A Guth Labs publication ↗
Back to News

Chips&Media says Mobilint production milestone opens a route into robotics

Conceptual AI semiconductor packages move through a precision production line over a circuit-board landscape.
AI-generated conceptual editorial illustration for guthlabs.newsroom.

guthlabs.newsroom | Physical AI Hardware | November 4, 2026

Chips&Media announced on November 4 that Mobilint’s next-generation AI chips incorporating its video-codec intellectual property were entering mass production. The companies are targeting physical-AI and robotics applications, including humanoids. The production milestone is company-reported; the announcement does not disclose shipment volumes or identify a completed robot deployment using the new chips. Chips&Media announcement

The relationship extends earlier licenses for H.264, H.265 and AV1 video technology, with AV2 and H.266/VVC added more recently. Chips&Media says its video-processing technology can handle 8K video at 60 frames per second on one core and emphasizes multi-camera workloads. Those specifications describe the vendor’s component capability, rather than measured end-to-end performance of an entire robot. Chips&Media announcement

The companies also plan to extend their cooperation into frame-buffer and data-compression IP. That expansion remains a stated plan, separate from the current video-codec production announcement. Chips&Media announcement

The visual pipeline matters before inference begins

A robot’s AI model depends on the information reaching it. Handling multiple camera feeds can place demands on bandwidth, memory and power before the model performs any inference. Dedicated video hardware is one way to manage those demands within a larger chip design.

That makes the announcement commercially different from a new foundation model or a finished humanoid. Chips&Media supplies reusable hardware designs that a semiconductor company integrates into its own product. The resulting chip still needs to fit the robot developer’s sensors, software and operating constraints.

The earlier safety work provides relevant context. In September, Chips&Media announced that its WAVE6331X FuSa video-codec IP had received ISO 26262 ASIL-B certification from DNV. It described fault-detection mechanisms, including protections intended to help detect corrupted data or abnormal operation, and a safety manual supporting customers’ integration work. Chips&Media certification announcement

That announcement characterizes the certified component as a safety element out of context. Its protections operate under defined assumptions of use and support chip-level safety work. Component certification should therefore not be treated as certification of a complete robot, its AI behavior or every use of the licensed video standards. Chips&Media certification announcement

From an IP license to a working machine

For chip buyers, the practical evaluation goes beyond a headline resolution or frame rate. Multiple simultaneous streams, memory traffic, latency and power consumption need to be tested together under the intended workload. A component’s maximum specification does not establish that every configuration can achieve it at once.

Robot developers also need a clear response when visual information is missing, delayed or unreliable. Detecting a fault is useful only when the rest of the system knows how to respond safely. That coordination crosses the boundary between the video block, the AI processor and the machine’s control software.

The Mobilint milestone gives Chips&Media a production-linked entry into a growing application area. Its significance will become clearer through disclosed design wins and deployed systems. For now, it is evidence of progress in the hardware supply chain, with broader multimedia integration and real-world robot performance still separate questions.

Sources and citations

Revision history

Published .