The growth of AI – including generative AI systems like ChatGPT and Grok – is changing the way the world is trying to build data centers, and that includes where they need to run. It’s also changing their size, their speed, and the technology itself is significantly different.

According to Kramer, this seachange must be supported with a modular approach, which requires a close look at data center design to ensure that any new data center not only meets today’s standards but can flex to future requirements.

“In the past, you could build every data center in a new and different way, and you could have a unique design because there was enough time, people and resources to do that, but today there isn’t,” he explains. Fast forward to today, where the need to build at such speed and scale has spearheaded the demand for standardization, ideally, through the lens of the manufacturer.

Standardized, modular approach

Data center design is no longer about looking at a thousand parts, but about boiling it down to five subsystems that can be modularized. Kramer explains that a modular data center is about building the whole subsystem of the facility, or even, in some cases, the whole thing, in a factory.

This means there is no need to build it on site, but the site in question can prepare to receive a much larger system while doing numerous parallel activities. “You get the benefit of speed, and you don’t have to have all the people at the site that you might otherwise need because you are using my factory’s labor,” clarifies Kramer.

Overcoming barriers

These days the major barriers, he says, are about how many data centers need to be built, how fast they need to be constructed and the availability of resources – which includes skilled construction labor.

“Because we're building in a factory, it's going to be a lot more reliable; it's going to start up faster, commissioning is going to be quicker, and you're going to be off to the races much faster with your equipment,” he remarks.

The key thing is to be adaptable to customers’ needs. One size definitely doesn’t fit all. One individual’s standards may be different to another’s. “So what we do is we spend our time working with our clients to figure out the things that fit their business, and then we try to take that to a high volume approach, so that they can get all the benefits that I just described for their business.”

Essential to move fast

While being adaptable, he says it’s often essential to move fast. However, in order to achieve this without compromising on quality and safety, Kramer advises engaging with partners from across the ecosystem as soon as possible.

“We want to be as close to the beginning as possible as we can help steer them with our knowledge,” he comments. The aim is to add extra layers of expertise to the whole solution, and so the earlier the company gets involved, the better chance he and his team has to add the most value.

Key benefits – modular approach

The key benefits of Scheider Electric’s modular data center approach include much faster deployment, parallel workstreams, shorter installation and commissioning while supporting AI buildouts to bring capacity rapidly online. With this comes standardization at scale, reduced onsite labor and resource constraints, higher reliability and quality, better alignment with high-density AI requirements, reduced integration headaches across vendors, de-risked operations via the use of digital tools, and increased flexibility with standardization.

“One of the clients that we work with is Compass Data Centers: we engaged early with them, and they’ve really bought into the modular and standardized approaches, so they're able to leverage that upstream with their clients,” he says. This allows them to go to the marketplace and say, ‘I can build it faster and more reliably than the others’, which has been a successful strategy for them.

Impacting value propositions

Subsequently, Schneider Electric is able to supply more than just equipment, getting to the heart of its customers’ value propositions in order to help them better communicate with their own customers.

Compass Data Centers has implemented a cookie-cutter design, which although standardized, is fungible enough to allow for optimization, thanks to a myriad of different options available to its customers. This flexibility has resulted in Compass having “grown dramatically in the last couple of years”.

With the march of AI, including the increasing demand for power to enable it to operate, he argues that there is no other way forward fast. Still, the market is going to continue to evolve and change, and so he believes his team has got to be ready to match or supersede whatever is going to happen next.

To find out more about a modular approach to building AI data centers, check out the full DCD>Talk here.