Nexo Earn with Nexo
Google launches Brazos liquid cooling system to tackle AI’s overheating problem

Google launches Brazos liquid cooling system to tackle AI’s overheating problem

The tech giant's new rack-mounted cooling solution handles chips drawing over 1,000 watts and will be open-sourced for industry-wide adoption

Google launched Brazos, a rack mounted liquid cooling system designed to help data centers run high power AI and high performance computing chips without rebuilding entire facilities.

The company said next generation AI and HPC chips now routinely exceed 1000 watts of thermal design power, pushing standard air cooling beyond its limits. 

Brazos is designed to solve that gap by letting operators deploy liquid cooled equipment inside existing air cooled data centers.

Advertisement

The system works as a closed loop liquid to air cooling setup. It captures heat from components using liquid, then rejects that heat into the data center’s hot aisle through liquid to air heat exchangers. 

Google said the approach separates the internal IT liquid loop from the facility water supply, allowing operators to install the system one rack at a time.

Brazos includes three cooling units and integrated rack manifolds. Each modular chassis takes up 11 Open Units of rack height and works with Open Compute Project ORv3 racks. The system supports a 60 kW nominal thermal load per rack across three modular units.

The system can run on deionized water or a 25% propylene glycol mixture, operates on 40 to 60 V DC input, and includes leak detection, pressure relief valves, local monitoring, and remote management through Modbus over TCP.

Google said Brazos is generally available, with manufacturing suppliers ready to market and produce the design for broader industry use. 

The company also plans to open source the technical specifications, design principles, and visual assets through industry forums in the coming months.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.

Google launches Brazos liquid cooling system to tackle AI’s overheating problem

Google launches Brazos liquid cooling system to tackle AI’s overheating problem

The tech giant's new rack-mounted cooling solution handles chips drawing over 1,000 watts and will be open-sourced for industry-wide adoption

Share

Add us on Google

Google launched Brazos, a rack mounted liquid cooling system designed to help data centers run high power AI and high performance computing chips without rebuilding entire facilities.

The company said next generation AI and HPC chips now routinely exceed 1000 watts of thermal design power, pushing standard air cooling beyond its limits. 

Brazos is designed to solve that gap by letting operators deploy liquid cooled equipment inside existing air cooled data centers.

Advertisement

The system works as a closed loop liquid to air cooling setup. It captures heat from components using liquid, then rejects that heat into the data center’s hot aisle through liquid to air heat exchangers. 

Google said the approach separates the internal IT liquid loop from the facility water supply, allowing operators to install the system one rack at a time.

Brazos includes three cooling units and integrated rack manifolds. Each modular chassis takes up 11 Open Units of rack height and works with Open Compute Project ORv3 racks. The system supports a 60 kW nominal thermal load per rack across three modular units.

The system can run on deionized water or a 25% propylene glycol mixture, operates on 40 to 60 V DC input, and includes leak detection, pressure relief valves, local monitoring, and remote management through Modbus over TCP.

Google said Brazos is generally available, with manufacturing suppliers ready to market and produce the design for broader industry use. 

The company also plans to open source the technical specifications, design principles, and visual assets through industry forums in the coming months.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.