized for falling behind in model performance. The company has been losing its lead in open-weight models to DeepSeek, prompting Zuckerberg to take a more hands-on approach. He has identified two core shortcomings: talent and compute, and is addressing both through significant investments.

Meta has acquired 49% of Scale AI at a valuation of around $30 billion, indicating that the company is willing to spend heavily to acquire top talent. The typical offer for the new "Superintelligence" team is $200 million over four years, a figure that is 100 times higher than what is typically offered to AI researchers and engineers in the industry [1].

In addition to talent acquisition, Meta is also rethinking its approach to compute. The company has scrapped its old datacenter blueprint and is now building GPU clusters in "tents," prioritizing speed over redundancy. This new approach is aimed at getting compute resources online faster and more efficiently [2].

Meta's new datacenter strategy is not limited to tents. The company is also building one of the world's largest AI training clusters in Ohio, known as Prometheus. This cluster is part of an "all of the above" infrastructure strategy that includes self-built campuses, leasing from third parties, AI-optimized designs, and multi-datacenter-campus training [3].

The investment in AI infrastructure is not without its risks. The company is facing intense competition from other tech giants and AI labs. However, Meta's aggressive approach to talent acquisition and infrastructure expansion suggests that the company is willing to take on these risks in order to maintain its leadership in the AI space.

References:
[1] https://semianalysis.com/2025/07/11/meta-superintelligence-leadership-compute-talent-and-data/
[2] SemiAnalysis Datacenter Model – as of 07/06/2025
[3] SemiAnalysis Datacenter Model – as of 07/01/2025