Fixstars, NTTPC, and Getworks launch a one-stop liquid-cooled GPU container data center package for on-premises AI.
The AMW Read
Productizes a prior Fixstars-NTTPC-Getworks PoC into a commercial liquid-cooled GPU container data-center package for on-prem/confidential AI workloads in Japan, an incremental but concrete infrastructure offering shaped by cooling and data-locality constraints.
Fixstars, NTTPC, and Getworks launch a one-stop liquid-cooled GPU container data center package for on-premises AI.
Tokyo-listed Fixstars (TSE Prime: 3687), NTTPC Communications, and Getworks began offering the "Liquid-Cooled GPU Container DC Package" on September 8, 2026, bundling Getworks' container-type data centers, liquid-cooled NVIDIA-accelerated servers, integrated monitoring, and Fixstars' operations support into one procurement. The package follows a joint PoC the three companies announced in December 2025 validating water-cooled GPU server operations for commercial use in Japan. Pricing is quote-based, minimum lead time is eight months, and components can be reconfigured to customer needs.
The launch targets a real bottleneck: as enterprises move generative and agentic AI from pilot to production, GPU servers are becoming more power-dense and harder to cool, while concern about keeping sensitive data off third-party clouds — plus network limits and the cost of moving large datasets — pushes some buyers toward on-premises deployment on their own land and power infrastructure. The partners cite two use cases: confidential-data AI analysis that cannot leave a customer's site, and manufacturing/physical-AI applications such as defect detection on production-line video.
For buyers, the package compresses vendor selection across cooling, compute, and monitoring into one contract, addressing the design expertise and staffing gaps the companies cite as adoption barriers. The eight-month minimum lead time signals this is a planning-cycle purchase, not fast provisioning. The partners plan to extend the package to NVIDIA's Vera Rubin platform and to distributed/edge AI deployments, positioning the consortium as an on-prem alternative to hyperscaler GPU cloud access for security-sensitive Japanese enterprises.