NTT Data, Daikin, and Start AI joined forces in July to run an advanced data center cooling test in Japan, marking a notable step in AI infrastructure innovation. The collaboration targets energy efficient cooling for high density workloads, aligning with rising demands across cloud and enterprise AI deployments.
Industry watchers tracking IBTimes JP coverage noted the test as a practical validation of AI optimized cooling strategies. By combining Daikin’s cooling hardware expertise with NTT Data’s digital services and Start AI’s analytics, the project aims to lower power usage effectiveness and improve thermal management at scale.
| Project | Key Partners | Primary Goal | Test Timeline |
|---|---|---|---|
| NTT Data Daikin Start AI Data Center Cooling Test | NTT Data, Daikin, Start AI | Validate AI optimized cooling for dense workloads | July trial period |
| Scope | Japanese data center environment | Energy efficiency and thermal stability | Measured under real AI load conditions |
| Innovation Focus | AI driven cooling control | Reduce PUE and improve server inlet air management | |
| Expected Outcomes | Cooling strategy insights, performance benchmarks | Path toward scalable deployment across Asian markets |
AI Workload Cooling Challenges in Modern Data Centers
High density AI racks generate concentrated heat patterns that traditional cooling architectures struggle to handle efficiently. NTT Data, in collaboration with Daikin and Start AI, examined how adaptive control strategies could match rapid fluctuations in AI compute demand.
The July test specifically targeted scenarios where GPU clusters operate at near peak utilization. Monitoring equipment tracked thermal gradients, fan energy consumption, and airflow distribution to refine the control logic underpinning the cooling system.
Daikin Cooling Hardware Integration with Data Center Infrastructure
Daikin’s modular cooling units were deployed to interface with existing air handling and chilled water systems. Precise staging of equipment allowed the team to simulate both partial and full scale AI workload profiles without disrupting other tenants.
By aligning airflow paths with the physical layout of server rows, the project reduced hot spots and improved thermal uniformity. Sensor data from multiple zones informed adjustments to pump speeds, refrigerant flow, and setpoint temperatures.
Start AI Analytics and Real Time Control Algorithms
Start AI contributed predictive models that forecast cooling load based on job schedules, historical power usage, and environmental conditions. These models feed into a real time optimization layer that adjusts setpoints dynamically while respecting equipment constraints.
The analytics stack also highlights anomalies, enabling operators to intervene early when deviations from expected performance occur. Dashboard views created for NTT Data emphasize actionable insights rather than raw data streams.
NTT Data Digital Services and Project Management
NTT Data orchestrated the integration across partners, ensuring that hardware, software, and networking components interoperate smoothly within the test environment. Their experience in large scale data center modernization helped align timelines and risk mitigation measures.
Governance structures defined clear responsibilities for monitoring, incident response, and reporting. Regular review meetings between NTT Data, Daikin, and Start AI supported rapid iteration on control parameters.
Roadmap for Scaling Energy Efficient Cooling Across AI Infrastructure
- Standardize measurement methodologies for cooling performance under AI workloads.
- Integrate control logic with existing building management and orchestration platforms.
- Expand trials to other regions with varying climate conditions to verify robustness.
- Develop service level agreements that guarantee efficiency and reliability targets.
FAQ
Reader questions
How does AI driven cooling differ from traditional data center cooling in this test?
AI driven cooling uses predictive models and real time analytics to adjust setpoints and airflow dynamically, whereas traditional cooling often relies on fixed setpoints and conservative overdesign that can waste energy.
What specific metrics were tracked during the July trial?
Key metrics included power usage effectiveness, inlet air temperature consistency, thermal gradients across server rows, fan energy draw, and refrigerant cycle stability under variable AI loads.
Why was a Japanese data center chosen for this collaboration?
Japan offers stringent efficiency regulations, high density urban facilities, and mature digital infrastructure, making it a representative market for testing scalable cooling solutions.
What are the next steps if the test results meet efficiency targets?
Partners plan to refine algorithms, validate performance across additional sites, and explore commercial offerings that bundle AI optimized cooling as a service for enterprises and cloud providers.