Axelera AI Challenges Nvidia With Europa Chip For High-Performance Edge Computing

Axelera AI has unveiled its Europa AI chip, a high-performance solution designed to compete directly with Nvidia in the edge computing and "AI factory" markets. Set for release in the first quarter of next year, the chip delivers 629 TOPS of performance within a 45-watt power envelope, targeting demanding workloads like large language models and advanced computer vision. This launch represents a significant expansion for the Netherlands-based startup, offering a high-tier alternative to its existing Metis lineup for use in robotics, surveillance, and edge servers.

Axelera AI’s newly announced Europa chip aims to disrupt the high-performance edge sector by offering superior performance-per-dollar and performance-per-watt compared to Nvidia’s L40 and Jetson products. Delivering 629 trillion operations per second (TOPS) in a 45-watt power envelope, the Europa chip represents a major leap over the company’s previous Metis lineup, which maxes out at 214 TOPS. The hardware is specifically engineered for complex tasks such as running Llama 3 models and high-density computer vision analytics in environments like stadiums and automated warehouses.

To ensure broad market adoption and ease of integration, Axelera AI opted for PCIe Gen 4 connectivity rather than the newer Gen 5 standard. According to CMO Alexis Crowell, this decision allows the chip to be retrofitted into existing workstations and edge servers, helping customers leverage current capital expenditures while upgrading their AI capabilities. The Europa architecture features eight AI processing units and 64 GB of memory, quadrupling the capacity of its predecessor, and includes 16 RISC-V vector processing cores to handle pre- and post-processing tasks directly on the chip, reducing the need to offload work to a system CPU.

Beyond hardware specs, Axelera AI is emphasizing software consistency by utilizing the same Voyager SDK across both its Metis and Europa product lines. This unified software approach is intended to simplify deployment for partners managing mixed use cases across different edge environments, such as humanoid robotics or AI-driven analytics. With a workforce of over 200 people—half of whom are focused on software—the company positions Europa as a versatile tool for "AI factories" and enterprise-grade rack-mount servers, aiming to lower the total cost of ownership for high-performance AI deployments starting in Q1 of next year.