The grid is aging and outages are rising. Its infrastructure is mechanical, slow, bulky, and carries no intelligence at all. Meanwhile everything being added to it, solar, wind, EV charging and storage, is natively DC. Managing distributed resources on a decentralized, bidirectional grid needs infrastructure faster and smarter than anything installed today, and load growth from AI data centers, transportation and electrification is arriving faster than the grid can absorb it.
Server racks are DC native and climbing, from 15 kW to over 100 kW and heading to 500 kW. Legacy AC infrastructure is inefficient and not reliable enough to carry that. Loads swing by the megawatt in milliseconds, which shows up on the utility as flicker and delays interconnection approval. And when a facility goes down it costs roughly $8,000 per minute per megawatt.
Legacy AC protection has to be derated, so a 1 MW circuit only ever delivers 800 kW. You buy the megawatts, then build extra rows to carry them.
On a ten-building gigawatt campus, an engineering firm put that at a whole building.
100 MW example. Row counts and footprint based on Merge4 engineering analysis. The 40% figure is power distribution floor given back by the rows; the campus figure comes from a third-party engineering analysis where power rows are one part of the floor.
