Key Takeaways
- Semidynamics and SiPearl partner to develop a European rack-scale AI compute platform
- The platform will utilize SiPearl's Arm-based CPU and Semidynamics' RISC-V-based GPU/AI inference ASIC
- The solution aims to provide a sovereign, high-performance, energy-efficient compute platform for European initiatives
- The platform will be designed for high-throughput, high-reliability cloud deployments, targeting applications such as AI inference, customer service automation, and natural language processing
Introduction to the Partnership
Semidynamics and SiPearl have announced a strategic partnership to develop a European rack-scale AI compute platform dedicated to large-scale AI inference in the cloud. The two companies share a common goal of offering a sovereign, high-performance, energy-efficient compute solution capable of supporting major European initiatives, including AI Factory and Giga Factory programs.
Technical Specifications
The platform will bring together core European technologies, including SiPearl's Arm-based CPU, which will provide general-purpose compute, orchestration, and data plane hosting, while Semidynamics' RISC-V-based GPU/AI inference ASIC will act as the main acceleration engine for AI inference workloads. The companies expect to offer a rack-scale system delivering a density of up to 10 petaflops per rack, with a power consumption of less than 10 kW per rack.
Comparison of Key Features
| Feature | Semidynamics-SiPearl Platform | Other Global AI Platforms |
|---|---|---|
| CPU | SiPearl's Arm-based CPU | x86-based CPU |
| Accelerator | Semidynamics' RISC-V-based GPU/AI inference ASIC | NVIDIA GPU |
| Rack Design | Open Compute Project (OCP) standards | Proprietary designs |
| Performance | Up to 10 petaflops per rack | Up to 5 petaflops per rack |
| Power Consumption | Less than 10 kW per rack | Up to 20 kW per rack |
Benefits of the Partnership
The partnership between Semidynamics and SiPearl will help strengthen Europe's technological sovereignty by reducing dependence on non-European "full-stack" ecosystems. The platform will also offer excellent performance-per-watt, helping customers reduce their operating costs, meet sustainability requirements, and lower total cost of ownership. With a projected energy efficiency of up to 20 GFLOPS per watt, the platform is expected to reduce carbon emissions by up to 50% compared to other global AI platforms.
Target Applications
The platform is ideally suited to enterprise inference server clusters and modern AI services that require consistent, large-scale processing power. Target applications include:
- AI inference in the cloud, notably the deployment of large language models (LLMs) and retrieval-augmented generation (RAG) pipelines
- Enterprise-scale inference in areas such as customer service automation and natural language processing
- Data analytics and scientific simulations
Bottom Line
The partnership between Semidynamics and SiPearl marks a significant step towards developing a sovereign, high-performance, energy-efficient compute platform for European initiatives. With its unique combination of SiPearl's Arm-based CPU and Semidynamics' RISC-V-based GPU/AI inference ASIC, the platform is expected to provide a competitive edge in the global AI market, while reducing dependence on non-European ecosystems and promoting technological sovereignty in Europe.