Watch my next video investigating AI Data Centers: https://nebula.tv/videos/stewarthicks-why-america-is-afraid-of-colorful-houses
Get Nebula using my link for 50% off an annual subscription: https://go.nebula.tv/stewarthicks
Meta’s Hyperion AI data center in Louisiana is projected to consume five gigawatts of power and cost more than $50 billion. It is one of the largest construction projects on Earth—built to house computing equipment that may become outdated in roughly the time it takes to finish the campus.
That raises a strange architectural question: why are we constructing permanent buildings around such temporary machines?
Factories, warehouses, cold-storage buildings, and even old post offices were once among the most technologically advanced structures of their era. Many have since become apartments, offices, galleries, stores, and entire urban neighborhoods. Meta’s own New York offices occupy part of the former James A. Farley Post Office, a building constructed more than a century ago for sorting mail.
Modern AI data centers appear to share many of the qualities that made those industrial buildings adaptable: repetitive structural bays, tall ceilings, heavy floor capacity, durable envelopes, and large open interiors. Yet today’s hyperscale data centers are usually treated less like lasting infrastructure and more like disposable appliances—single-purpose campuses built around one generation of computing equipment.
In this video, we look inside the architecture of an AI data center, from GPU racks and liquid cooling systems to substations, generators, switchyards, and enormous single-story shells. We compare Meta’s Hyperion campus with Chicago warehouses, converted industrial lofts, the former U.S. Steel South Works, and the emerging Illinois Quantum and Microelectronics Park.
The AI building boom is transforming land, electrical grids, and communities across the United States. But what happens when the servers are replaced, the lease expires, or the next generation of computing requires an entirely different kind of facility? The history of the Rust Belt suggests that buildings designed without an afterlife can leave towns paying the bill for decades.
#Architecture #ArtificialIntelligence #DataCenters
__Special Thanks__
Evan Montgomery — coproducer
Daniela Osorio Sanudo — research and graphics
BIBLIOGRAPHY
1. Meta Platforms, Inc. “Annual Report for the Fiscal Year Ended December 31, 2025.” Form 10-K, U.S. Securities and Exchange Commission, 2026. Meta reports an estimated useful life of approximately 5.5 years for most servers and network assets.
2. Meta Platforms, Inc. “Meta Announces Joint Venture with Funds Managed by Blue Owl Capital to Develop Hyperion Data Center.” October 21, 2025. The announcement describes Meta’s initial four-year operating leases for the completed campus.
3. Reuters. “Meta Expands Louisiana Data Center to 5 Gigawatts, Investment Crosses $50 Billion.” July 13, 2026.
4. Meta. “The Largest Meta Data Center Yet Brings Big Impact to Louisiana.” Richland Parish Data Center project overview.
5. U.S. Department of Energy. *Best Practices Guide for Energy-Efficient Data Center Design.* Federal Energy Management Program and National Renewable Energy Laboratory, 2024.
6. Virginia Joint Legislative Audit and Review Commission. *Data Centers in Virginia.* 2024. Includes employment, land-use, infrastructure, and economic-impact analysis of hyperscale data-center development.
7. Kohn Pedersen Fox. “Meta Farley.” Project description of Meta’s offices inside New York’s former James A. Farley Post Office.
8. ComEd. “Powering the Future: How ComEd’s Partnership Sparked the Illinois Quantum Campus.” August 26, 2025. Discusses the approximately 350-megawatt power requirements of the planned quantum campus at the former U.S. Steel South Works site.
__Attributions__
Stock video and imagery provided by Getty Images, Storyblocks, and Shutterstock.
Music provided by Epidemic Sound and includes music from Chromatic by Tom Fox
https://www.youtube.com/@chromaticbytomfox”
#architecture #urbandesign
source
