• Fixstars, NTTPC and Getworks launched a liquid-cooled container AI service on 8 September, combining computing hardware, software, monitoring and operational support
  • Reference pricing starts at ¥300 million plus ¥750,000 a month, while civil works, power-supply work and other site construction are charged separately

The fact

Fixstars, NTTPC Communications and Getworks began offering a containerised AI-infrastructure service in Japan on 8 September. The package is designed for installation on customer premises and combines a liquid-cooled container, GPU computing equipment, software deployment, monitoring and ongoing operational support.

The standard offer uses a 40-foot container. AI Watch reported equipment options including Nvidia HGX B300 and HPE systems, while the partners said customers can select configurations according to their requirements. The service also includes cooling equipment, network and security equipment and tools for monitoring the installed system.

Reference pricing starts at ¥300 million for a configuration with one GPU server, with a monthly charge starting at ¥750,000, excluding tax. Civil works, power-supply work and other construction needed at the installation site are priced separately. The minimum stated delivery period is eight months and may vary with the configuration, site conditions and equipment procurement.

The assessment

Bundling the equipment changes where the customer has to coordinate the project. The partners can deliver the container, cooling, computing hardware, software and monitoring as one package, but the site still has to be prepared for it. Civil works, electrical supply and other installation requirements therefore remain part of the customer's project even when more of the technology comes from one supplier group.

That boundary matters before an order is placed. Site preparation and equipment procurement have to be planned together so that the location is ready when the container arrives and the installed system can move into commissioning. Buyers also need to agree which work is included in the package, which remains their responsibility and what conditions must be met before the system is accepted for operation.

For BTW readers, the container does not remove the need to prepare AI infrastructure at the site. It packages more of the computing and cooling work into one delivery. The first deployments will show whether that arrangement makes installation easier to coordinate and how much work customers still need to complete outside the package.

What to watch

Watch for the first named customer installation and details of the delivered configuration, site works and commissioning schedule. The clearest test will be whether customers can move from site preparation to an operating AI system within the planned timetable, and which parts of that process still require separate contractors or additional spending.