A data center will be built in the middle of a California desert oasis, and in its place a 32,000-square-foot bathhouse will be built in a nearby community where people can relax in pools heated by underground servers.
New York-based architecture studio Forma designed the “Pink Thermal Bus” concept in 2021, before generative AI ignited a race to build computing facilities across America.
Forma founding partner Miroslava Brooks told Business Insider that the concept is not intended as a blueprint for turning a hyperscale AI campus into a spa. Instead, it’s asking questions that are becoming more urgent as computing complexes spring up near residential neighborhoods. What can a data center give back to the place that hosts it?
“I really think the first question is what is this building going to deliver? And it has to be more than just energy and data,” Brooks said.
Cloud is getting harder to ignore
Data centers are not new. Before Frontier AI Model Lab, we supported banks, websites, streaming services, and cloud storage.
Thomas McGoldrick, a managing director at Gensler who has been designing data centers for nearly 20 years, remembers a time when data centers were treated as “behind the scenes” facilities that supported individual businesses.
“Transferring data is becoming increasingly important today,” he said. “It has become part of our strategic infrastructure.”
A Business Insider analysis found that by the end of 2025, more than 1,400 data centers will have been built or approved across the United States.
Virginia is a hotbed for data centers, some of which are built near residential areas. Nathan Howard/Getty Images
Some data centers are being developed close to home. A 2024 Virginia study found that 29% of operational data center assets are within 200 feet of residential areas, and said neighborhood impacts could increase as suitable land dwindles.
Residents have raised concerns about constant noise, water usage and pressure on electricity bills.
A March Gallup poll found that 71% of U.S. adults oppose building AI data centers in their hometowns, including 48% who “strongly oppose.”
The scale of data centers and the opposition around them has forced architects into a debate over whether their designs can alleviate their burdens or simply camouflage them.
Adapt the gray box
San Francisco-based architecture firm Gensler has several hyperscalers among its customers, including Microsoft, and works with developers building for major cloud companies.
McGoldrick said these customers prioritize speed to market, scale, access to power and buildings that can adapt to changes in computing equipment.
“They’re all trying to get their products out as quickly as possible to grow their business as much as possible,” he says.
Within these constraints, Gensler is trying to make data centers more than empty industrial boxes. One approach, McGoldrick said, is to treat them like “office buildings that house computers.”
In one complex, Gensler repurposed an old call center campus into a 1 million square foot computing facility.
Gensler designed the COR-TEN® steel data center to blend into the local environment. gensler
The company used Corten steel to complement the rustic texture of the local landscape and added an acre of public parkland through efficient planning.
McGoldrick said the company often starts with reproducible prototypes and then adjusts materials and layouts to suit each site.
Beyond aesthetics, data centers face increased scrutiny from the community regarding energy usage and noise. McGoldrick said there are limits to what architects can do.
“There’s only so much we can control in that environment,” McGoldrick said. “So we’re being open and honest about what we’re doing and what we’re seeing in other communities.”
Data center that gives back
UK-based architecture and engineering firm Arup is exploring how standardized data center boxes will change when they are brought into cities.
Arup director Rachel Attis said traditionally long, low buildings may need smaller footprints and higher heights to expand to multiple floors. Bringing a data center to the city means architects have to “change it completely,” she said.
Arup associate director Marco Mugnai said noise issues could be addressed through acoustic screens, landscaped buffers and changes to the topography of the site.
The company also envisions data centers that would reuse structures such as redundant offshore oil rigs or combine them with tomato farms that would harness waste heat.
Arup envisions a data center that can contribute to local agriculture. Arup
“I think it’s about giving back,” Attis said. “I think every building should do that in some way.”
Some ideas require the participation of local authorities and developers to build supporting infrastructure such as district heating networks. Security and backup power requirements also mean some of Arup’s more ambitious concepts may be years away from becoming a reality, Attis said.
Forma’s Pink Thermal Baths offer a similar proposition. Rather than a linear system where electricity flows in and heat flows out, Brooks, a founding partner of Forma, envisions a “circular model” that converts excess heat into public use.
Forma’s “Pink Thermal Baths” imagine underground data centers heating above-ground public baths. Forma
Successful data centers operate across “ecosystems, infrastructure, culture and citizenship,” she said.
When design alone is not enough
Marina Otero-Vergier, an architect and lecturer at the Harvard Graduate School of Design, looked for another use for server heat.
Her “Computational Compost” project pumps heat generated by computers into a vermicomposting system, where earthworms and microbes thrive and create fertile compost for local gardens.
Otero cautioned that heat recycling is not a complete solution.
“I don’t think just reusing heat is enough because heat is already waste,” she says.
Otero said parks, low-carbon materials and shared heating systems can improve data centers, but “they are not enough.”
“Computational Compost” proposes a computer that provides heat to fertile soil. Tabacalera
She proposed challenging the data center blueprint itself, raising questions such as whether all types of data need to be readily available, whether round-the-clock operations are required, and whether the competitive demands of a company should dictate how a community’s resources are spent.
This could mean designing facilities for different “data ecologies”: hot, cold, private and ephemeral, rather than defaulting to the same high-security, always-on model, she said. It could also mean starting with what communities need, rather than reorganizing housing, energy, and other local infrastructure around data centers.
“The needs of OpenAI, Google and Meta are not the needs of the majority of the world,” Otero said. “Those are the needs of those businesses and their owners.”