As artificial intelligence workloads accelerate demand for computing power, data center operators face a mounting challenge: how to expand AI and compute capacity when access to additional electricity is increasingly constrained. Hitachi Vantara, the data storage, infrastructure and hybrid cloud management subsidiary of Hitachi Ltd., is demonstrating a new approach to this problem at its data center in Zaltbommel, the Netherlands, where it has deployed one of the world's largest operational Phase Change Material (PCM) installations using technology developed by Dutch thermal energy company Duraflow.
The deployment addresses a fundamental inefficiency in modern data centers: every watt consumed by AI ultimately becomes heat, and managing that heat can require as much as 40% of a data center's total energy needs. In regions where grid congestion or limited access to additional electricity constrains data center expansion, improving infrastructure efficiency offers a path to making more computing capacity available within existing power connections. Hitachi Vantara's installation provides a real-world proof point for this approach at operational scale, showing how organizations can use up to 90% less electricity for thermal management depending on the application, while requiring zero water for cooling.
The installed system delivers a total cooling capacity of 240 kW across two 120 kW PCM units. According to operational monitoring data, the two units collectively consume just 5–8 kW of electricity depending on operating conditions. By comparison, a conventional DX cooling system delivering the same 240 kW cooling capacity would require an estimated 60–80 kW of compressor power. These figures illustrate how PCM-based thermal management can significantly reduce the electricity required for cooling, freeing up more of the data center's available power capacity for computing rather than diverting it to thermal management.
"This project started with a very practical challenge: how do we grow our computing capacity when access to additional electricity is limited," said Simon Ninan, Senior Vice President of Business Strategy at Hitachi Vantara. "We realized that if we could dramatically reduce the amount of power required for thermal management, we could make more of our existing electrical capacity available for computing. By partnering with Duraflow, we are demonstrating how smarter energy management can become an important enabler of AI growth."
Developed in the Netherlands, Duraflow's PCM technology absorbs thermal energy during periods of peak demand before releasing it in a controlled manner when demand decreases. Unlike conventional cooling approaches that rely continuously on electrically driven refrigeration to remove heat, PCM technology manages and stores thermal energy more efficiently without water consumption. "Developing technology is one thing; proving that it can operate at this scale in a live, mission-critical environment is another," said Simon Loerakker, CEO of Duraflow. "Our collaboration with Hitachi Vantara has allowed us to demonstrate the potential of innovation at an unprecedented scale. Together, we are showing that sustainable innovation and the growth of AI infrastructure can go hand in hand."
The deployment is part of Hitachi Vantara's broader focus on helping organizations extract more value from their infrastructure while managing growing energy demands. The company's Virtual Storage Platform One Block includes a 4:1 Data Reduction Guarantee and features Dynamic Carbon Reduction technology, which reduces energy consumption by switching CPUs into eco-mode during periods of low activity, helping reduce CO2 footprint by as much as 40%. The approach also supports Hitachi Vantara's goal of becoming carbon neutral in Scope 1 and 2 GHG emissions by FY2030. "Addressing the energy demands of AI isn't about optimizing one part of the business in isolation," said Courtney Hadden, Sustainability Director at Hitachi Digital. "It's about working in concert across the organization and with innovative ecosystem partners to find efficiencies at every stage. The answer cannot simply be to consume more power. We need to become smarter about how we use the power we already have."
