SEMIFIVE will develop a robotics-focused AI chip together with Mobilint, according to an announcement reported by EE Times Asia. The collaboration pairs a custom silicon design specialist with an AI accelerator company aiming squarely at robotics workloads.

For practitioners, the deal signals where robot-brain silicon is heading: away from repurposed data-center accelerators and toward designs that match the constraints of embedded robotics — tight power budgets, real-time inference, and edge deployment.

Who is doing what?

SEMIFIVE handles the chip-design side. The company has built its business on custom SoC design and chiplet-based platforms, which lets partners assemble domain-specific silicon without funding a full from-scratch design effort. That model shortens the path from architecture to tape-out for firms that don't keep large internal silicon teams.

Mobilint brings the AI-inference muscle. The startup develops high-performance NPU accelerators and already fields products targeting computer vision and machine-learning inference at the edge — the workload class robotics depends on.

What does the partnership change?

The joint effort focuses on an AI chip tailored for robotics applications. That's a meaningful shift for robot OEMs and integrators: a design conceived for robotics from the start, rather than an inference chip adapted after the fact.

Custom design work also gives the partners room to tune the usual trade-offs that dominate embedded AI — performance per watt, latency guarantees for control loops, and memory bandwidth — against the specific demands of robotic perception and decision-making.

Why does this matter for the market?

Robotics has become one of the fastest-growing demand pools for edge AI silicon, and suppliers are racing to offer platforms rather than generic chips. A design-services firm and an accelerator developer combining forces is a familiar playbook in this segment — it mirrors how much of the current edge-AI silicon wave gets built, through partnerships rather than single-company vertical integration.

For engineers specifying compute for next-generation robotic systems, the announcement adds another candidate to the qualified-silicon shortlist. For teams already working with either company, it opens a route to robotics-tuned accelerators without changing design partners.

The companies have not yet disclosed the chip's target specifications, timeline, or expected availability. Further technical details are expected as development progresses.

Die Signal will track this story as the partners release specification and schedule information.