Axelera AI

What Axelera AI does

Axelera AI delivers an AI-native game-changing hardware and software platform to accelerate artificial intelligence. They provide datacenter-level AI inference wherever it's needed, combining purpose-built hardware and software for maximum efficiency and accuracy. Their platform is built using proprietary digital in-memory computing (D-IMC) technology and RISC-V dataflow architecture, delivering industry-leading performance and usability at a fraction of the cost and energy consumption of current solutions. Axelera AI aims to lead the democratization of Artificial Intelligence, offering faster, easy-to-use AI acceleration while minimizing power and cost, to simplify AI at the Edge.

Products

• Metis® AIPUsAI Processing Units starting at 214 TOPs, designed for high performance and power efficiency at the edge. Available as M.2 Max Accelerator Card, PCIe 1-AIPU Accelerator Card, and PCIe 4-AIPUs Accelerator Card. It's a 214 TOP powerhouse packed into a 4-8watt budget, bringing inference to the edge in a compact M.2 module or PCIe card, delivering hundreds to thousands frames per second of real time analysis.
• Europa®AI Processing Unit at 629 TOPs, incorporating on-die pre/post processing, doubling the AIPU core-count, and adding 16 vector processing units. It scales from "heavy edge" to enterprise data center.
• Voyager® SDKSoftware Development Kit with 100+ models, providing ease of use for building and deploying high-performance computer vision applications. It simplifies AI vision system deployment with YAML-based pipeline configuration that works with pre-trained models. It allows deployment of AI applications across both Metis and Europa with the same SDK, enabling development once and scaling across the entire Axelera platform.
• Axelera Metis Compute BoardA complete AI system combining the Metis AIPU with the Voyager® SDK. It is a single-board computer (SBC) based on a Metis AIPU with a four-lane PCIe connection to the CPU, suitable for running multi-stream and multi-model applications. It features an ARM-based application processor and a complete development stack.
• Axelera AI Mini PC with M.2 MaxA complete AI system combining the Metis AIPU with the Voyager® SDK, available in multiple configurations. It is supported by ASRock NUC Box-125H, providing inference at the Edge powered by Metis® AIPU with state-of-the-art performance and energy-efficiency.
• Metis PCIe System with Dell Pro Slim Plus XE5A complete AI system combining the Metis AIPU with the Voyager® SDK, available in multiple configurations.
• Titania®A chiplet that scales the Digital In-Memory compute architecture into high performance compute and AI Gigafactories.

Key capabilities

• AI inference brings intelligence to video monitoring, LLMs, and VLMs out of the cloud to the edge.
• Metis® AIPUs starting at 214 TOPs, Europa® at 629 TOPs.
• Voyager® SDK with 100+ models.
• High performance, power efficiency, and ease of use.
• Proprietary Digital In-Memory Computing (D-IMC) technology.
• Outstanding Energy Efficiency at 15 TOPs/W.
• 99.9% Relative Accuracy with post-training quantization.
• RISC-V controlled dataflow technology.
• Standard CMOS processing for accelerator technology.
• Neural Network optimization with quantization techniques and mapping tools.
• Local compute for Edge/IoT devices.
• ARM-based application processor for general-purpose processing.
• Scalable platform with continuously evolving model zoo.
• Secure Boot and Root of Trust security features.

Use cases

• AI inference for video monitoring
• Bringing LLMs and VLMs out of the cloud to the edge
• Accelerating AI in Computer Vision
• Real-time processing at the edge
• Enhanced privacy and security by minimizing data movement
• Increased scalability in remote or bandwidth-constrained environments
• Accelerating AI in Industrial Manufacturing
• Enhancing productivity in production lines and warehouses
• Machine vision and Industry 4.0 equipment
• AI-based video analytics in Security Systems
• Quick incident investigation and risk prevention in security
• Intelligence for Retail
• Real-time analytics across customer touchpoints in retail
• Accelerating AI in Robotics
• Autonomous mobile robots navigation
• Mountain search and rescue with drones
• Path planning, object detection, and target tracking for drones
• Object detection and tracking across optical and thermal cameras for drones
• Smart Cities
• Medical & Healthcare
• Agriculture
• High Performance Compute
• Industrial Automation
• Visual inspection systems in manufacturing
• Real-time sorting of garbage or vegetables
• Faster and more flexible Optical Character Recognition (OCR)
• Advanced defect analysis
• Precise pick-and-place operations
• Sophisticated 3D vision in robotic applications
• Smart Warehouses powered by AI
• Monitoring occupancy storage and traffic areas to optimize picking and packing
• AI-enabled products for machine builders
• Fraud prevention at checkout in retail
• Real-time inventory monitoring in retail
• Personalized digital signage in retail
• AI Powered Customer Experience in retail
• AI-enabled Point of Sales
• Optimizing shopping with new insights (shopper traffic, queues)
• Streamlining curbside pickup processes
• Enabling large scale autonomous stores
• Real-time threat detection and response in security
• Workplace safety monitoring
• Optimizing traffic flow, decreasing congestion, enhancing safety measures in Intelligent Transportation Systems (ITS)
• AI-Powered PPE detection

AI approach

Axelera AI develops a hardware and software platform for accelerating AI inference, particularly for computer vision on edge devices. Their proprietary Digital In-Memory Computing (D-IMC) technology and RISC-V controlled dataflow technology are central to their approach, aiming for high performance, power efficiency, and accuracy. They provide AI accelerators (AIPUs) and complete AI systems, along with an SDK (Voyager® SDK) that includes a model zoo and tools for optimizing neural networks.

Tech named: Digital In-Memory Computing (D-IMC), RISC-V controlled dataflow technology, SRAM (Static Random-Access Memory), post-training quantization technique, Voyager® SDK, Metis® AIPU, Europa® AIPU, Titania® chiplet, Tensorflow lite, YOLOv8s, ResNet-50, object detection, target tracking, path planning, Optical Character Recognition (OCR), pose estimation

Industries served

• Semiconductor Manufacturing
• Industrial Manufacturing
• Security
• Retail
• Robotics
• Computer Vision
• Smart Cities
• Space
• Medical & Healthcare
• Agriculture
• High Performance Compute
• Industrial Automation

What it says sets it apart

• AI-native game-changing hardware and software platform.
• Datacenter-level AI inference at the edge.
• Proprietary Digital In-Memory Computing (D-IMC) technology.
• Outstanding Energy Efficiency (15 TOPs/W).
• 99.9% Relative Accuracy with post-training quantization.
• RISC-V controlled dataflow technology.
• Cost-effective compared to GPUs.
• Simplified AI at the Edge through faster, easy-to-use AI acceleration.
• Strong investor backing including European Innovation Council Fund, Innovation Industries, Invest-NL, Samsung Catalyst Fund, Verve Ventures, Fractionelera, CDP Venture Capital SGR, RVO, Bitfury, SFPIM, and imec.
• Largest investment ever in an EU AI semiconductor company (over $450 million total equity, grants, and venture debt).
• Taped out four chips and launched Metis AI Platform, Europa platform, and Titania chiplet.
• Voyager SDK for effortless machine learning model deployment.
• Unbeatable combination of high performance with low power consumption for industrial edge applications.
• Industrial-grade reliability, low power consumption for extended temperature range, and fanless design.
• Low latency and high throughput for industrial AI hardware.
• Ability to stream 24 concurrent 1080p video feeds for comprehensive store coverage.
• Consistent accuracy for real-world retail environments due to in-memory computing.
• Purpose-built hardware combining cloud power with edge device power budget, working with standard camera infrastructures.

Funding rounds we track

From the AI funding tracker — rounds as reported by the linked publications.