Strategic Collaboration for High Density AI Infrastructure
Siemens Smart Infrastructure and Rittal have entered into a strategic collaboration for tackling the increasing energy requirements of AI-based data centers. With an increase in AI-based workloads, power consumption per rack is often above 100 kW, and it is estimated that it will reach as high as 1 MW per rack by the end of this decade.
Improving Power Distribution Efficiency
The partnership brings Siemens’ expertise in intelligent power systems as well as Rittal’s data centre hardware solutions. The partnership aims to design a system that can support greater density without compromising efficiency. The outcome of the partnership is a next-generation “sidecar” power rack that can be placed in close proximity to the server cabinet in the white space. This design minimises the distance between the power equipment and the servers to ensure minimal losses in efficiency.
Facilitating Modular and Scalable Design
The proposed sidecar architecture is based on a modular architecture, enabling scalability. This is possible due to its ability to easily add capacity, thereby avoiding complex architectural changes. This is expected to help in more scalable growth as AI compute requirements continue to scale. Additionally, the power electronics will be easy to maintain within the data centre.
Prioritising Standardisation and Deployment Speed
With the pace of AI adoption increasing, the importance of faster data center build and expansion cannot be overstated. Siemens and Rittal focus on standardised infrastructure models to deliver faster deployment solutions. This approach, which aims at reducing “time-to-compute,” allows operators to quickly deploy additional computing capacity in multiple places. According to Andreas Matthé, CEO of Electrical Products at Siemens Smart Infrastructure, there is a need to provide smart, reliable, and efficient power solutions quickly. Rittal’s expertise in modularity can support this requirement well.
Improving Energy Performance Metrics
The partnership is expected to bring greater efficiencies, especially with regards to “tokens per watt,” which refers to the extent of productivity achieved per watt of power used by the AI technology. This partnership will help optimize the performance and efficiency at the rack level through better integration with power distribution and computing resources. Such integrations will help prevent inefficiencies and downtime in power supplies.
Future Development and Broader Applications
Although the power rack for sidecars was the first outcome of the partnership, further projects are already in the development stage. However, the primary application for such integrations seems to be for data centers. However, the primary application seems to be for data centers, while the increasing intensity of AI workloads will require greater efficiencies.