How Crusoe builds data centers around AI needs

AI is moving so fast that the infrastructure to support it can barely keep up. 

Goldman Sachs predicts that data center demand will skyrocket by 50% by 2027, reaching 92 gigawatts a year. This exponential increase in demand is why flexibility is so important in AI infrastructure. However, adaptability and optimization for AI aren’t something that traditional hyperscalers considered when building data centers for the cloud era. 

It’s a problem that Crusoe is committed to fixing. 

With each contributor to the AI stack – energy, data centers and software – thinking in “different time horizons,” Crusoe is working to connect dots and fill the gaps, said Erwan Menard, the company’s SVP of product. “I think this is the biggest challenge for the industry.” 

Crusoe aims to address this issue with its Spark modular data centers, a strategy that enables data centers to be constructed in a matter of months rather than years. These modular data centers essentially pack a lot of chips into minimal space, bringing power as close to customers as possible. 

“When you look at the economics of that modular data center, they're extremely compelling for the scale,” Menard told The Deep View. “They're actually very, very compelling on a per-token basis, and they compare well with a gigawatt data center.” 

And on Thursday, Crusoe unveiled a way to get infrastructure online even faster: Crusoe Edge Zones. This offering is enabled by Crusoe’s modular approach and provides low-latency AI by deploying in “geographically targeted locations,” the company said. 

To support this effort, the company debuted a new manufacturing facility, called “Spark Factory,” in Brighton, Colorado. The factory is dedicated solely to producing its modular data centers, enabling it to scale its compute capacity even faster. 

“Speed is core to Crusoe’s DNA, and is the primary driver behind our new Spark Factory – allowing us to move beyond traditional construction methods to deliver AI infrastructure in a modular productized form factor with velocity,” Cully Cavness, co-founder, president, and chief strategy officer of Crusoe, said in a statement. 

Crusoe’s efforts target an emerging use case for its technology: Serving customers that want to support “one place and one place only,” Menard said. 

Sovereign AI, or AI that only serves one geographical location, is a growing area of interest for governments as they seek to figure out their region’s place in the AI market. A McKinsey survey of 300 executives, investors, and government officials, conducted in December, found that 71% saw sovereign AI as an “existential concern” or “strategic imperative” for their organizational goals. 

But these data centers serve more than just sovereign AI customers. Crusoe’s ability to rapidly deploy infrastructure can also help “bootstrap a site,” supporting larger data center sites and allowing them to be operational before construction is completed, said Menard, a key benefit as demand outpaces deployment. 

“That's why, when OpenAI looks at Oracle Cloud to give them an infrastructure service, they select us to do their data center development under the hood,” said Menard. “Because we have that unique expertise of doing this for AI.” 

Start building with Crusoe.