Accelsius LLC

What Accelsius does

• Designs and manufactures NeuCool, a complete two-phase, direct-to-chip liquid cooling system.
• Helps operators support current and future generations of high‑TDP CPUs and GPUs.
• Focuses on mission-critical reliability, uptime, and ease of service in live data centers.
• Works with partners across the data center ecosystem (CDU manufacturers, server OEMs, channel and deployment partners, and colocation providers) to deliver end-to-end solutions.

Who NeuCool is for

NeuCool is intended for:

• AI and machine learning infrastructure
• High-performance computing (HPC)
• High-density enterprise and cloud data centers
• Colocation providers supporting GPU-heavy workloads

Company background

• Formed in 2022 to commercialize a two-phase liquid cooling platform based on extensive thermal science research.
• Founded by Innventure and led by an executive team with deep experience in servers, IT infrastructure, data centers, and thermal engineering.
• Focused on enterprise-grade engineering, operational innovation, and sustainable technologies for mission-critical environments.

Location and contact

• Headquarters: 1835-B Kramer Lane, Suite 2-180, Austin, TX 78758, USA
• General email: info@accelsius.com
• Website: https://accelsius.com

To request more information, arrange a demo, or inquire about deployments, customers are directed to use the contact form on the Accelsius website or email the company directly.

Accelsius describes an all-inclusive support model designed to enable peak NeuCool performance, from initial planning through ongoing operations. Typical services include:

• Pre-production site assessment and solution design modeling to ensure NeuCool integrates smoothly with existing data center infrastructure.
• Tailored solution designs for specific power densities, rack layouts, and facility conditions.
• Contract manufacturing through service partners to support hyperscale and large multi-rack deployments.
• Step-by-step project validation by thermal experts for stress-free assembly, commissioning, and handover.

Manufacturing and supply chain

• NeuCool products are manufactured and assembled in the United States.
• A large share of Tier 1 vendors is based in North America, supporting supply chain resilience.

Warranty and protection

Accelsius references the following warranty and protection elements for NeuCool deployments:

• NeuGuard™ limited warranty options (for example, 2‑, 3‑, and 5‑year limited warranty options backed by CNA, as described on the website).
• Additional server and integration warranties available through partners such as Unicom and Park Place.

Customers should confirm current warranty terms, coverage, and durations directly with Accelsius or its partners, as these may vary by product, configuration, and region.

Operations, maintenance, and SLAs

To keep NeuCool systems operating reliably in mission-critical environments, Accelsius highlights:

• Preventative and corrective maintenance services aimed at maintaining optimal performance.
• Tailored service level agreements (SLAs) for full lifecycle management.
• Remote and on-site troubleshooting to resolve issues quickly.
• Technician dispatches for routine site inspections, training, and asset support.

Specific service levels, response times, and coverage details should be confirmed with Accelsius as part of a customer's contract or SLA.

Global services and support

• Accelsius works with global service partners to support deployments in multiple regions.
• Customers can request help, schedule services, or learn more about deployment options by using the contact form on the Accelsius website or emailing info@accelsius.com.

• Accelsius develops NeuCool, a two-phase, direct-to-chip liquid cooling system specifically for data centers.
• Unlike single-phase water cooling or immersion cooling, NeuCool uses a dielectric refrigerant that boils directly on the chip to remove heat.
• This approach is described as enabling much higher heat removal per watt and supporting very high-TDP AI and GPU processors while reducing energy and water use compared with traditional cooling.

How does NeuCool compare with immersion cooling?

From the FAQ themes:

• Direct-to-chip focus: NeuCool targets heat directly at the processors rather than submerging the entire server in fluid.
• Less fluid and easier integration: Direct-to-chip uses much less fluid and works with standard racks and servers, avoiding many of the weight, retrofit, and maintenance trade-offs of immersion tanks.
• Serviceability: Dripless quick disconnects and hot-swappable components are intended to make servicing easier in live data centers.

