Axelera AI 1GB Quad-Core Metis M.2 M-Key Accelerator card (214 TOPS)

Axelera AISKU: AXE-BME20M1AB01A01-1GB-M.2-M-QuadCoreMetis-2
Delivery time: 1-2 days
No reviews

Price:
Sale price279,00 €

Shipping calculated at checkout

Configurator

Total price: Sale price279,00 €

Axelera Metis M.2 AI Edge Accelerator – High Performance for Computer Vision

The Axelera Metis M.2 AI Edge Accelerator Card is a compact and highly efficient AI accelerator for demanding computer vision applications . With up to 214 TOPS (INT8) of computing power in the M.2-2280 M-Key form factor , the card enables the parallel processing of numerous image and video streams – with a typical power consumption of only 4–8 W.

The specially developed Metis AI Processing Unit (AIPU) is optimized for high performance and low power consumption. It is ideal for Industry 4.0, robotics, logistics, security systems, and medical technology . Thanks to its industrial-grade temperature range (-40 to +85 °C) and long-term availability , the module is the perfect choice for long-life B2B applications in the edge sector.

Highlights:

  • Up to 214 TOPS INT8 computing power

  • Energy efficient: only 4–8 W power consumption

  • Industrial temperature range (-40 to +85 °C)

  • Compact M.2-2280 M-Key form factor

  • Optimized for Edge AI & Computer Vision

Product variants

The Axelera Metis M.2 AI Edge Accelerator Card is available in two versions:

  • With passive heatsink:
    Suitable for systems without additional airflow. Heat is reliably dissipated via the heatsink, ensuring stable performance even under continuous load.

  • Without heatsink:
    Ideal for embedded or OEM solutions that already incorporate their own cooling strategy (e.g., case fans, heat pipes, or active cooling modules). This variant saves height and enables more flexible integration into compact systems.

Both variants are based on the same Metis AIPU architecture and offer up to 214 TOPS (INT8) performance. The choice depends primarily on thermal requirements and the planned installation environment.

Technical data

feature specification
Model number AXE-BME20M1AB01A01
architecture Quad-Core Metis AIPU
performance Up to 214 TOPS (INT8)
Form factor M.2 2280 M-Key
interface PCIe Gen 3, 4 lanes (4 GB/s bidirectional)
On-board memory 1 GB DRAM
Operating temperature -40 to +85 °C
Power consumption 4–8 W typical
Security features Secure Boot, Root of Trust
compatibility ARM & x86 systems, Linux & Windows

Software support (inferencing)

For software integration, Axelera provides the Voyager SDK , allowing developers to quickly and efficiently deploy deep learning models on the Metis AIPU.

Features:

  • Runs deep neural networks with high energy efficiency

  • Especially optimized for CNNs and computer vision applications

  • Supports parallel processing of multiple models and streams

  • Runs on Linux and Windows

  • Compatible with ARM and x86 systems

Framework & model support:

  • PyTorch and ONNX natively supported

  • TensorFlow via ONNX

  • Automatic quantization (FP32 → INT8) and graph optimization for maximum performance and accuracy

Example models from the model zoo:

  • Image Classification: ResNet-50, MobileNetv2

  • Object Detection: MobileNetv2-SSD, YOLOv5, YOLOv6, YOLOv7, YOLOv8, YOLOv9

  • Segmentation: DeepLabv3, YOLOv8-seg, U-Net

  • Pose Estimation: YOLOv8-pose

Areas of application

  • Industry 4.0 & Automation

  • Security and surveillance systems

  • Robotics & Autonomous Systems

  • Smart Mobility & Logistics

  • Medical image processing

  • Agritech & Smart Farming

Advice from buyzero

We can advise you on whether your model is compatible with the Axelera AI Metis chip and whether you can implement your project using Axelera AI. Please contact us.

FAQ

What are example applications for this module?

This module can be used to accelerate so-called deep learning algorithms, especially CNNs (convolutional neural networks).

This includes:

  • Pose Estimation - Estimating the pose of a person
  • Object recognition with bounding boxes (e.g. where is a bicycle in the image and what is it)
  • Segmentation - mark the areas of an image that a banana occupies
  • License plate recognition
  • Facial recognition
  • and much more

Can I run NVIDIA-based models with this chip? What are the advantages of Axelera vs. NVIDIA?

