Axelera AI launches Europa AIPU, a purpose-built inference accelerator with software toolchain for enterprise and data center deployment.

Axelera AI announced the launch of the Europa architecture on September 15, 2026, with a growing ecosystem of OEM, ODM and technology partners, including validated systems with Dell and Supermicro. Europa is a next-generation AI Processing Unit (AIPU) that extends Axelera's edge-first architecture to more demanding enterprise and data center workloads, giving organizations a practical path to run production AI on infrastructure they control.

With a sales pipeline exceeding $1.5 billion, Axelera's performance-per-watt leadership is driving customer wins across AI factories and gigafactories in Europe. The company has deployments with more than 600 customers globally, spanning defense, robotics, drones, retail and security applications.

Inference represents the bulk of the AI compute load, yet rising cloud costs, exponential token usage, power constraints and increasingly capable smaller models are shifting the economics of where AI should run. Many enterprise applications don't require the largest frontier models or most expensive hardware—they need infrastructure designed to run persistent AI workloads securely and sustainably. Europa delivers up to 6x the tokens per second per watt compared to GPU-based solutions, enabling customers to deploy reliably within their own environments without sacrificing performance.

Europa gives organizations greater control over where AI runs, where their data goes and what it costs—particularly critical for Physical AI deployments and markets such as financial services, healthcare, legal, defense and government, where security, compliance and data sovereignty requirements can limit public cloud AI use.

"Dell Technologies is working with AI systems integrator E4 and Axelera AI to support the deployment of next-generation AI infrastructure for the EU AI Factories in Italy and Luxembourg," said Peter Barnes, General Manager EMEA HPC at Dell Technologies. "Together, the collaboration brings the performance, control and sovereignty organizations need to advance AI innovation while keeping critical infrastructure and data aligned with Europe's digital autonomy goals."

Fabrizio Del Maffeo, CEO and co-founder of Axelera AI, said: "AI is an imperative at most companies, but ensuring that an enterprise gets the highest return from infrastructure while controlling their data is imperative. We built our architecture around some of the hardest constraints in computing: power, energy, cost and the need to process data locally. Europa applies those same principles, expanding from Physical to Enterprise AI, giving organizations the performance they need for increasingly sophisticated workloads while keeping control of their data, infrastructure and economics."

Europa is designed for the growing mix of AI workloads, including agentic systems, vision-language models, generative AI and machine and computer vision. It gives organizations a practical way to add AI inference capacity in existing server environments while addressing real-world power, cost, security and operational constraints. Built on the same architecture principles proven across Axelera's existing Embedded, Edge and Server portfolio, Europa is available as a chip for custom board design, as a half-height half-length PCIe card (Axelera Edge 232p) or a full-height full-length PCIe card (Axelera Server 250p). The standard PCIe form factors enable organizations to add AI capacity without rebuilding infrastructure around a new platform.

Shipping now, Europa is supported by Axelera's growing ecosystem of OEM, technology, software and industry partners. The Axelera Edge 232p is available in a fully validated Dell XE5 and Supermicro 111AD, joining validated systems with global OEMs including Advantech, Axiomtek, HPE, Lenovo and Seco, with additional systems to be announced soon.

Axelera's Voyager Toolchain helps developers optimize existing AI models for Axelera hardware and move quickly from development to production, supporting all Axelera hardware with a consistent foundation for deploying AI inference from edge devices to enterprise infrastructure. Developers can also use Voyager Wingman, an agentic platform that can port existing inference pipelines, create new pipelines and optimize code.

Model support has been a limiting factor for many customers and silicon providers. Axelera recently disclosed AxeleraScript (AxScript), a novel approach to model compilation—the first of its kind—to address this gap.

Fabrizio Del Maffeo presented the infrastructure shifts driving the next phase of enterprise and Physical AI at AI Infra Summit in Santa Clara on September 15, speaking on "The Physical AI Inflection Point: Industry Shifts, Adoption Barriers, and the Infrastructure Imperative" from 11:00–11:20 a.m., with Europa demonstrations available at booth 930.