A new high-performance server platform promises to change how artificial intelligence (AI) infrastructure is built and cooled, offering a scalable way to power increasingly dense compute systems without the need for data centre overhauls.
Technology firms ZutaCore and ASRock Rack have announced the launch of the ASRock Rack 4U16X-GNR2/ZC, a system that integrates the NVIDIA HGX B300 platform with ZutaCore’s HyperCool waterless direct liquid cooling technology. The turnkey server, was unveiled ahead of its live showcase at the OCP Global Summit in San Jose, is designed to support up to eight NVIDIA HGX B300 systems in a standard rack configuration while maintaining performance and reliability in high-intensity AI environments.
The launch reflects a wider shift across the data centre industry as AI’s appetite for power, performance and cooling pushes existing infrastructure to its limits. Traditional air and water-based cooling systems are struggling to cope with the thermal output of next-generation chips, particularly those used to train large-scale AI models. The new system aims to provide a factory-integrated, production-ready solution that reduces complexity and risk in deploying such hardware.
Cooling for the AI era
At the core of the new platform is ZutaCore’s two-phase, waterless cooling loop, which uses a non-conductive dielectric fluid to absorb and remove heat directly from the chip. Unlike water-based systems, the dielectric fluid poses no risk to electronics if a leak occurs and eliminates the corrosion and maintenance issues associated with traditional liquid cooling.
This approach allows for direct-to-chip heat removal from CPUs, GPUs and other power-hungry components, providing consistent temperature control and higher compute density. According to ZutaCore, this enables operators to run more powerful AI workloads within existing data centre footprints without the need for costly retrofits or expansion.
As artificial intelligence scales up, with data centres now consuming unprecedented amounts of energy, efficient heat management has become a decisive factor in sustaining growth. The collaboration between ASRock Rack and ZutaCore highlights the growing demand for integrated, pre-certified systems that shorten deployment timelines and meet stringent performance targets in an increasingly power-constrained market.
From hardware challenge to strategic advantage
The system’s turnkey design aims to simplify what has traditionally been one of the most complex aspects of AI infrastructure deployment. Delivered fully assembled and factory tested, it eliminates the need for on-site customisation or specialist integration work, allowing operators to accelerate installation and reduce operational downtime.
By integrating the NVIDIA HGX B300 platform, a high-performance computing standard for AI training and inference, within a waterless liquid-cooled environment, ASRock Rack and ZutaCore are positioning their joint system as a reference point for the next generation of AI data centres. The companies argue that efficient thermal design is no longer just an engineering problem but a business imperative, with energy costs, sustainability commitments and uptime now determining competitive advantage.
The ZutaCore HyperCool technology represents one of several new cooling innovations gaining traction as chipmakers and system integrators adapt to AI’s rising thermal footprint. With AI models demanding exponential increases in compute power, the ability to cool processors safely and efficiently has become central to maintaining performance per watt, a key metric for both operators and regulators.
As AI continues to reshape global computing infrastructure, the race to develop reliable, scalable cooling solutions may prove just as transformative as advances in silicon design. For an industry under growing scrutiny over energy consumption, innovations like these could mark an important step toward building data centres that are not only more powerful but also more sustainable.
