Nvidia on Monday announced a new AI system designed to reduce water consumption by operating at higher temperatures and requiring less cooling equipment.
The company unveiled the technology during London Climate Week. Josh Parker, Nvidia's chief sustainability officer, said the water consumption challenge for data centers has been "largely solved," according to Axios. The system uses a recirculated liquid mixture containing water and propylene glycol and operates at 113 degrees Fahrenheit.
Nvidia said the higher operating temperature reduces the need for chillers that consume large amounts of energy and water. The coolant runs in a closed loop, allowing facilities to rely less on mechanical cooling systems.
Steve Solomon, Microsoft's vice president of data center engineering, said the technology would eliminate the need for mechanical chillers in many climates, including hot areas such as Arizona.
The announcement comes as large technology companies face increasing questions over the environmental impact of AI infrastructure. Google and Amazon have both defended their water practices in recent weeks amid opposition to new data center projects in some communities.
Amazon said last week that its operations in India had become "water positive," meaning the company replenishes more water than it consumes. Amazon also said its Indian data centers do not use water for cooling, Reuters reported.
Earlier this month, Google published water management standards aimed at reducing the impact of data centers in water-stressed regions. The company said the guidelines include greater use of reclaimed wastewater and efforts to replenish local water supplies.
Concerns over water use have grown alongside the rapid expansion of AI infrastructure. Investor groups earlier this year called on Amazon, Microsoft and Alphabet to provide more information about the water and power consumption of their data centers, Reuters reported in April.
Nvidia declined to discuss the cost of the new systems. The company said the design is intended to lower cooling expenses for operators.
Many existing data centers continue to rely on older cooling technologies. Water use inside facilities represents one part of the broader resource demands associated with AI infrastructure, as electricity generation for data centers can also require significant amounts of water depending on the energy source, Axios reported.
Parker said in a blog post that AI workloads continue to increase and that efficiency improvements are necessary to support rising computing demand.
