Semidynamics and SiPearl have entered into a strategic cooperation to develop a European rack-scale AI compute platform aimed at large-scale AI inference in cloud environments. The effort combines SiPearl’s Arm-based CPU technology with Semidynamics’ RISC-V-based GPU/AI inference ASIC, targeting what the companies describe as an EU-sovereign approach to rack-scale AI infrastructure.
The planned platform splits roles along familiar system boundaries: SiPearl’s Arm-based CPU is slated to handle general-purpose compute, orchestration, and data plane hosting, while Semidynamics’ accelerator is intended to be the primary inference engine, with headroom for future performance scaling. The rack design is expected to follow Open Compute Project (OCP) standards, positioning the system for interoperability with established cloud and data center infrastructure practices.
For data center engineers, the most concrete architectural takeaway is the intent to deliver a rack-scale reference system that integrates host CPU, accelerator, and enclosure design as a coordinated package, rather than leaving integration to an OEM or operator. That can simplify qualification, but it also puts a spotlight on practicalities the market will care about quickly, including how the platform’s power, cooling, and serviceability characteristics behave under sustained inference throughput.
The cooperation is structured in stages. In the first iteration, SiPearl will provide its Arm-based CPU technology and platform support for host compute and orchestration. Semidynamics will provide its RISC-V-based GPU/AI inference ASIC, accelerator enablement, and the design of the enclosure and rack solution integrating both technologies. As a second step, the companies said further integrations at the chiplet level will be disclosed. They also plan to jointly manage the reference architecture, marketing materials, and coordinated tender responses for European procurement opportunities, including initiatives referenced as AI Factory and Giga Factory programs.
SiPearl’s first-generation CPU, Rhea1, is described as featuring 80 Arm Neoverse V1 cores and 61 billion transistors, and is currently being manufactured. SiPearl also said its CPUs will equip two EuroHPC JU exascale supercomputers: Rhea1 in JUPITER (Germany) and Rhea2 in Alice Recoque (France).
“SiPearl is thrilled to see the impact of years of work in the European Processor Initiative and the EU sovereign ecosystem come to fruition with this platform,” said Philippe Notton, SiPearl’s CEO and Founder. Roger Espasa, CEO of Semidynamics, said, “Combining SiPearl’s high-performance CPU with Semidynamics’ RISC-V-based GPU/AI inference technology gives Europe a credible path towards sovereign, rack-scale AI infrastructure built around European-controlled compute.”
Source: Semidynamics
