Highlights:

  • AWS stated that its new Trainium2 chips and Nvidia’s GB200 NVL72 rack-scale AI supercomputing chip will both benefit from the newly implemented cooling system.
  • The major update from AWS is the introduction of a new liquid cooling system for its AI servers and other high-performance systems.

At its recent annual re: Invent conference, Amazon Web Services (AWS) Inc. announced significant updates to its cloud computing infrastructure. AWS is integrating a new liquid cooling system and optimizing the power distribution across its data centers, among other innovations aimed at improving efficiency and sustainability.

The cloud computing leader stated that these upgrades make its infrastructure services four times more effective than traditional on-premises environments, positioning it to better manage the growing demands of next-generation artificial intelligence tasks. This improved efficiency will help AWS meet the increased computational needs of AI applications moving forward.

The company also emphasized that its data centers are not only more efficient but can reduce the carbon footprint associated with AI and other computing workloads by up to 99%. This represents a significant step towards achieving more sustainable and eco-friendly operations within the tech industry.

The most significant update AWS is introducing is the implementation of a new liquid cooling system designed for its AI servers and other high-performance systems. These include infrastructure powered by AWS’s Trainium chips and Nvidia’s advanced graphics processing units, enhancing the efficiency and performance of its data centers.

AWS revealed that its new Trainium2 chips and Nvidia’s GB200 NVL72 rack-scale AI supercomputing chip will leverage the recently introduced liquid cooling system. This system is designed to improve efficiency, reducing mechanical energy consumption by up to 46% during peak cooling conditions, while requiring no additional water resources.

AWS’s cooling infrastructure is designed to remain adaptable, with the new system capable of using air cooling when liquid cooling isn’t necessary. For instance, servers handling regular computing tasks, networking, and storage—typically powered by less energy-intensive central processing units (CPUs)—are unlikely to rely on liquid cooling. These CPUs generate significantly less heat, making liquid cooling unnecessary and air cooling sufficient for efficient operation.

The company says, “This flexible multimodal cooling design allows AWS to provide maximum performance and efficiency at the lowest cost, whether running traditional workloads or AI models.”

AWS has introduced a new streamlined electrical distribution system designed to enhance efficiency and reliability by reducing potential failure points by up to 20%. This improvement is achieved by minimizing the number of electrical conversions required, positioning backup power sources closer to the data center racks, and reducing reliance on fans for expelling hot air. Instead, the system uses a natural pressure differential to expel heat, allowing more energy to be directed toward servers and lowering overall power consumption.

AWS is rolling out a new power shelf design that will significantly boost rack power density. Over the next two years, the design is expected to achieve a six-fold increase, with an additional three-fold boost anticipated later. This advancement will enable a 12% increase in computing power at each AWS facility, allowing the company to optimize its operations and reduce the number of data centers it needs to maintain.

AWS has revealed that many of these upgrades are aimed at supporting AI workloads, while also leveraging AI technologies to enhance the efficiency of its data center designs. By developing AI models, AWS can optimize rack configurations within its facilities to minimize power wastage and improve utilization. Additionally, AI will play a key role in a new control system for managing electrical and mechanical devices in its data centers. This system will incorporate integrated telemetry services, enabling real-time diagnostics and more effective performance monitoring.

AWS is working to reduce its carbon footprint by switching to a new renewable diesel for its backup generators. This new fuel is described by the company as biodegradable and non-toxic, and it can reduce carbon emissions by up to 90% over its lifetime when compared to conventional fossil fuel-based diesel. This move is part of AWS’s broader commitment to sustainability and reducing its environmental impact.

AWS has announced plans to make its data center construction more sustainable by incorporating lower-carbon steel and concrete. The steel will be produced using electric arc furnaces instead of traditional gas-fired furnaces, which will significantly reduce the carbon emissions associated with its production.

AWS has reduced the carbon content in its concrete mix by up to 35% compared to the industry standard. Additionally, the company is minimizing its overall use of steel and concrete by leveraging a new AI-optimized rack placement system, which allows for more efficient use of materials in its data centers.

AWS plans to implement these changes globally across all new data centers it builds, with some of the upgrades also being introduced to its existing facilities.

“These data center capabilities represent an important step forward with increased energy efficiency and flexible support for emerging workloads. 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,” said Prasad Kalyanaraman, Head of AWS Infrastructure Services.