Specifically, models and software stacks need to be ported (and, for example, quantized from fp32 to int8). This is especially interesting for applications that operate in industrial temperature ranges or have a limited power budget.

We will advise you whether your model is capable of running on the Axelera AI Metis chip and whether you can implement your project using Axelera AI.Please contact us for this.

Further advantages are long-term availability, Axelera's industrial focus, and the availability of more compact solutions (M.2) with higher TOPS compared to, for example, the NVIDIA Jetson platform.

I need more TOPS because I have a complex application. What can I do?

Axelera prepares a PCIe card with four Metis chips -Please contact us, if you would like more information. Another alternative is a cluster of Biscotti modules (based on HAILO-8):

We offer a solution in our shop with52 TOPS- theBiscotti ModuleIt uses two HAILO-8 modules and has an EDSFFE1.SConnector - used in storage servers.

This allows theBiscotti Dual HAILO Modulecan also be installed in clusters, and in a 19'' servere.g. on 1600 TOPSbe scaled.

In cooperation with the manufacturer Unigen, we can also implement Axelera clusters for you with a lead time of approximately 6 months.

We are happy to answer any questions you may have. Please contact us to tell us about your requirements!

Is Axelera AI / the Metis chip a good alternative to Google Coral?

Google Coral has 4 TOPs (8 TOPs in the Dual Edge variant). For applications that have previously deployed many Coral chips in a decentralized manner, Axelera Metis could be an interesting alternative.

For an upgrade with less power requirements, we recommend considering HAILO-8 (13 TOPS - 26 TOPS).

Can I run LLMs on Axelera Metis, e.g., LLAMA?

LLMs = Large Language Models (something like ChatGPT / GPT4-o, or LLAMA, Falcon, etc.)

No, the Axelera Metis chip is not designed for that.

Can I use it to train models?

No, Axelera Metis is intended exclusively for inference (int8). A traditional GPU is typically used to train models.

We are happy to assist you with technical questions about models.

Misc

  • Axelera SKUs / alternative names:

    AXE-BME20M1AB01A01
    M.2 AI Edge accelerator card. Powered by a quad-core Metis AIPU
    Metis Omega M.2 Card
    91-02AXEO-0000
    AD-M28-AM1A (T96AM000***)
  • HS Code: 85176200

Documentation & Downloads

Assembly instructions

  • Installation inM.2-2280 M-Key Slotnecessary
  • Host CPU required
  • Software integration via Voyager SDK

GPSR - EU Responsible Person: Axelera AI BV, High Tech Campus 5, EINDHOVEN, Noord-Brabant, 5656 AE, Netherlands, e-mail: info@axelera.ai
GPSR - Product manufacturer (Contact details for GPSR): Axelera AI BV, High Tech Campus 5, EINDHOVEN, Noord-Brabant, 5656 AE, Netherlands, e-mail: info@axelera.ai
GPSR - Economic operator: pi3g GmbH & Co. KG, Zschochersche Allee 1, 04207 Leipzig, Germany, support [@] pi3g.com

Safety information

  • Please read the instruction manual carefully before using the product.
  • Ensure that all manufacturer's assembly and installation instructions are carefully followed.
  • Use the product only for its intended purpose.
  • Improper use of this product may result in serious injury or property damage.
  • Not suitable for children under 10 years.
  • Improper use may result in injury.
  • This product complies with the applicable safety requirements of the European Union.
  • This product has been tested in accordance with the GPSR, which ensures that all relevant safety requirements for consumer products are met.

Traceability information

Each product has one or more of the following features:

  • A CE marking indicating compliance with the safety, health and environmental protection requirements of the European Union.
  • A unique serial or batch number to ensure traceability and support recalls if necessary.
  • Manufacturer and importer information for customer support and safety inquiries.

Monitoring and reporting of incidents

In the unlikely event of a product issue, we have implemented procedures to:

  • Handle customer complaints promptly.
  • Report serious incidents via the EU Safety Gate/RAPEX system.
  • Cooperate with market surveillance authorities to ensure public safety.

Contact:

  • Email: support [@] pi3g.com
  • Phone: 0341 / 392 858 40

This product is fully compliant with all applicable EU regulations to ensure the safety of our valued customers.

Similar products

Recently viewed