AWS Reveals New AI Data Center Designs
As the data center industry adapts to the rapidly evolving computing needs of artificial intelligence, cloud giant Amazon Web Services has unveiled significant redesigns of many of the components and systems in its facilities.
The design changes are meant to ensure that AWS’ data centers can handle the skyrocketing electricity needs of AI while simultaneously becoming more energy-efficient, the company said in a release Monday.
The largest data center user in the world, AWS has made significant changes to cooling and power systems, as well as implementing a new system for how information technology equipment is arranged in data centers to maximize energy efficiency.
These redesigned systems, which will be implemented globally in new and existing data centers, come as AWS and other tech giants are under growing scrutiny for the enormous amounts of electricity required to power data centers amid the AI arms race.
“These data center capabilities represent an important step forward with increased energy efficiency and flexible support for emerging workloads,” Prasad Kalyanaraman, vice president of infrastructure services at AWS, said in a statement. “But what is even more exciting is that they are designed to be modular, so that we are able to retrofit our existing infrastructure for liquid cooling and energy efficiency to power generative AI applications and lower our carbon footprint.”
Firms like AWS, Microsoft, Google and Meta are locked in an AI arms race that Blackstone projects will lead to $2T in global AI investment in the next five years.
The newest IT equipment used for generative AI uses far more power per square foot and produces far more heat than the servers that have traditionally lived in data centers. These higher rack densities require fundamentally different cooling systems, power infrastructure and other supporting equipment than what is available in all but the newest data centers.
Many of the redesigned data center components announced by AWS are meant to better support rising rack densities for AI. Taken together, the company says these innovations enable it to deliver 12% more compute power per site for customer workloads, reducing the overall number of data centers needed to deliver the same amount of compute capacity.
The new systems will also enable AWS to increase rack power density sixfold over the next two years and subsequently triple those densities further in the future, according to the company.
Among the design changes highlighted by AWS are improvements to its data center cooling system, including broader adoption of the liquid cooling needed for the most powerful AI processors. The cooling improvements are expected to reduce mechanical energy consumption by up to 46% compared to its previous design during peak cooling conditions, without increasing water usage on a per-megawatt basis, according to AWS.
The company is also rethinking how servers are arranged inside the data center, using AI to develop a new configuration that the company claims will maximize the use of available power.
The company also touts a simplified power management system that it says will improve efficiency by eliminating the energy loss that occurs as electricity goes through multiple conversion and distribution systems in a data center before reaching IT equipment.
Other design changes are focused more directly on sustainability. AWS plans to reduce the amount of embodied carbon in the concrete used in its data center buildings by up to 35% and is standardizing backup generators that can run on renewable biodiesel.
