Two Innovative Technologies Aim to Reduce A.I.’s Carbon Footprint
A wave of startups is developing cost-effective technologies to curb the rising carbon footprint of artificial intelligence (A.I.) data centers. Data centers consume massive amounts of electricity for running A.I. models, a concern highlighted by Google’s recent report which revealed a significant increase in the company’s greenhouse gas emissions since 2019, partly due to A.I. operations.
Energy consumption in data centers has become a critical issue, with industry giants like Microsoft and Amazon investing heavily in building new facilities, which could potentially double the U.S.’s electricity consumption by 2030. According to Abhijit Sunil, a Forrester analyst, there’s now more focus than ever on finding energy-efficient solutions for data centers.
Immersion cooling is one promising technology being explored. Traditional cooling methods, such as air conditioners and fans, are energy-intensive. Immersion cooling submerges data center equipment in a special liquid that absorbs heat, potentially reducing cooling energy requirements by up to 90%, states Peter Boulin of Green Revolution Cooling. This is particularly crucial as new chips, like Nvidia’s GPUs, demand more electricity for A.I. workloads.
Another approach is the development of software to optimize current workloads, such as the platform created by Incredibuild. This technology leverages unused computing power, transforming it into a usable energy source for other tasks. Dori Exterman of Incredibuild notes that this energy optimization software is already in use by companies like Amazon Web Services and Intel, helping reduce carbon emissions through efficient resource utilization.
However, these innovations face challenges. Immersion cooling lacks standardized best practices, which can impede widespread adoption, as per Forrester analyst Sunil. Additionally, the cost and effort of retrofitting existing infrastructure can be a significant hurdle for major data center providers.
Despite these obstacles, as A.I. technology continues to evolve and demand more power, the industry must adopt these energy-efficient technologies on a large scale to meet the growing energy needs while mitigating environmental impact.
Source: These 2 Innovative Techs May Help Reduce A.I.’s Carbon Footprint—If Data Centers Are Ready.
