Intel Challenges NVIDIA and Dedicated Accelerators: 'No More Double Hardware in Robots'

At Computex, Intel announced a new phase of its strategy focused on robotics and edge artificial intelligence, revealing that it has made over 130 design commitments based on the Intel Core Ultra Series 3 and Intel Core Series 3 processors for edge AI applications.

The U.S. company particularly aims to strengthen its presence in autonomous robotic systems and intelligent industrial platforms, a sector where the increasing demand for automation is pushing companies and operators to invest in solutions capable of processing data locally, reducing latency, and managing real-time AI inferences.

Among the most relevant innovations is the debut of OpenVINO Physical AI, a new open-source framework optimized for Intel hardware, designed to simplify the transition from experimental prototypes to large-scale commercial deployments. According to Intel, one of the main obstacles to the widespread adoption of AI robotics is the fragmentation of software stacks and the need to create custom integrations for each individual robot.

The goal of OpenVINO Physical AI is to provide a unified environment for managing multimodal models, inference pipelines, sensors, and actuation systems, thereby reducing development complexity and implementation costs. The framework integrates an optimized silicon-level inference runtime and supports open-source environments such as LeRobot and Physical AI Studio. The latter represents an additional piece of Intel’s Robotics AI Suite and offers dedicated tools for data collection, model fine-tuning, quantization, and optimization of neural networks before operational deployment.

Intel’s approach is also concretely demonstrated by the Ella project developed by Sensory AI, described as the first operational multi-agent Physical AI store in a public commercial environment. The company has abandoned a previous architecture composed of separate CPUs and discrete accelerators, moving to a single platform based on the Intel Core Ultra Series 3.

According to reports, the system allows for the simultaneous execution of three specialized AI agents on the same SoC: an Avatar for customer interaction, a Guardian for system operations, and the Ella Agent for retail business intelligence. A deterministic orchestrator coordinates the robot's activities in real time.

Intel claims that this setup allows for the elimination of additional hardware components, reduces software complexity, and improves return on investment, as well as creating a more easily scalable foundation for future robotic systems.

The company also highlights how the market is rapidly shifting from traditional deterministic robotic systems to autonomous Physical AI platforms capable of perceiving the environment, making decisions, and acting safely with latencies on the order of milliseconds. In this context, Intel believes that an integrated hardware-software platform can help companies and integrators reduce TCO and reuse code across different types of robots.

A preview of OpenVINO Physical AI is already available on GitHub, while general availability is expected in the second half of 2026. Physical AI Studio is already accessible to developers.