Unpacking Meta’s Strategy for Data Centre Layouts

Meta is fundamentally reimagining how data centres are designed and built, abandoning traditional rigid approaches in favour of flexible layouts that can adapt to AI's unpredictable trajectory. 

From temporary tent structures to gigawatt-scale campuses, the company is deploying an unconventional strategy that prioritises speed and agility over conventional infrastructure wisdom.

The social media giant's shift reflects a broader industry trend: AI workloads have become so demanding and fluid that yesterday's facility designs simply cannot keep pace. 

Meta operates 30 data centres globally, representing billions of dollars in investment, yet the company has fundamentally rescoped its entire approach to accommodate AI's unique requirements.

How is Meta building for uncertainty?

At the heart of Meta's strategy lies a critical acknowledgement: no one can predict AI's infrastructure needs with certainty. This has led to design principles that embrace flexibility rather than fight it.

"AI, and its inference and training needs, is still evolving, so our design needs to balance what we know today with what we might know in the future," Meta explained when breaking ground on its El Paso facility in October 2025. "Different AI configurations will require different approaches to hardware and network systems designs, so our new data centres are built to accommodate flexibility.

The company's 1GW El Paso data centre exemplifies this philosophy. The 1.2 million square foot campus has been designed with systems that can support both traditional servers and future generations of AI-enabled hardware. This dual capability allows Meta to hedge its bets, avoiding the risk of building infrastructure that becomes obsolete as AI technology evolves.

Setting up camp: tents and deployment speed

Perhaps Meta's most striking departure from convention involves its use of temporary tent structures to house AI computing equipment while permanent facilities are under construction. 

The company dubs them “rapid deployment structures”. These tents are long rectangular buildings made of puncture-resistant and waterproof fabric supported by aluminium substructures.

The tent approach was deployed at Meta's Prometheus facility in New Albany, Ohio, which is expected to draw more than 1GW of power by 2026. The structures represent a calculated trade-off between speed and traditional redundancy, utilising prefabricated power and cooling modules without backup diesel generation.

“Meta is united around an ambitious mission: to build the world’s largest AI compute infrastructure and enable personal super intelligence at scale,” said Rachel Peterson, Vice President of Data Centers at Meta, writing on LinkedIn.

“A cornerstone of this mission is the agility of our data centre strategy. We are laser-focused on operational excellence, ensuring we can bring online gigawatt-scale superclusters – like Hyperion and Prometheus – with incredible velocity and adaptability. I am proud to see our teams not just embracing change but actively driving it, taking thoughtful risks and breaking new ground every day.

“To fuel this rapid expansion, we’ve pioneered solutions to deploy new data centres in a matter of months—not years.  We are building weatherproof ‘tents’ to house our GPU clusters. This approach challenges the status quo of our traditional data centre delivery, allowing us to move at unprecedented speed and demonstrating our spirit of innovation.

“I am energised every day by the ingenuity, passion and purpose our teams bring to the table. Together, we’re setting a new bar for what’s possible in the data centre industry.”

The tent strategy highlights the impetus behind Meta's digital infrastructure development. Traditional data centres require 18–24 months to build and commission, timelines that simply do not align with the pace of AI development. By deploying temporary structures, Meta can bring computing capacity online in a fraction of that time.

Ongoing adaptation as a mindset

Beyond temporary solutions, Meta has shown a willingness to completely overhaul its permanent facility designs and engineering approach. 

The company paused construction on multiple projects in late 2022 to "rescope" them for AI workloads, a decision that affected a dozen sites already under construction at the time.

In 2023, after this temporary hiatus, Meta then claimed its new designs were 31% cheaper and would take half the time to build.

Since then, the company has needed to evolve its facility design to support new technology, hardware and building innovations – let alone the surging demand from AI in the past couple of years. 

Gigawatt-scale ambitions

Meta's layout strategy extends beyond individual facility design to encompass massive multi-gigawatt campuses. 

