Frore Systems has released LiquidJet Diamond, a new AI Factory liquid cooling coldplate featuring an integrated diamond spreader.
The accelerating demand for AI Tokens, coupled with energy scarcity, is spurring hyperscalers to find new ways to increase the efficiency of AI Factories.
LiquidJet Diamond Coldplate boosts AI Tokens/Watt and revenue by 35% compared to traditional skived coldplates, with the increased revenue flowing directly to profit.
LiquidJet Diamond builds on the previously announced Frore Systems LiquidJet coldplate. LiquidJet coldplate employs a unique 3D ultra short-loop multi-stage design which can be precisely customised to the power map of each GPU, including extreme hotspots.
This precision is critical as GPUs grow in complexity and power draw to meet AI demand. LiquidJet delivers 12 oC reduction in GPU die temperature which directly translates to 25% higher Tokens/Watt efficiency.
LiquidJet Diamond adds an additional spreader with integrated diamond wafers to the coldplate bottom, positioned precisely at the hotspot locations.
LiquidJet Diamond delivers an additional 10 oC reduction in hotspot die temperature, boosting increase in Tokens/Watt to 35%.
Founder & CEO of Frore Systems, Seshu Madhavapeddy, said AI Factories are facing major performance barriers.
"GPUs are increasing power density and creating extreme hotspots which, if not properly cooled, hinder the GPU from maximising performance," he said.
"LiquidJet Diamond, which features an integrated diamond spreader in addition to innovative 3D ultra short-loop multi-stage design, can cool extreme hotspots up to 770 W/ cm2, unleashing high-performance AI GPUs and delivering even more AI Tokens/Watt."
Madhavapeddy said when each GPU is more efficient, each AI Token costs less to produce, and more AI Tokens can be generated within the available power budget, using the same AI Factory infrastructure.
"This increases AI Factory profitability, while conserving local resources," he said.
Power available to AI Factories is finite and expensive to expand. Utility interconnection, transformer capacity, and physical power delivery infrastructure are all limiting factors on total AI Factory power draw and are areas which have local communities increasingly concerned.
Facing this power constraint, AI Factory operators understand that power is a strategic asset, and every watt allocated must be used to generate tokens as efficiently as possible.
"If each AI Factory could produce 35% more AI Tokens with the same resources by deploying LiquidJet Diamond, hyperscalers would reach their growth goals faster with fewer AI Factories," he said.
"Each new AI Factory costs tens of billions of dollars to build and more to operate, so the ability to generate more AI Tokens/Watt with existing resources can significantly increase profits, conserve energy and make a significant difference to all stakeholders."
LiquidJet Diamond performance compared to a skived coldplate:
- Hotspot cooling: over 150% higher max power density
- 75% higher heat transfer efficiency
- 16 - 22 ºC cooler GPU
- 30 - 35% more AI Tokens/Watt
- Seamless drop-in upgrade
"LiquidJet addresses the industry's most pressing thermal challenges — not only through superior performance and hotspot management, but also through its ability to scale to hyperscaler volumes," he said.
Traditional skiving cannot deliver the exacting thermal performance demanded by today's AI GPUs.
Newer approaches such as ECAM (electrochemical additive manufacturing, also known as 3D metal printing) offer design flexibility but cannot meet hyperscaler quality, reliability, volumes, or timelines.
He said LiquidJet combines the precision of semiconductor fabrication with high-volume manufacturing capability — satisfying both the performance and scale requirements of leading AI data centre coldplate deployments.
Higher thermal performance and high-volume manufacturing are not the only challenges LiquidJet solves. GPU die sizes are increasing dramatically.
Die sizes range from 1,488 mm2 NVIDIA Blackwell Ultra all the way to 46,225 mm2 Cerebras Wafer-Scale Engine, roughly 30x the footprint.
"With LiquidJet, Frore Systems has pioneered the only coldplate manufacturing process capable of supporting not only the power and performance, but also the size requirement of these larger AI GPUs," he said.