Axelera AI has introduced Europa, a next-generation AI Processing Unit architecture that extends its edge-focused technology to enterprise and data-center inference, giving organizations more control over their AI infrastructure and data.

Europa is designed for agentic systems, vision-language models, generative AI, and machine and computer vision, with intended applications in financial services, healthcare, legal, defense, and government.

Axelera AI says Europa can deliver up to six times more tokens per second per watt than GPU-based solutions, targeting persistent workloads where energy use, cost, security, and data sovereignty matter.

Europa will be offered as a chip for custom board designs, a half-height, half-length PCIe card called Axelera Edge 232p, and a full-height, full-length PCIe card called Axelera Server 250p, allowing upgrades to existing server environments.

The Axelera Edge 232p is available in validated Dell XE5 and Supermicro 111AD systems, with additional validated systems from 2CRSi, Advantech, Axiomtek, HPE, Lenovo, and Seco.

Dell Technologies is working with AI systems integrator E4 and Axelera AI on next-generation infrastructure for EU AI Factories in Italy and Luxembourg, emphasizing performance, control, and European digital autonomy.

Axelera’s Voyager Toolchain helps developers optimize models and deploy inference across its hardware, while Voyager Wingman can port or create inference pipelines and optimize code; the company has also disclosed its AxScript model-compilation approach.

Axelera AI reports a sales pipeline exceeding $1.5 billion and deployments with more than 600 customers worldwide across defense, robotics, drones, retail, and security.