Ahead of its annual re:Invent conference, Amazon Web Services (AWS) has announced significant updates to its data center strategy, including the adoption of liquid cooling for its AI servers and a shift towards simplified electrical and mechanical designs. The move is aimed at providing maximum performance and efficiency at the lowest cost, while also reducing the company's carbon footprint.

The most notable of these updates is the introduction of liquid cooling for AWS's AI servers, including those based on its homegrown Trainium chips and Nvidia's accelerators. This move is particularly significant, as it marks a departure from traditional air-cooling methods and is expected to provide a significant boost to performance and efficiency. According to AWS, the new cooling system will be capable of integrating both air and liquid cooling, allowing the company to provide maximum performance and efficiency while minimizing costs.

In addition to the adoption of liquid cooling, AWS also announced that it is moving towards more simplified electrical and mechanical designs for its servers and server racks. This includes the use of DC power to run servers and HVAC systems, reducing the number of times electricity is converted and minimizing energy losses. The company claims that this will enable infrastructure availability of 99.9999% and reduce the potential number of racks impacted by electrical issues by 89%.

AWS's Vice President of Infrastructure Services, Prasad Kalyanaraman, highlighted the significance of these updates, stating that they represent an important step forward in terms of energy efficiency and flexible support for emerging workloads. He also noted that the new designs are modular, allowing AWS to retrofit its existing infrastructure for liquid cooling and energy efficiency to power generative AI applications and lower its carbon footprint.

The company expects the new multimodal cooling system and upgraded power delivery system to support a 6x increase in rack power density over the next two years, with another 3x increase in the future. To achieve this, AWS is also leveraging AI to predict the most efficient way to position racks in the data center, reducing the amount of unused or underutilized power. Additionally, the company will roll out its own control system across its electrical and mechanical devices in the data center, complete with built-in telemetry services for real-time diagnostics and troubleshooting.

Nvidia's Vice President of Hyperscale and HPC, Ian Buck, also weighed in on the announcement, noting that data centers must evolve to meet the transformative demands of AI. He highlighted the importance of enabling advanced liquid cooling solutions to efficiently cool AI infrastructure while minimizing energy use, and expressed excitement about the potential of AWS's liquid cooling rack design to allow customers to run demanding AI workloads with exceptional performance and efficiency.

These updates are significant not only for AWS but also for the broader cloud computing and AI industries. As AI workloads continue to grow in complexity and demand, the need for efficient and sustainable data center designs will only increase. AWS's move towards liquid cooling and simplified electrical and mechanical designs sets a new standard for the industry, and is likely to have a lasting impact on the way data centers are designed and operated in the future.