MPN: Metis M.2 / Metis PCIe / M.2 Max · Axelera AI
Axelera Metis AIPU
Edge AI · 214–856 TOPS · 15 TOPS/W

M.2 2280 · PCIe HHHL · D-IMC + RISC-V · Voyager SDK · ResNet-50 @ 3,200 FPS · Europa next-gen 629 TOPS

Product Lineup

Metis M.2 AI Accelerator

M.2 2280 M-key module for drop-in edge/embedded AI inference in resource-constrained systems.

Metis PCIe 1-AIPU Card

Single-slot HHHL PCIe card for desktops, servers and embedded racks.

Metis PCIe 4-AIPU Card

Four quad-core Metis AIPUs on one card for the most demanding multi-stream vision workloads.

Metis M.2 Max

Unlocks LLM and VLM inference on edge and embedded devices with up to 8GB memory.

Metis Compute Board (RK3588)

Single-board computer integrating a quad-core Metis AIPU with Rockchip RK3588 host.

Europa (Next-Gen)

Next-generation Axelera platform scaling to 629 TOPS with 8K vision and robotics/SLM support.

Specifications

Vendor

Axelera AI
Eindhoven, Netherlands

Architecture

Metis AIPU — quad-core
RISC-V + D-IMC

AI Performance

214 TOPS (1 AIPU)
856 TOPS (4 AIPUs)

Energy Efficiency

15 TOPS/W

Form Factors

M.2 2280 · PCIe HHHL
Compute Board (SBC)

Interface

PCIe Gen3 x4 / x8

Memory

1GB (M.2) · 4/16GB (PCIe)
8GB (M.2 Max)

Power

5–9W (M.2)
8–15W (PCIe)

Throughput

3,200 FPS ResNet-50

Security

Secure Boot · Root of Trust

Software

Voyager SDK / Toolchain

Next-Gen

Europa — 629 TOPS, 8K

Stock

Available

Price

Quote

Overview

Axelera AI is a Netherlands-based (Eindhoven) European edge AI specialist focused on high-efficiency inference — a complement and alternative to GPU-centric acceleration. At the heart of the platform is the Metis AIPU (AI Processing Unit), a quad-core chip combining a proprietary RISC-V orchestration layer with Digital In-Memory Computing (D-IMC) for matrix-vector multiplication, delivering up to 214 TOPS at an industry-leading 15 TOPS/W.

The Metis portfolio spans the M.2 2280 module (1GB DRAM, 5–9W) for embedded integration, the PCIe HHHL card in 1-AIPU (4GB/16GB, 8–15W) and 4-AIPU (856 TOPS) variants for servers and industrial racks, the M.2 Max (up to 8GB) for on-device LLM/VLM inference, and a Compute Board pairing the AIPU with a Rockchip RK3588 host. All products ship with the Voyager SDK for PyTorch/TensorFlow/ONNX model deployment, Secure Boot, and Root of Trust.

Axelera's roadmap extends to Europa, a next-generation platform scaling to 629 TOPS with 8K vision, robotics, and small-language-model support — positioning Axelera among the leading European alternatives in the edge AI accelerator market.

Need Axelera Metis AIPU?

Contact QS Compute for availability, configuration, and volume pricing.

Request Quote