DEEPX

What DEEPX does

DEEPX is a global leader in on-device AI, specializing in the development of AI semiconductors that bring intelligence to every electronic device. Their cutting-edge AI chip product line aims to revolutionize industries like robotics, smart cities, surveillance, and more by optimizing performance, reducing power consumption, and enhancing cost efficiency for on-device AI operations. DEEPX focuses on making AI accessible and sustainable by embedding intelligence directly into devices, enabling real-time decision-making at the source of data generation. They are building a strong IP portfolio with over 400 patents filed and more than 130 granted worldwide, aiming to lead the AI edge market.

Products

• DX-M1An AI chip designed for edge devices, delivering 25 TOPS (INT8) performance at 1-5W power consumption. It supports multiple AI models simultaneously and is optimized for breakthrough efficiency, offering superior thermal management and GPU-level AI accuracy with INT8 precision.
• DX-M2An AI semiconductor in development, extending DEEPX's vision for accessible, sustainable AI, bringing Generative AI to portable, battery-powered devices. It aims to advance efficient LLMs for the edge with lower power, heat, and cost.
• DX-M1MAn AI processor delivering 25 TOPS (INT8) at 3W (Typical) power consumption, prioritizing Inferences Per Second (IPS) per Watt. It integrates effortlessly with host systems, accelerating computer vision and audio analysis, and features a dedicated NPU and robust hardware-based security.
• DX-H1 QuattroA low-profile PCIe card delivering 100 TOPS (INT8) performance at a 20W TDP. Its 'Quattro' design integrates four AI processors for parallel processing, optimized for AI workloads in data center and edge environments, and designed for easy installation in existing servers.
• DX-H1 V-NPUA low-profile PCIe card with 50 TOPS (INT8) AI performance and integrated decoding, encoding, and transcoding capabilities, operating at 30W. It's designed for real-time, large-scale monitoring, offering an all-in-one architecture with dual hardware video codecs and dual NPUs.
• AI ChipsCutting-edge AI chips engineered for maximum efficiency and effortless integration into any device, including the DX-M series.
• AI ModulesHigh-performance AI modules with exceptional inference, low power, and minimal space needs, compact modules that integrate DEEPX chips for rapid AI deployment.
• AI SystemsComprehensive AI computing solutions designed for peak performance, reliability, and efficiency.
• Single Board ComputersAll-in-one boards with built-in AI acceleration for development and prototyping.
• SoM BoardsScalable, plug-and-play compute modules optimized for on-device AI applications.
• AI GatewayEdge gateways enabling multi-channel, real-time AI inference without cloud dependency.
• Industrial PCsAlways-on AI computing: DEEPX-powered industrial PCs for 24/7/365 reliability in demanding environments.
• DX-AIPlayer N97An AI product validated and deployed, showcasing solutions that combine technical reliability with commercial success in real-world deployment.
• DX-All-SuiteA software suite that includes DX-Compiler and DX-RT, offering features like pip-based Python integration, full YOLO (v3–v26) acceleration, and optimized performance for faster, more stable inference. It focuses on Runtime Efficiency, Security Hardening, and a Unified Application Architecture.

Key capabilities

• Ultra-low power consumption (1-5W for DX-M1, 3W for DX-M1M, 20W for DX-H1 Quattro, 30W for DX-H1 V-NPU)
• High AI performance (25 TOPS for DX-M1/M1M, 100 TOPS for DX-H1 Quattro, 50 TOPS for DX-H1 V-NPU)
• GPU-level AI accuracy with INT8 quantization
• Industry-leading Total Cost of Ownership (TCO)
• Superior thermal management
• Innovative memory design allowing multiple AI models to work side by side
• Dedicated NPU (Neural Processing Unit) architecture
• Robust, hardware-based security (Root of Trust, True Random Number Generator)
• Integrated decoding, encoding, and transcoding (DX-H1 V-NPU)
• PCIe Gen3 x4/x16 host interface support
• Broad OS support (Windows 11/10, Debian-based Linux, Yocto Project, Docker)
• Compatibility with popular AI frameworks (Ultralytics, TensorFlow, PyTorch, ONNX, Keras)
• Support for x86 and ARM-based architectures
• Compact form factors (M.2 modules, PCIe cards, SoMs)
• Product Longevity Policy (minimum 10 years product life cycle)
• Global standard quality management system (ISO 9001, FCC, ISED, UKCA, VCCI, KS certifications)

