Intel Corp. announced 130-plus design engagements for its Series 3 processor family, intended for edge artificial intelligence and edge computing. The Series 3 lineup, consisting of Core Ultra Series 3 and Core Series 3 processors, debuted in January at CES and utilizes Intel’s 18A manufacturing process. This launch aims to consolidate fragmented robotics hardware, replacing mixed central processing units and discrete accelerators with a unified platform.

SensoryAI Inc. has migrated its multi-agent retail robot, Ella, to the Intel Core Ultra Series 3 platform. The system now manages real-time robot control and AI inference on a single chip, running three specialized agents concurrently: an Avatar agent for customer interaction, a Guardian agent for operations, and an Ella agent for business intelligence. Other design applications for the Series 3 include industrial generative AI, vision-based defect detection, humanoid robots, and multimodal AI for medical imaging.

Intel also unveiled OpenVINO Physical AI, an extension of its 2018 edge computer vision toolkit. This open-source robotics library provides a silicon-optimized inference runtime, allowing developers to move robot policies and multimodal models from experimentation to production. It integrates with existing environments such as Physical AI Studio and the open-source LeRobot project. Intel intends to address the deployment gap by providing a consistent, scalable software stack that reduces reliance on complex, one-off integration pipelines.

In performance comparisons shared by Intel, the Core Ultra X7 358H is positioned against Nvidia Corp.’s Jetson AGX Orin and Jetson Thor T5000 modules. Intel claims competitive performance with the Thor module on medium-sized vision-language-action models while estimating the total system cost at roughly half of Thor’s price. Additionally, Intel reports lower latency than the AGX Orin on a three-camera Pi0.5 Droid workload. OpenVINO Physical AI is currently available in preview on GitHub, with general availability scheduled for the second half of 2026.

Intel Corp. announced 130-plus design engagements for its Series 3 processor family, intended for edge artificial intelligence and edge computing. The Series 3 lineup, consisting of Core Ultra Series 3 and Core Series 3 processors, debuted in January at CES and utilizes Intel’s 18A manufacturing process. This launch aims to consolidate fragmented robotics hardware, replacing mixed central processing units and discrete accelerators with a unified platform.

SensoryAI Inc. has migrated its multi-agent retail robot, Ella, to the Intel Core Ultra Series 3 platform. The system now manages real-time robot control and AI inference on a single chip, running three specialized agents concurrently: an Avatar agent for customer interaction, a Guardian agent for operations, and an Ella agent for business intelligence. Other design applications for the Series 3 include industrial generative AI, vision-based defect detection, humanoid robots, and multimodal AI for medical imaging.

Intel also unveiled OpenVINO Physical AI, an extension of its 2018 edge computer vision toolkit. This open-source robotics library provides a silicon-optimized inference runtime, allowing developers to move robot policies and multimodal models from experimentation to production. It integrates with existing environments such as Physical AI Studio and the open-source LeRobot project. Intel intends to address the deployment gap by providing a consistent, scalable software stack that reduces reliance on complex, one-off integration pipelines.

In performance comparisons shared by Intel, the Core Ultra X7 358H is positioned against Nvidia Corp.’s Jetson AGX Orin and Jetson Thor T5000 modules. Intel claims competitive performance with the Thor module on medium-sized vision-language-action models while estimating the total system cost at roughly half of Thor’s price. Additionally, Intel reports lower latency than the AGX Orin on a three-camera Pi0.5 Droid workload. OpenVINO Physical AI is currently available in preview on GitHub, with general availability scheduled for the second half of 2026.