DEEPX
About
About DEEPX
DEEPX designs neural-processing accelerators for running computer-vision and other inference workloads directly in cameras, robots, vehicles and industrial devices. DX-M1 delivers 25 TOPS at one to five watts as a 17-millimetre chip or M.2 module, supporting x86 and Arm hosts, Windows and Linux, and models from PyTorch, TensorFlow, ONNX and related frameworks.
Higher-throughput modules include the 100-TOPS, 20-watt DX-H1 Quattro and the DX-H1 V-NPU, which pairs 50 TOPS with hardware decoding for 64 simultaneous 1080p video channels. DEEPX also packages accelerators in evaluation kits, single-board computers and edge servers. Industrial-temperature and automotive-qualified options position the technology for continuously operating machines that cannot depend on cloud latency or connectivity.
Products
Supply
Supply chain relationships
- Wind River
Outbound - Compute & Control - Wind River integrates DEEPX edge-AI accelerator technology into a mission-critical robotics stack
Source - Hyundai Motor Company
Outbound - Compute & Control - Hyundai Motor Robotics Lab incorporates DEEPX DX-M1 NPUs for perception in its AMR development architecture
Source
Funding
- 2024-05-09 - Series C - $80.5M - DEEPX raises KRW 110B Series C
DEEPX completed a KRW 110 billion Series C, stated as USD 80.5 million, led by SkyLake Equity Partners. - source: prnewswire.com
Investors
- AJU IB - Participant - Angel / Individual - 1 funding event - first seen 2024-05-09
- BNW Investments - Participant - Angel / Individual - 1 funding event - first seen 2024-05-09
- SkyLake Equity Partners - Lead - VC - 1 funding event - first seen 2024-05-09
- TimeFolio Asset Management - Participant - Unknown - 1 funding event - first seen 2024-05-09
Team
- Lokwon Kim - Founder and Chief Executive Officer
Timeline
Events
- 2026-10-06 - Award - Doosan Robotics consortium selected for AI welding research project
The 45-month project targets at least 95% alignment with skilled-worker paths, path error within ±3 mm, automated welding of five or more layers, and more than 50% shorter working time. DEEPX will develop NPU-based edge hardware and a vision pipeline; SAGE will handle 3D recognition and defect models, while Changwon contributes path generation and physics-based digital twins. Planned trials at Doosan Enerbility’s Nuclear Center will cover tack welding, preheating and welding before broader application to shipbuilding, plant equipment and defence.
Source - 2026-08-14 - Market Signal - DEEPX reports 77 DX-M1 purchase orders worth more than $13 million
Order count rose from 29 at the end of March to 56 at the end of May and 77 at the end of July 2026. More than half of cumulative revenue through July came from overseas markets. DX-M1 is fabricated on Samsung's 5-nanometre process for low-power inference in robotics, smart factories, CCTV and other edge systems.
Source - 2026-07-14 - Capability - DEEPX expands NPU integrations across YOLO, PaddlePaddle and Raspberry Pi
The Ultralytics integration lets developers export YOLO models into DEEPX format and run them through the company’s edge-AI runtime. DEEPX and PaddlePaddle have enabled PP-OCR 5 workloads on the DX-M1 accelerator, extending the stack into industrial OCR and real-time inference. Raspberry Pi 5 provides a development and validation layer before applications move into robots, industrial cameras, edge gateways and smart-factory equipment.
Source - 2026-07-03 - Partnership - DEEPX signs Japan distribution partnership with Koshida Tech
The agreement was signed at Koshida Tech?s Tokyo headquarters on July 2. Target markets include manufacturing, robotics, drones, smart factories, industrial PCs, security cameras and edge servers. The partnership builds on DEEPX?s Tokyo Physical AI Expo presence with Koshida.
Source - 2026-05-13 - Partnership - Ultralytics partners with DEEPX to bring YOLO inference to edge NPUs
The integration added a supported format=deepx export path for compiling and running YOLO models on DEEPX NPU hardware. Low-power edge deployment targets robots, industrial automation and autonomous systems that require real-time vision. The workflow covers export, quantisation and validation for DEEPX devices.
Source - 2026-04-15 - Partnership - DEEPX and Hyundai Motor Group expand collaboration on next-generation on-device AI chips for robots
DEEPX and Hyundai Motor Group are expanding their collaboration to develop a generative AI robotics platform built on DEEPX's next-generation low-power DX-M2 chips DEEPX's current chips already power Hyundai Motor Group delivery robots, including the Robotics LAB's DAL-e Delivery platform The expanded collaboration builds on the companies' earlier robotics chip work and extends the partnership toward next-generation mass-produced robot systems
Source - 2026-03-10 - Product Launch - DEEPX unveils DX-M1M and DX-AIPlayer mass-production product lines at Embedded World 2026
DEEPX introduced the mass-produced DX-M1M M.2 module and the DX-AIPlayer integrated edge-AI acceleration platform at Embedded World 2026. The company said the products were designed to simplify deployment of intelligent vision AI in industrial environments including robotics, smart factories, and automated logistics. DEEPX positioned the launches as part of its push to become a physical-AI hardware standard in Europe through immediate commercial deployment paths.
