As tech giants gear up for the future of artificial intelligence, Meta has a plan that exceeds imagination in scale.
Meta is rapidly expanding its infrastructure at an unprecedented pace. The company believes that such infrastructure is crucial for training and deploying future AI models. Recently, Meta CEO Mark Zuckerberg unveiled ambitious plans that include building massive compute clusters with multi-gigawatt capacities. One interesting move is the construction of tent-like data center structures—a strategy inspired by Elon Musk’s xAI, aimed at significantly reducing both cost and construction time.
While Meta’s Llama language models haven’t garnered as much attention as expected, the company is now focusing on developing native infrastructure for AI processing. According to a report from SemiAnalysis, Meta will be the first company to build an AI-dedicated data center at the scale of one gigawatt. This new center will deliver over 3 billion teraflops of computing power. However, Zuckerberg says this is just the beginning.
One of the new computing centers, named Prometheus, is set to go online in 2026. In addition, a much larger project called Hyperion is underway, which will eventually consume up to 5 gigawatts of power. Some of these data centers are said to be as large as major sections of Manhattan in New York City.
Zuckerberg has promised that “Meta’s AI super labs” will offer the highest compute capacity among competitors, with an unmatched compute-to-researcher ratio. This statement may signal a new revenue model for Meta—providing AI infrastructure to other companies, much like Amazon or Groq.
To achieve this goal, Meta must rapidly scale its projects. Reports suggest the company has completely redesigned its data center construction approach—eliminating unnecessary components like backup diesel generators and prioritizing speed and efficiency. The use of prefabricated modules for power and cooling, sourcing energy from local substations, and even building two 200-megawatt gas power plants in Ohio are all part of Meta’s bold plan to dominate the future of AI processing.
