Schneider standardizes the power train, but not the grid connection

Schneider Electric has released a line of prefabricated power modules in 2MW, 2.25MW and 2.5MW configurations built on the company's Galaxy VXL UPS technology, the largest capacity the vendor says it has ever produced, and claims it can turn out prefabricated power infrastructure in as little as six months. The modules, sold alongside open-format power skids, target the hyperscale, neocloud and colocation data center environments Schneider names in its positioning.

Where the design came from matters more than the rating, since the modules rest on reference designs Schneider produced with Nvidia and carry integrated power management and liquid cooling controls for the GB300 NVL72 rack-scale platform, which aligns the electrical specification with the GPU platform rather than the hall. Nvidia's design choices function as procurement choices for everyone selling into the AI buildout, so a vendor whose power train lands in the reference design inherits distribution into every project that follows the blueprint.

Inside each enclosure sits the Galaxy VXL UPS, along with switchgear, cooling systems, lithium-ion battery technology, busway and power distribution, all tested in the factory before shipment, while the skid variant carries the same core electrical infrastructure in an open format for indoor deployment in grey space. Engineering, manufacturing and testing run through Schneider's plant in Sant Boi de Llobregat, outside Barcelona, where the company recently expanded modular data center production to 5,000 square meters, or 53,819 square feet, with European orders open now, North America later this year, and additional international markets and APAC to follow.

Dedicating factory floor to modular production is a bet in itself: a vendor adds space for prefabricated data center infrastructure when it expects an order book to justify it, and the Barcelona expansion puts roughly 53,800 square feet behind the new range, which is the conviction behind the product line.

The sales argument is schedule, and schedule is what this market pays for. Tarunjeet Sarao, Schneider's SVP for data center systems, frames the standardization as enabling providers to deploy high-density power infrastructure "in months rather than years" for what he calls AI factories, AI hubs and large-scale digital infrastructure. That is the vendor's claim, and the six-month figure deserves the skepticism any lead time attracts, because manufacturing is one clock in a data center project and not usually the one that runs longest.

The schedule claim runs into the constraint this publication has been tracking: the binding limit has moved off the site and onto the grid; connection rights now trade ahead of the electrons they carry, and a prefabricated skid compresses the piece of the timeline that was already the most controllable—assembly labor behind the meter. It does nothing for the interconnection study, the transformer order or the energized feed in front of it, and the queue does not run on a factory schedule.

Standardization is nonetheless the part worth taking seriously, because a uniform product line, the company argues, simplifies the process of designing and building power systems—a claim about the front end of every project rather than the box on the truck. Prefabrication moves labor off a site and into a factory by definition, and factory work is repeatable in a way site work is not; if the power train becomes a catalog component with a published lead time, procurement starts to behave less like a construction calendar and more like a supply chain, and a supply chain can be scaled against demand, which is what a buildout this fast requires.

The product fits the construction constraint but not the grid constraint, and construction is the one Schneider can invoice. The open question is lock-in: if Nvidia specifies the same power train for the GB300 NVL72's successors, a hyperscaler that standardizes on Schneider's module is buying into the GPU vendor's electrical blueprint as much as the modules themselves, and switching costs rise at the moment speed is worth more than optionality. The coverage does not include module pricing, one number that would tell buyers how hard Schneider intends to defend the slot.

Europe can order the modules now, but the test that matters arrives with the North American launch later this year, when the first large orders have to be squared against sites that may not yet have a feed, and the six-month lead time will then read either as a solution or as inventory sitting in a warehouse.