NVIDIA's 800V DC architecture and a modular gigawatt data-center design attack the power wall at the rack and the site
NVIDIA released an 800-volt DC power architecture and, with a 31-company ecosystem, is standardising how power moves inside the rack — while a Bechtel partnership modularises a 1-gigawatt data center design to accelerate 'AI Factory' build-out. The response to the grid ceiling is engineering at both ends: more efficient power delivery inside, faster construction outside.
800V DC is an efficiency play at the point of highest loss. As racks climb past 120 kilowatts, the resistive losses in traditional power distribution become a first-order cost, and moving to higher-voltage direct current cuts the current for the same power, which cuts the loss. It is the kind of unglamorous infrastructure change that only becomes worth standardising when the power budget is the constraint.
The modular gigawatt design attacks the other bottleneck: time. If the grid interconnection takes 24-36 months, shaving months off the construction of the facility that sits behind it is one of the few levers a chip vendor controls. Standardising and prefabricating a 1-GW design is an admission that build speed is now part of the competitive surface.
The through-line is that NVIDIA is no longer only a silicon company; it is increasingly a power-systems and construction-standards company, because the thing gating its customers' growth is no longer the chip. When the bottleneck moves downstream of your product, you either follow it downstream or watch demand stall against it.
EnkiAI — NVIDIA's 2026 power play: how AI is reshaping the grid → · GPU Insights — AI data center power infrastructure 2026 — GPU clusters →