Lenovo and Intel Flag Power and Cooling Constraints for AI Data Centers in Japan
DataNews Editorial Team
Lenovo Enterprise Solutions and Intel have identified grid capacity, heat dissipation and rising rack densities as primary constraints on AI data center deployment in Japan. In an industry interview, the companies positioned liquid cooling as a response to high-density AI infrastructure requirements.
- Lenovo Enterprise Solutions and Intel identify power availability, heat rejection and rack density as key AI data center constraints in Japan.
- The interview cites AI rack requirements of approximately 100–150 kW per rack.
- Limited substation capacity could restrict new data center development in Tokyo, according to Lenovo.
- Lenovo positions liquid cooling as a way to raise compute density and reduce cooling-related energy demand.
- The cited 2030 electricity outlook for Japanese data centers comes from an unnamed study and is not independently substantiated in the source.
Lenovo Enterprise Solutions and Intel said Japanese data center operators preparing for AI workloads face mounting constraints in power supply, thermal management and rack density. The companies cited limited available grid capacity—particularly in Tokyo—and the limits of conventional air cooling as AI infrastructure moves toward substantially higher-density deployments.
According to the interview, high-density AI infrastructure is expected to require approximately 100–150 kW per rack. An Intel representative said existing rack designs are poorly suited to AI workloads at around 100 kW per rack. Lenovo also said insufficient substation capacity could impede new data center construction in Tokyo.
The companies presented server liquid cooling as a potential route to accommodate these loads. Lenovo said liquid cooling can more than double resource density compared with air cooling and claimed it could reduce total data center power consumption by 30–40%. Those figures should be treated as vendor-stated potential benefits rather than universally validated operating outcomes.
Intel described the practical limit of air cooling, even with a large heatsink, as roughly 500 W. This illustrates the widening gap between conventional thermal approaches and the power demands associated with dense AI systems.
The interview also cited an unnamed study projecting that Japanese data center electricity consumption could triple by 2030, reaching about 10% of national electricity consumption. If that forecast proves directionally accurate, it would reinforce the importance of grid interconnection capacity, cooling-system upgrades and the potential for development or expansion outside metropolitan areas where substation capacity is constrained.
This was a supplier interview rather than an announcement of a specific deployment, contract or operational data center project. Nevertheless, it indicates that power availability and liquid-cooling readiness are becoming central infrastructure considerations for AI-oriented data center capacity in Japan.
The discussion underscores that AI data center deployment is increasingly constrained by site power availability and thermal design, not only by server and GPU supply. For Japan, this could shift competitive advantage toward operators with grid access, liquid-cooling-ready infrastructure and expansion options outside areas facing substation constraints.
- Impress Watch — クラウド&データセンター / October 26, 2026 · JapaneseSource →
This article was prepared by the DataNews editorial team based on the sources listed above.