NTT Data, Daikin to Test AI-driven Datacenter Cooling Prediction at Osaka Firm's Facilities

The partners will verify technology that predicts server internal thermal states without sensor access, targeting commercial launch in fiscal 2027 after trials at NTT Data datacenters.

OSAKA — Daikin Industries and NTT Data Corporation will begin verifying AI-based datacenter cooling technology in July 2026 that predicts server internal thermal conditions without direct sensor access, targeting commercial rollout in fiscal 2027.

The joint project tackles a constraint intensifying with AI server adoption: high heat loads and rapid fluctuations that traditional cooling systems cannot track efficiently. Most datacenters cannot access server internal data for security reasons, relying instead on ambient temperature sensors near equipment—a mismatch that produces over-cooling and wasted energy or under-cooling and operational risk. The new solution uses AI to infer actual server thermal states from indirect indicators like power draw and peripheral temperature readings, then coordinates air conditioning, heat source equipment, and liquid cooling across the facility.

Key project elements

• AI thermal prediction: Learns correlations between server behavior and heat output from NTT Data’s operational datasets; estimates internal states without breaching server security perimeters.
• Integrated equipment control: Unifies cooling devices under single optimization logic rather than operating each system independently.
• Verification period: July 2026 through March 2027 at NTT Data datacenters, measuring energy savings, cost reduction, and automation effectiveness.

Why it matters: AI workloads now account for the fastest-growing segment of datacenter power demand globally. A scalable cooling solution from two Tier-1 Japanese firms—NTT Data operates datacenters worldwide; Daikin dominates commercial HVAC—positions Japan’s infrastructure efficiency technology for export. The trial begins in fourteen months.