As artificial intelligence infrastructure expands rapidly, rising GPU temperatures are becoming a critical constraint on computing performance, energy efficiency and profitability. LiquidJet Diamond Coldplate technology promises to address this challenge through advanced liquid cooling.
The newly introduced cooling solution reportedly delivers 30–35% more AI tokens per watt, potentially transforming the economics of hyperscale data centres and AI factories operating thousands of high-performance GPUs.
According to the company's performance claims, LiquidJet Diamond can reduce GPU temperatures by 16–22°C compared with traditional skived coldplates, enabling processors to operate under demanding computational workloads with improved thermal management.
Its innovative architecture targets extreme GPU hotspots, delivering over 150% higher maximum power-density handling and approximately 75% greater heat-transfer efficiency. These improvements could help overcome thermal limitations increasingly affecting next-generation AI accelerators.
Unlike conventional cooling solutions, LiquidJet Diamond is positioned as a seamless drop-in upgrade, potentially reducing infrastructure modification requirements while improving thermal performance. This approach could accelerate adoption among hyperscalers seeking greater computing efficiency.
The financial implications are significant. AI factories increasingly measure productivity through tokens generated per watt, making energy efficiency a competitive differentiator. Improved thermal management can potentially increase computing output without proportionately expanding electricity consumption.
For hyperscalers operating massive GPU clusters, even modest efficiency improvements can translate into substantial operational savings. However, a 35% improvement in tokens per watt does not automatically mean a corresponding 35% increase in net profits.
The technology also arrives as AI infrastructure operators face mounting pressure from electricity costs, cooling requirements, power availability and sustainability commitments. Advanced liquid cooling is becoming essential for supporting increasingly power-dense accelerator architectures.
For data centre operators, cloud providers and enterprise AI infrastructure companies, thermal innovation could become as strategically important as processor performance. Cooling efficiency will increasingly influence infrastructure investment decisions, deployment density and total cost of ownership.
The larger industry message is clear: the future of AI computing depends not only on faster chips but also on smarter cooling. LiquidJet Diamond's reported performance gains highlight how thermal engineering could unlock additional AI capacity, improve energy economics and reshape hyperscale infrastructure competitiveness.
