Independent Benchmarks Show Accelsius Two-Phase Direct-to-Chip Cooling Delivers 9°C Lower NVIDIA B200 Junction Temperatures Than Single-Phase in Warm-Water Conditions

Key Terms

direct-to-chip liquid cooling technical

A cooling method that pumps liquid directly over the surface of a computer chip (processor or graphics unit) using a fitted plate, removing heat more efficiently than air or indirect systems. Think of it like a water jacket on an engine: keeping the hottest part cooler lets the equipment run faster, longer, and in a smaller space, which can lower energy and space costs and enable higher-performance products—factors that affect operating margins, capital needs, and adoption potential for companies and data centers.

high-bandwidth memory (HBM) technical

High-bandwidth memory (HBM) is a type of computer memory engineered to move very large amounts of data quickly between memory and processors by stacking memory chips and using wide, short connections for faster, more efficient transfers. It matters to investors because HBM can significantly boost performance and lower power use in data centers, AI systems and high-end chips, affecting demand and profitability for chipmakers and their suppliers — like widening a highway so more traffic moves faster with less congestion.

lpm/kw technical

lpm/kw is a performance ratio showing how many liters per minute of fluid (usually air or gas) a machine moves for each kilowatt of power it consumes. It is a simple measure of energy efficiency for pumps, fans, compressors and medical ventilators — higher values mean more flow for the same power, like miles per gallon for a car. Investors use it to compare product efficiency, running costs and competitive positioning across manufacturers.

free cooling technical

Free cooling is a way to cool buildings or data centers using outside air instead of relying on expensive air conditioning systems. It saves energy and money by taking advantage of naturally cooler weather, much like opening windows on a cool day to lower indoor temperature. This method helps reduce overall costs and is better for the environment.

junction temperatures technical

Junction temperature is the internal temperature at the active semiconductor junction inside an electronic component, like where current actually flows in a chip or transistor. It matters to investors because this temperature controls a device’s performance, reliability and lifespan—higher junction temperatures can cause slower operation, greater failure risk, or the need for more cooling, which affects product quality, warranty costs and manufacturing requirements.

See more from StockTitan in Google Search and AI answers. Adds StockTitan as a preferred source · opens Google

Add on Google Not now

Third-party testing across roughly 40,000 operating points, detailed in a new white paper, found that Accelsius NeuCool® improved liquid cooling performance at warm facility water temperatures from NVIDIA's new standard of 45°C to as hot as 54°C.

AUSTIN, Texas--(BUSINESS WIRE)-- Accelsius, the leader in two-phase, direct-to-chip liquid cooling technology for AI and high-performance computing, today announced results from independent third-party benchmark testing demonstrating that two-phase direct-to-chip liquid cooling better supports NVIDIA’s goal of enabling global free cooling with warm facility-water temperatures. The results show that single-phase performance at 45°C facility water can be matched by Accelsius with 54°C inlet water temperatures.

As AI infrastructure expands into hotter climates and operators seek to reduce energy consumption and local community impact, the ability to maintain GPU performance at higher facility water temperatures is becoming a critical requirement for purpose-built AI factories. Every 1°C increase in facility water temperature equates to approximately 4 percent in annual energy savings. By enabling operation with facility water temperatures beyond 50°C, NeuCool enables free cooling in most locations around the world for much of the year.

Independent benchmarking was conducted by a third-party systems integrator using a commercially available eight-way NVIDIA B200 GPU server. The evaluation compared the server’s factory-installed single-phase direct-to-chip cooling system with Accelsius’ NeuCool two-phase technology across approximately 40,000 operating points under simulated AI workloads. By varying GPU core, high-bandwidth memory (HBM) and CPU power, the testing created a comprehensive comparison of thermal performance across realistic operating conditions.

An additional advantage of two-phase direct-to-chip cooling is how well it capitalizes on higher facility water flow rates. Raising facility water flow from 1.5 liters per minute per kilowatt (LPM/kW) to 3.0 LPM/kW widened the thermal advantage over single-phase from 9°C to 14°C, enabling previously unattainable facility water supply temperatures of 59°C. Single-phase cooling does not realize comparable performance gains from increased facility water flow.

Leading ecosystem players, from server OEMs to chip manufacturers, recognize the advantages two-phase will bring to future generations of data center operators.

To measure the effects of two-phase, direct-to-chip cooling in warm water facilities, Accelsius has detailed these technical improvements in a whitepaper called “Warm-Water Ready: Benchmarking Two-Phase vs. Single-Phase Direct-to-Chip Cooling on an NVIDIA B200 Server.”

NVIDIA CEO Jensen Huang recently highlighted the industry’s move toward warmer facility water temperatures as a key opportunity to improve AI infrastructure efficiency. As Accelsius has long maintained, the physics of vaporization allow two-phase cooling to realize disproportionately greater performance benefits as facility water temperatures rise. The independent benchmark results validate that advantage, demonstrating increasing separation between two-phase and single-phase cooling under the warm-water conditions required for global free cooling.

About Accelsius

Founded by Innventure, Inc. (NASDAQ:INV), Accelsius empowers data center operators and AI neoclouds to achieve their business, financial and sustainability goals through advanced cooling solutions. The proprietary NeuCool® platform provides best-in-class thermal efficiencies through a safe, two-phase, direct-to-chip liquid cooling system that scales from single racks to entire data centers. For more information, visit accelsius.com or follow us on LinkedIn.

View source version on businesswire.com: https://www.businesswire.com/news/home/20260721991081/en/

media@accelsius.com

Source: Accelsius