South Korean AI-chip startup DeepX is expanding its push into physical AI through technical cooperation with Intel and preliminary partnership discussions with AMD, as it seeks a foothold in systems such as robots, autonomous vehicles, drones and smart factories.
The companies are aiming for a plug-and-play connection between DeepX chips and Intel systems, with an initial result expected early next year.
The AMD discussions followed Kim’s meeting with CEO Lisa Su during her October visit to South Korea. DeepX was among the Korean AI-chip companies invited to a roundtable with Su, and Kim said an AMD technical representative subsequently emailed about follow-up discussions. DeepX is preparing for working-level talks, but no finalized AMD partnership has been announced.
DeepX was founded in 2018 and has developed four AI-chip types. Its DX-M1 flagship is in full-scale production on Samsung Electronics’ 5-nanometer process. The company recorded 77 purchase orders from more than 10 countries between August 2025 and July 2026, has contracts with 27 semiconductor distributors worldwide, and said roughly 300 organizations are evaluating its products.
The company is also pursuing a Samsung-based 2-nanometer chip, with tape-out planned for October and the first chips expected around March or April. Separately, DeepX plans to introduce DX-Newton later this year, software intended to manage robot operations and distribute updated AI models across large fleets; its inputs and outputs were designed to align with Nvidia’s Isaac robotics platform.
DeepX’s broader commercialization effort includes a three-year agreement with Taiwan’s AAEON to integrate its NPUs into industrial computers, embedded boards and edge gateways. The company has also worked with Hyundai Motor Group’s Robotics LAB on low-power on-device computing for robots, including an edge-AI chip designed to operate without cloud or network connectivity.