The facility would harness NVIDIA's next-gen Vera Rubin architecture as Japan races to become an AI infrastructure heavyweight.
Japan is positioning itself as a serious contender in the global AI infrastructure race, with plans for a datacenter powered by NVIDIA’s Vera Rubin GPUs targeting a June 2028 launch. The project reflects a broader push by Japanese and allied entities to build out the kind of compute capacity that next-generation AI workloads demand.
NVIDIA’s Vera Rubin platform is the company’s next step beyond its current Blackwell architecture. Each NVL72 rack packs 72 Rubin GPUs alongside Vera CPUs and NVLink 6 interconnects. Volume shipments for the Vera Rubin platform are scheduled to begin in fall 2026.
South Korea’s SK Group has announced plans to construct an AI-specialized “AI Factory” datacenter in Japan, with an operational timeline between 2028 and 2029.
NVIDIA is exploring a partnership with Mitsubishi Heavy Industries, reported in July 2026, focused on developing advanced cooling and power systems specifically designed for AI datacenters.
Hitachi has also entered the picture with a memorandum signed alongside Mitsui OSK Lines to develop a floating datacenter in Japan.
For anyone watching the AI hardware and infrastructure space, Japan’s datacenter push creates several investment vectors worth tracking. NVIDIA remains the most obvious beneficiary, with each NVL72 rack representing 72 Rubin GPUs per purchase order. The cooling and power angle is also notable, with Mitsubishi Heavy Industries addressing thermal management demands as GPU power consumption climbs with each generation. The risk to watch is execution. A June 2028 target for Rubin GPU operations depends on NVIDIA hitting its fall 2026 shipment schedule, construction timelines holding firm, and power grid capacity being available.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our