Koukaryoku (High-Power Computing) GPU services

Offering overview

Domestic GPU capacity for Japan's AI developers is the focus of the Koukaryoku line, a brand that predates the current generative AI boom. Three formats are sold: PHY bare-metal servers with eight NVIDIA H100, H200 or B200 SXM GPUs per node on monthly contracts, VRT virtual machines, and DOK, a container service where customers upload a Docker image and pay by the second for V100 or H100 time. Older P40 and V100 plans remain listed for lighter workloads. Everything runs in the company's Ishikari data center in Hokkaido, which operates on renewable electricity and reports zero annual CO2 emissions, an attractive detail for buyers with sustainability reporting obligations. Keeping data within Japan, with a long-established domestic operator, is the main draw for universities, research labs and enterprises training or fine-tuning models under local compliance expectations. Entry pricing is low: a DOK V100 container costs JPY 0.016 per second (about JPY 57.6 per hour including tax) with one-second minimum billing, while an H100 container runs JPY 0.28 per second. Multi-GPU PHY servers are priced on request as monthly subscriptions.

Technical specifications

Containers

Unknown

Kubernetes

Unknown

Root access

Unknown

API access

Unknown

Autoscaling

Unknown

Persistent processes

Unknown

AI capabilities

Model serving

Not stated

AI frameworks

Not stated

Inference endpoints

Unknown

Fine-tuning

Unknown

Vector database

Unknown

Model catalog

Unknown

Monitoring

Not stated

Evaluation

Not stated

Security & compliance

Geography

Provider headquarters are separate from offering infrastructure and availability.

Available plans

Pricing can change. Promotional prices are shown separately from standard prices.

Questions & answers

What GPU hosting formats does Koukaryoku offer?

Koukaryoku offers PHY bare-metal servers, VRT virtual machines and DOK GPU containers. PHY nodes have eight NVIDIA H100, H200 or B200 SXM GPUs.

How does the DOK container service work?

Customers upload a Docker image and use V100 or H100 GPU time billed by the second. The service has a one-second minimum billing period.

Where are Koukaryoku workloads hosted?

All workloads run in the Ishikari data center in Hokkaido, keeping data within Japan. The facility operates on renewable electricity and reports zero annual CO2 emissions.

Who uses Koukaryoku GPU services?

The service serves Japan's AI developers, universities, research labs and enterprises training or fine-tuning models under local compliance expectations.