Key Takeaways
- Siemens, NVIDIA, and Fluence have developed a 136 MW AI factory power architecture reference design for hyperscalers and cloud infrastructure providers
- The design provides an electrical, power, and controls framework for high-density AI infrastructure deployment
- The architecture is scalable, allowing for phased capacity additions from tens of megawatts to larger installations
- The reference design includes nVent-aligned electrical design parameters and Fluence battery energy storage for operational flexibility
Introduction to AI Factory Power Architecture
The rapid growth of AI factories in the data center sector has led to increased demands on power and cooling infrastructure. To address these challenges, Siemens, NVIDIA, and Fluence have collaborated to develop a reference design for NVIDIA DSX Vera Rubin-aligned AI factory power architecture. This design provides a comprehensive electrical, power, and controls framework for hyperscalers, colocation providers, and specialized cloud infrastructure providers.
Technical Specifications
The reference design is sized for a total facility capacity of 136 MW, with an IT load of 100 MW. The architecture includes:
- A nominal 34.5 kV utility connection
- Medium-voltage distribution
- Modular low-voltage power blocks
- Rack interface
The design is based on Tier III concurrent maintainability, allowing for the removal of any single component without affecting IT operations.
Comparison of AI Factory Power Architectures
| Feature | Siemens Reference Design | Traditional AI Factory Design |
|---|---|---|
| Total Facility Capacity | 136 MW | Variable |
| IT Load | 100 MW | Variable |
| Utility Connection | 34.5 kV | Variable |
| Scalability | Phased capacity additions | Full redesign required |
| Maintainability | Tier III concurrent maintainability | Variable |
Benefits of the Siemens Reference Design
The Siemens reference design provides several benefits, including:
- Predictable implementation and deployment timelines
- Flexibility for future IT platforms and changing energy requirements
- Support for high-density AI infrastructure deployment
- Integration with centralized Integrated Data Center Management Suite
- Operational flexibility and reliability through Fluence battery energy storage
Conclusion
The Siemens reference design for NVIDIA DSX Vera Rubin-aligned AI factory power architecture provides a comprehensive and scalable framework for hyperscalers and cloud infrastructure providers. With its emphasis on predictability, flexibility, and reliability, this design is well-suited to support the growing demands of AI factories in the data center sector.
Bottom Line
The collaboration between Siemens, NVIDIA, and Fluence has resulted in a robust and scalable AI factory power architecture reference design, capable of supporting high-density AI infrastructure deployment. With its technical specifications, benefits, and comparison to traditional AI factory designs, this reference design is poised to play a significant role in shaping the future of AI factory power architecture. As the data center sector continues to evolve, the importance of reliable, efficient, and scalable power infrastructure will only continue to grow, making this reference design a crucial component in the development of next-generation AI factories.