Use cases

• IoT applications
• Remote monitoring
• Drones (autonomous navigation, aerial surveillance)
• Autonomous Mobile Robots (AMRs) (navigation, object avoidance, logistics, delivery, inspection, warehouse automation)
• Smart Factories (automation, predictive maintenance, quality control, crane automation, automated packet sorting, safety monitoring)
• Smart Cities (intelligent traffic systems, energy-efficient buildings, traffic management, environmental monitoring, public safety)
• Video Management Systems (VMS) (anomaly detection, facial recognition, traffic pattern analysis, advanced video analytics, multi-channel video analytics)
• AI CCTV (real-time object detection, behavioral analysis)
• ADAS (real-time recognition of traffic signs, pedestrians, lanes)
• Industrial Robot Arms (real-time visual inspection, precise assembly)
• AI Dashcams (advanced incident detection, video analysis)
• Edge Servers (managing and processing data from multiple sensors and cameras locally)
• AIoT Devices (smart home and retail devices with on-device data analysis)
• Remote Patient Monitoring (RPM) (real-time, on-device analysis of patient vitals)
• Data Center AI Servers (cloud-based AI services, large-scale AI model inference)
• Unmanned Systems
• Warehouse Management
• Worker Safety Monitoring

AI approach

DEEPX specializes in developing AI semiconductors (NPUs) for on-device AI, focusing on ultra-low power, high-performance, and cost-effective solutions. They aim to bring AI intelligence to every electronic device, from edge devices to industrial PCs and servers, by optimizing performance, reducing power consumption, and enhancing cost efficiency for on-device AI operations. Their products support various AI frameworks and operating systems, and they emphasize their proprietary NPU architecture and intelligent quantization techniques to achieve GPU-level accuracy with lower power consumption.

Tech named: Deep Learning, Processor, Machine Learning, Edge Devices, Embedded Devices, Artificial Intelligence, Software Stack, Edge Computing, Smart Mobility, Leading Edge Technology, AI semiconductors, AI chip product line, on-device AI, DX-M1, DX-M2, GPGPUs, DEEPX IQ8: Intelligent Quantization with INT8, NPU (Neural Processing Unit), CNN, Transformer-based models, Generative AI Semiconductor, Large Language Models (LLMs), DXNN, Ultralytics, TensorFlow, PyTorch, ONNX, Keras, YOLO (v3–v26), Root of Trust, True Random Number Generator (TRNG), hardware video codecs

Industries served

• Robotics
• Smart Cities
• Surveillance
• IoT
• Transportation
• Manufacturing
• Automotive
• Healthcare
• Retail
• Data Centers

What it says sets it apart

• Unmatched performance/power/cost efficiency with maximum accuracy
• Lower TCO than even free chips due to reduced power consumption
• Ability to run SLMs fully on-device
• Sustainable AI for a greener future, reducing energy use and emissions by 90%
• Breakthrough performance at 10% of GPU costs
• Intelligence liberated to the edge, not just in the cloud
• Innovative memory design allowing multiple AI models to work side by side
• Superior thermal management for industrial temperature ranges
• Achieves GPU-level AI accuracy using quantized Int-8 precision
• Exceptional performance efficiency, outperforming GPGPUs by 20x and competitors by more than 2x
• Seamless developer experience with comprehensive software environment for model optimization
• Robust, hardware-based security features
• All-in-one architecture for video processing pipelines (DX-H1 V-NPU)
• Effortless scalability with compact, fanless designs for high-density deployments
• Strong IP portfolio with over 400 patents filed and 130+ granted
• Commitment to product longevity (10-year minimum life cycle)
• Global standard quality management system and certifications (ISO 9001, FCC, ISED, UKCA, VCCI, KS)

Funding rounds we track

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