News

Startups

Artificial Intelligence

Asia

Panmnesia and Openchip Form Strategic Alliance at MWC26 to Advance AI Infrastructure

Panmnesia and Openchip have signed a Memorandum of Understanding at MWC26 to co-develop high-performance AI and HPC infrastructure solutions.

Panmnesia and Openchip have signed a Memorandum of Understanding at MWC26 to co-develop high-performance AI and HPC infrastructure solutions.

NewDecoded

Published Mar 7, 2026

Mar 7, 2026

4 min read

Image by Panmnesia

Panmnesia announced a strategic partnership with Openchip at the Mobile World Congress (MWC26) in Barcelona to revolutionize AI and High-Performance Computing (HPC) infrastructure. The Memorandum of Understanding (MOU) establishes a technical collaboration focused on enhancing performance, scalability, and resource efficiency within data center environments. Both companies aim to integrate advanced link technologies with open-standard processors to solve critical data movement challenges.

The collaboration leverages Panmnesia's expertise in Compute Express Link (CXL) and Openchip's unique portfolio of RISC-V based accelerators. By joining forces, the firms intend to explore new system architecture configurations that allow for seamless connectivity between AI accelerators and infrastructure resources. This approach is designed to meet the growing demands of large-scale AI models that require massive memory and high-speed communication.

Myoungsoo Jung, CEO of Panmnesia, emphasized that link technologies are becoming critical to overall system performance as AI environments evolve. He stated that Panmnesia will continue to advance next-generation solutions through global partnerships to support the rapidly changing infrastructure environment. Cesc Guim, CEO of Openchip, added that the collaboration is vital for building a scalable and efficient AI infrastructure using open standards.

A primary goal of the partnership is to tackle the "Memory Wall" by disaggregating resources like GPUs and memory. Using Panmnesia's CXL 3.2 fabric switches, data centers can create pools of shared memory that are accessible to multiple RISC-V accelerators. This method significantly reduces the total cost of ownership for operators while boosting the performance of complex AI training and inference tasks.

Panmnesia is a leader in advanced link solutions, offering a portfolio that includes Link Controllers and Fabric Switches. The company was recently recognized for its innovation at both CES 2024 and CES 2025 for its reference architecture. Their technology enables flexible memory expansion and resource sharing without requiring additional GPU deployment in large-scale architectures. Visit https://panmnesia.com/ for more information.

Openchip is a Barcelona-based systems company focused on developing RISC-V based accelerators and full-stack hardware infrastructure. Their products are designed to advance digital sovereignty and provide best-in-class performance for Europe's critical computing needs. By focusing on open-standard silicon, they aim to define the next generation of supercomputing and AI applications. Learn more at https://openchip.com/ for technical details.


Decoded Take

Decoded Take

Decoded Take

This strategic partnership represents a significant shift in the AI hardware landscape by challenging the dominance of proprietary ecosystems like NVIDIA's NVLink. By combining South Korean CXL 3.2 connectivity with European RISC-V compute power, the alliance builds a credible, open-standard alternative that addresses the "Memory Wall" bottleneck. For the industry, this signals that CXL technology has matured into a deployable reality capable of supporting digital sovereignty goals in Europe while providing the scalability needed for next-generation large-scale AI models.

Share this article

Related Articles