Ecosystem partners and compatibility

• Accelsius works with partners across the liquid cooling value chain, including CDU manufacturers, server OEMs, channel and deployment partners, and colocation providers.
• NeuCool is designed to integrate with industry-standard coolant distribution units (CDUs) in both liquid‑to‑liquid and liquid‑to‑air configurations.
• It is intended to work with standard server and rack form factors used in AI and high-performance computing environments.

Facility water, operation, and efficiency

• NeuCool is designed so that facility water connects to the CDU/iCDU, not directly to the chips.
• The chips only contact the dielectric refrigerant, not water.
• Because of its large thermal margin, NeuCool can operate with warmer facility water than some single-phase direct‑to‑chip systems, which can increase energy efficiency and the potential for free cooling.

Serviceability and maintenance

• NeuCool is described as being built for serviceability in live data centers, including:
• Dripless quick-disconnect fittings
• Hot-swappable components
• Designs that minimize the need to drain loops or take racks offline for routine service

Deployment scale and order sizes

• Accelsius indicates that NeuCool is in volume production.
• The company supports deployments ranging from single-rack trials to large multi-rack rollouts, with manufacturing capacity intended to scale with AI-driven demand.

How to see NeuCool in action

According to the FAQ themes and company information:

• Customers and prospects can see NeuCool operating at the Accelsius lab in Austin, Texas, either via onsite or virtual demonstrations.
• Organizations may apply to the NeuCool Field Trial Program to evaluate the system in their own environment.
• Accelsius also showcases NeuCool at major data center and AI infrastructure events and conferences.

To request a demo, inquire about field trials, or arrange a lab visit, customers are directed to use the contact form on the Accelsius website or email info@accelsius.com.

• Two-phase direct-to-chip cooling: A dielectric refrigerant boils directly on the chip surface, removing large amounts of heat at a stable temperature and supporting very high power densities.
• Dielectric refrigerant: Non-conductive and non-flammable, so a leak does not damage IT equipment.
• High thermal performance:
• Cools 4500W+ per socket.
• Supports rack densities up to 250kW.
• Features approximately 0.020°C/W thermal resistance at 700W+ TDP.
• Scalability: Designed to cool current and next-generation CPUs and GPUs used in AI and HPC, extending the runway for higher-density systems without major facility redesigns.
• OpEx and TCO impact (as stated by Accelsius):
• NeuCool's design and thermal performance enable 35–44% annual OpEx savings compared with single-phase direct-to-chip cooling.
• Accelsius cites 5-year TCO savings of 8–17% versus single-phase direct-to-chip designs, depending on region.

NeuCool products

Accelsius offers multiple products built on the NeuCool platform for different deployment scenarios:

MR250

• In-row cooling system designed to deliver up to 250kW across up to four racks.

Multi-GPU solutions

• Configurations designed specifically to cool advanced GPU and CPU combinations used in AI and high-performance computing.

TSR (Thermal Showcase Rack)

• A "thermal bank" or showcase system that allows customers to evaluate NeuCool in their own environment.

IR150

• A reimagined in-row cooling solution configured for the mission-critical liquid cooling era.

Key system components

NeuCool iCDU (Intelligent Coolant Distribution Unit)

• Central control and distribution unit for the NeuCool loop.
• Includes core logic, integrated industrial condenser, N+1 pumps, and system sensors.
• Provides a touchscreen control interface and supports remote monitoring.
• Integrates with a variety of heat rejection options on the facility side.
• Uses redundant, hot-swappable components (including power supplies, pumps, sensors, and core logic board) to maximize uptime.

NeuCool Vaporator

• Direct-to-chip heat exchanger optimized for two-phase boiling on CPUs and GPUs.
• Capable of cooling up to 4500W per socket, supporting current and next-generation processors.

NeuCool Manifolds

• Aluminum manifolds that route refrigerant between servers and the iCDU.
• Split design for installation flexibility and optimized vapor and liquid flow.
• Low-profile layout to preserve rear-aisle access and avoid PDU interference.
• Designed and tested for 3× normal operating pressure.
• Stainless-steel, dripless quick disconnects for one-handed, audible connections.