Source - 2026-01-08 - Partnership - DEEPX and Hyundai Motor Group Robotics Lab validate on-device AI for next-generation robots and security systems
At CES 2026, DEEPX said Hyundai Motor Group Robotics Lab had validated its on-device AI capabilities in real-world conditions. A Hyundai Motor Group Robotics Lab executive stated that the two sides planned to deploy the solutions in next-generation robots and security systems starting in 2026. This was a robotics-specific commercialization signal tied directly to DEEPX''s physical-AI chip strategy rather than only a conference appearance.
Source - 2026-01-08 - Product Launch - Hyundai and DEEPX Unveil On-Device AI Chip for Robots
• Hyundai Motor Group's Robotics LAB and DEEPX co-developed an on-device AI chip for robots, unveiled at CES Foundry 2026. • The ultra-low power 'edge brain' chip, consuming less than 5 watts, enables real-time autonomous decision-making without external connectivity. • This technology has been integrated into Hyundai's Facey facial recognition system and DAL-e Delivery robot, with plans for broader application in future AI security solutions and mobile robots.
Source - 2025-12-18 - Partnership - DEEPX signs strategic partnership with Avnet Silica to expand into Europe, the Middle East, and Africa
DEEPX signed a franchise agreement with Avnet Silica to distribute its ultra-low-power NPUs across EMEA. The company said the partnership targeted industrial edge markets including autonomous mobile robots, machine vision, smart factories, and smart-city systems. DEEPX framed the agreement as the start of a full-scale European market expansion built around commercial access to its physical-AI semiconductor platform.
Source - 2025-08-13 - Partnership - DEEPX and Wind River collaborate on mission-critical edge AI for robotics and industrial systems
DEEPX announced a collaboration with Wind River that combines DEEPX on-device AI semiconductors with Wind River''s VxWorks RTOS and Helix virtualization platform. The companies positioned the stack for mission-critical edge deployments across aerospace, defense, industrial systems, and robotics. The announcement marked a concrete software-plus-silicon partnership aimed at speeding deployment of reliable physical-AI systems rather than a generic ecosystem mention.
Source - 2024-05-09 - Funding - DEEPX raises KRW 110B Series C
SkyLake Equity Partners led the completed round and became DEEPX’s second-largest shareholder. BNW Investments and AJU IB participated alongside existing investor TimeFolio Asset Management. DEEPX said the capital would support mass production and next-generation on-device large-language-model chips.
Source
Relations
Relationships
- Raspberry Pi
Physical AI Ecosystem Integration - DEEPX integrates its edge NPU platform with Raspberry Pi for physical-AI development and deployment workflows.
- Hyundai Motor Company
On Device AI Chips / Robotics Platforms - DEEPX and Hyundai Robotics LAB co-develop on-device AI semiconductors for robot platforms.
- Doosan Robotics
Technology Partner - Doosan Robotics and DEEPX are developing AI welding robots using DEEPX's NPU hardware and SDK, with nuclear-equipment trials planned.
- Koshida Tech
Japan Distribution Partner - DEEPX signed a strategic partnership with Koshida Tech to expand distribution of low-power AI semiconductor products in Japan.
- Avnet Silica
Emea Distributor - Avnet Silica signed a franchise distribution agreement for DEEPX NPUs across EMEA.
- Ultralytics
Partnership Counterparty - Ultralytics announced a DEEPX partnership and a supported export integration for deploying YOLO models on low-power NPU hardware.
- Hyundai Motor Group
Robot AI Development Collaborator - DEEPX supplies compute control to Hyundai Motor Group.
Context
Comparable profiles
- Hailo
stack adjacency - Edge AI Compute Supplier
- NVIDIA
stack adjacency - Edge AI Compute Supplier · Robotics Compute Infrastructure · Edge AI Compute
- Huixi Intelligence
stack adjacency - Edge AI Compute Supplier · Robotics Compute Infrastructure · Edge AI Compute
- ZOTAC
stack adjacency - Robotics Compute Infrastructure · Edge AI Compute
- ModalAI
stack adjacency - Edge AI Compute Supplier · Edge AI Compute
- SiMa.ai
stack adjacency - Edge AI Compute Supplier · Edge AI Compute
- One Stop Systems
stack adjacency - Edge AI Compute Supplier · Edge AI Compute
- Acrab
stack adjacency - Edge AI Compute Supplier · Edge AI Compute
- Sugon
stack adjacency - Robotics Compute Infrastructure
- Husarnet
stack adjacency - Robotics Compute Infrastructure
- Husarion
stack adjacency - Robotics Compute Infrastructure
- ThunderSoft Technology
stack adjacency - Edge AI Compute Supplier
Context tags
- Edge AI compute - Capability - NPU and edge AI processors for on-device inference
- Robotics compute infrastructure - Stack Layer - Compute hardware layer for robotics and edge AI systems
- Component supply - Commercial Stage - Commercial position is as a component and chip supplier
- Edge AI and robotics compute layer - Stack Layer - Core peer group: AI edge compute chip supplier
- Factories - Customer Environment - Industrial automation as target compute application