Meta’s push into frontier AI is moving beyond talent and models. The company is now laying out an infrastructure plan built around enormous data centers, including a project called Hyperion that CEO Mark Zuckerberg says is expected to supply Meta’s new AI lab with five gigawatts of computational power.
The plan places compute at the center of Meta’s competition with OpenAI, Google, Google DeepMind, and Anthropic. It also sharpens a larger question around artificial intelligence: how much electricity, water, and local infrastructure will the next phase of AI require?
Meta’s AI data center strategy is getting larger
Zuckerberg said in a Monday post on Threads that Meta is building Hyperion, a data center intended to support the company’s AI work. According to Meta spokesperson Ashley Gabriel, Hyperion will be located in Louisiana, likely in Richland Parish where Meta previously announced a $10 billion data center development.
The scale is unusually large even by modern data center standards. Zuckerberg said Hyperion’s footprint will be large enough to cover most of Manhattan. Gabriel said Meta plans to bring two gigawatts of data center capacity online by 2030 with Hyperion, with the project scaling to five gigawatts in several years.
Meta is also planning a separate 1 GW super cluster called Prometheus. Zuckerberg said Prometheus is expected to come online in 2026, which would make Meta one of the first tech companies to control an AI data center of this size. Gabriel said Prometheus is located in New Albany, Ohio.
Together, Hyperion and Prometheus signal that Meta sees physical infrastructure as a core part of its AI strategy. The company has already moved to strengthen Meta Superintelligence Lab by bringing in top talent, including former Scale AI CEO Alexandr Wang and former Safe Superintelligence CEO Daniel Gross. Now the focus is shifting toward the computational power needed to train frontier AI models and serve them at scale.
Why compute has become a competitive weapon
Training and serving leading AI models requires more than researchers and software. It also requires access to large amounts of computational power. Meta’s build-out appears aimed at making the company more competitive with OpenAI, Google DeepMind, and Anthropic in that area.
Control over AI data center capacity can matter in several ways. It can help a company train larger or more capable systems. It can also support the deployment of AI products once those systems are in use. For companies racing to build frontier models, compute is not just an operating expense; it is part of the product roadmap.
Meta’s plans may also affect hiring. The source article notes that the effort could help Meta attract additional talent, since researchers and executives may be drawn to work at a company with the computational resources needed to compete in the AI race.
That creates a reinforcing cycle. Companies with stronger AI infrastructure can pursue more ambitious model development. More ambitious model development can help attract people who want to work on frontier systems. Those teams, in turn, increase the demand for more infrastructure.
The local cost of AI infrastructure
The same scale that makes Hyperion and Prometheus strategically important also raises concerns for nearby communities. Together, the two projects will soak up enough energy to power millions of homes, according to the source article. That level of demand could pull significant amounts of electricity and water from surrounding areas.
Those risks are not theoretical. The New York Times reported Monday that one of Meta’s data center projects in Newton County, Georgia, has already caused the water taps to run dry in some residents’ homes. Other AI data center projects may create similar pressure elsewhere.
Bloomberg reported that AI hyperscaler CoreWeave is planning a data center expansion projected to double the electricity needs of a city near Dallas, Texas. That example shows that Meta is not alone. The wider AI industry is expanding physical infrastructure in ways that can reshape local utility needs.
The core tension is straightforward:
- AI companies want more computing capacity to train and serve advanced models.
- Large data centers require major electricity and water resources.
- Nearby communities may face stress if infrastructure growth outpaces local capacity.
- Without rapidly increased energy production, those pressures could grow.
This is why the AI data center boom is becoming both a technology story and an infrastructure story. The race is not only about who builds the strongest model. It is also about who can secure the land, power, water, and construction capacity to run it.
A wider AI build-out is underway
Meta’s projects are part of a broader surge in AI infrastructure. Other major efforts include OpenAI’s Stargate project with Oracle and SoftBank, as well as xAI’s Colossus supercomputer. These projects point to the same industry-wide conclusion: tech companies are determined to build massive data center capacity to support their AI ambitions.
The Trump administration has largely supported the technology industry’s AI data center buildout. President Donald Trump helped OpenAI announce its Stargate project and has spoken about efforts to expand America’s AI infrastructure.
In a column featured in The Economist on Monday, U.S. Secretary of Energy Chris Wright called for the U.S. to “lead the next major energy-intensive frontier: artificial intelligence.” He also wrote that AI transforms electricity into the “most valuable output imaginable: intelligence,” and said the federal government would accelerate energy production from coal, nuclear, geothermal, and natural gas.
That policy support matters because AI data centers are expected to consume far more energy in the years ahead. Experts estimate that data centers could account for 20% of America’s energy consumption by 2030, up from just 2.5% in 2022.
Meta’s Hyperion and Prometheus plans show where the industry is heading. Frontier AI is becoming a contest over infrastructure as much as algorithms. The companies that can marshal enough power and compute may gain an advantage, but the communities around those projects will be left to manage the practical consequences.