Intel launches its own robotics platform

Intel is positioning itself as the preferred infrastructure for developers who want to build robots and deploy them in the field, rather than just selling them chips. Under the new brand Intel Robotics, unveiled at the CompEx 2026 conference in Taiwan and expanded thereafter, the company bundles hardware, development cards and a software package intended to shorten the gap between a laboratory prototype and an industrial robot on a production line. The effort is currently carried out within the company’s personal computing division, which has been renamed Client Computing and Physical AI, but the scale of the move points to an attempt to create a separate, focused business unit.

The hardware portfolio relies on the Core Ultra Series 3 family of processors, including a version that has been specially adapted for robots and edge AI with GPU and NPU accelerators. Development cards accompany the processors to let customers build new platforms. The software component is central to the offering: everything is packaged in the Intel Robotics AI Suite, which contains two main modules. The first, OpenVINO Physical AI, is an open-source library that provides an inference runtime optimized at the silicon level. The second, Physical AI Studio, handles data collection, fine-tuning and export of Vision-Language-Action (VLA) models.

VLA models form the core of modern robotics: they perceive the environment and execute actions based on that perception, with inference cycles that can run dozens of times per second. OpenVINO Physical AI is intended to supply a unified interface for moving multimodal models and robot policies from the development environment to the physical robot while maximizing inference performance. Dan Rodriguez, senior vice president of Edge Computing at Intel, explains that the deployment of Physical AI models has been delayed until now because of fragmented software layers and robot-specific integrations that were required for each robot individually.

Physical AI Studio is already available for use, but OpenVINO Physical AI is currently only in a Preview release on GitHub. General availability is expected only in the second half of 2026. Intel argues that the current market approach forces customers to maintain duplicate systems with excess compute resources, and that combining Core Ultra Series 3 with OpenVINO will lower total cost of ownership, enable code reuse across different robot types and streamline deployment in factories, warehouses and retail environments. For now, the claim remains a promise awaiting the code’s arrival on GitHub.