NeuCool Reservoir

• Provides refrigerant expansion volume to handle changes in compute load.
• Helps create head pressure that extends pump life and stabilizes fluid flow.
• Incorporates redundant level sensors that feed into the leak detection system.

NeuView – NeuCool Software Management System

• Integrated software platform for monitoring and controlling NeuCool systems.
• Accessible via a local touchscreen LCD on the iCDU or remotely through a web interface or DCIM tools.
• Interfaces with common data center and server management systems (e.g., SNMP, IPMI, Redfish, DCIM), enabling integration with NOC workflows.

Supported processors (as stated on the website)

NeuCool is described as supporting leading high-performance CPU and GPU families, including for example:

• CPUs: Intel Sapphire Rapids, Intel Emerald Rapids, AMD EPYC Genoa, AMD EPYC Turin, NVIDIA Grace
• GPUs: NVIDIA H100, NVIDIA H200, NVIDIA B200, AMD MI325X, AMD MI355X

For the latest list of supported platforms and validated configurations, customers should contact Accelsius directly.

• A dielectric refrigerant circulates through direct-to-chip cold plates (vaporators) attached to CPUs and GPUs.
• Heat from the processors causes the refrigerant to boil (change phase from liquid to vapor) at a controlled temperature.
• This phase change absorbs a large amount of heat (latent heat), allowing NeuCool to remove significantly more heat per unit of fluid than single-phase water systems that only raise fluid temperature.
• Vapor returns to the NeuCool iCDU, where it is condensed back to liquid and recirculated.

This two-phase process allows NeuCool to support very high chip and rack power levels while maintaining tight thermal control.

Safety characteristics of the coolant

Based on the information Accelsius publishes:

• The refrigerant used in NeuCool is non-conductive (dielectric), so it does not cause electrical damage to IT equipment in the event of a leak.
• It is described as non-flammable and formulated for use around live electronics.
• The fluid has low global warming potential (GWP) compared with many traditional refrigerants.
• In the event of a breach, two-phase systems typically operate at relatively low pressure, and the dielectric fluid is designed to vaporize rather than pool on electronics, further reducing risk.

Leak avoidance and detection

NeuCool is engineered with features intended to minimize leak risk and mitigate consequences:

• Leak avoidance by design, including robust materials, pressure ratings, and connection hardware.
• Redundant leak detection sensors integrated with the NeuCool Reservoir and system controls.
• Continuous monitoring via the NeuView software platform.
• If a leak does occur, the dielectric fluid is designed not to damage electronics and to vaporize instead of accumulating on components.

Energy efficiency and facility integration

Accelsius highlights several efficiency-related benefits of its two-phase direct-to-chip approach:

• Higher heat removal per unit of fluid and lower flow rates can reduce pump energy compared with single-phase liquid cooling.
• The large thermal margin of two-phase cooling allows NeuCool to operate with warmer facility water temperatures than many single-phase systems.
• Warmer water support can increase the number of free-cooling hours and reduce the need for mechanical chilling, contributing to lower overall data center PUE.
• NeuCool is designed to interface with industry-standard CDUs and a range of heat rejection options, so it fits into existing liquid cooling and facility water architectures.

Mission-critical design philosophy

Accelsius positions NeuCool as a mission-critical solution, emphasizing:

• Redundant, hot-swappable components (power supplies, pumps, sensors, and control logic) in the iCDU.
• Real-time monitoring and control via local touchscreen and remote interfaces.
• A design focus on maximum uptime, serviceability, and ease of deployment for onsite technicians, with features such as dripless quick disconnects and rear-aisle serviceability.

For detailed engineering specifications, performance test data, and integration guidance, Accelsius advises contacting the company directly or requesting technical documentation through its website.

Ask anyway — the agent will say plainly that it is not published rather than guess.