The Hyperion site in Richland Parish, Louisiana, is projected to reach 2GW by 2030, with potential to scale to 5GW. The site spans approximately three square miles and could encompass up to nine buildings, totalling more than four million square feet.

The Beaver Dam, Wisconsin facility, announced in November 2025, spans over 700,000 square feet and represents a US$1bn investment. Like other recent Meta facilities, it has been optimised specifically for AI workloads.

Each new campus demonstrates Meta's commitment to building infrastructure that can scale rapidly. 

Sustainability within flexible frameworks

Despite the focus on speed and flexibility, Meta maintains that sustainability remains central to its layout decisions. The company aims to reach net zero emissions across its value chain by 2030, and Meta’s data centre design is a significant component of this endeavour.

All of the company’s owned facilities have LEED Gold certification or higher, and the company matches 100% of its data centre electricity use with clean and renewable energy.

Meta’s ‘Design for Sustainability’ philosophy directs its approach to IT hardware, technical design principles and optimisation – with the ultimate goal of reducing emissions and cost. 

Moreover, the company’s sustainability ambitions are being realised by tangible design strategies which include reuse, retrofitting, modularity, extended hardware lifecycles and greener materials.

Water management has also become increasingly sophisticated in Meta's layouts. The Kansas City facility, which came online in August 2025, utilises cooling technology that is significantly more water-efficient than industry standards. During construction, the site captured and repurposed stormwater from onsite retention ponds, saving more than one million gallons of potable water.

In El Paso, Meta is deploying a closed-loop, liquid-cooled system designed to use zero water for the majority of the year. The company has committed to restoring 200% of the water consumed by the facility to local watersheds.

The South Carolina data centre, expected online in 2027, will be powered by a 100MW solar farm in Orangeburg County.

"As Meta progresses construction on our data centre in South Carolina, we are pleased to join our new utility partners Central and Aiken Electric Cooperative and our long-term trusted partner Silicon Ranch to announce this project together," said Urvi Parekh, Head of Energy at Meta.

“We’re grateful to Central for supporting our energy goals in South Carolina and thrilled to expand our relationship with Silicon Ranch, who shares our commitment to have a positive impact in the communities where we locate.”

Regional layout considerations

Meta's site selection strategy reveals how layout requirements influence location choices.

Kansas City was selected because it “offered excellent infrastructure, a robust electrical grid, a strong pool of talent for construction and operations jobs, and incredible community partners,” said Brad Davis, Data Center Community and Economic Development Director at Meta.

The US$1bn Kansas City facility represents an investment of 1.4 million square feet across three data centre buildings and warehouses. The layout employed an average of 1,500 skilled trade workers at peak construction and will support more than 100 operational jobs.

El Paso's selection followed similar criteria, with Brad noting how the site is “right in the heart of the Borderplex region that embodies innovation, talent and opportunity.  

“We chose El Paso as the home for our next data centre for a variety of reasons, including strong talent, robust energy resources and great community partners,” he added.

What does the future hold for data centre layouts at Meta?

Meta's data centre layout strategy represents a fundamental rethinking of infrastructure planning. By embracing uncertainty, deploying temporary structures and designing for maximum flexibility, the company is attempting to build facilities that can evolve alongside AI technology rather than becoming obsolete the moment they are completed.

Whether this approach proves successful remains to be seen. The tent structures face questions about resilience and environmental vulnerability. The reduced redundancy in rapid deployment facilities could create service risks. And the substantial capital expenditure required to build flexibility into permanent facilities may prove wasteful if AI development takes unexpected turns.

Yet Meta appears committed to the strategy, with billions being invested in both gigawatt-scale permanent campuses and temporary solutions. As AI continues to transform computing requirements, the company's willingness to experiment with unconventional layouts may prove either prescient or cautionary – a lesson other hyperscalers will undoubtedly watch closely.