IN Brief:
- NEURA and SECO will develop electronic compute modules for cognitive robots including the 4NE1 humanoid.
- Qualcomm Dragonwing processors will sit within NEURA's distributed Brain + Nervous System architecture.
- SECO will engineer and industrialise the hardware for European series production, alongside semiconductor and electronics manufacturing applications.
NEURA Robotics and SECO will jointly design, engineer, and manufacture electronic compute modules for NEURA’s cognitive robots, including its 4NE1 humanoid platform. The modules will use Qualcomm Dragonwing processors and form part of NEURA’s distributed computing architecture, with development and series production planned in Europe.
The hardware programme is built around NEURA’s Brain + Nervous System architecture, which distributes sensing and compute through the robot rather than concentrating every function in a single central processor. Its Smart Limb concept places processing closer to joints, sensors, and other points where low-latency decisions are required. SECO will take responsibility for engineering and industrialising the corresponding compute modules, bringing embedded design and manufacturing capability into a platform intended for repeatable production.
Distributed robot compute changes the electronics problem substantially. A central high-performance processor remains useful for perception, planning, and large AI workloads, but motor control, sensor processing, safety functions, and local decision loops often have tighter latency and determinism requirements. Pushing selected processing closer to those functions can reduce communication delays and bandwidth demand, although it also increases the number of electronic modules that must be qualified, powered, cooled, networked, and maintained across the machine.
Qualcomm’s Dragonwing robotics processors are being used as the compute foundation for that architecture, while SECO already develops industrial platforms around the wider Dragonwing family. The partnership therefore combines a robotics system architecture with an embedded supplier experienced in turning processor platforms into manufacturable boards and modules. Stable hardware configuration, component lifecycle management, thermal design, and production test become increasingly important once a robot moves beyond limited development builds.
The companies also plan to use deployments in semiconductor and electronics manufacturing to collect production data and develop automation capabilities around NEURA’s Physical AI platform. A NEURA Gym in Italy is planned as the company’s first training hub in southern Europe, providing another environment for developing and validating robot skills. The stated aim is to reuse knowledge generated in individual industrial processes rather than treating every automation deployment as an isolated engineering project.
Electronics manufacturing provides a demanding environment for that approach because production lines combine repetitive handling with fine tolerances, variable product flows, machine interfaces, inspection tasks, and strict uptime requirements. Distributed compute modules used in that setting have to support more than raw AI throughput: deterministic interfaces, thermal control, electromagnetic compatibility, serviceability, and long-term component availability all become part of the design. Series manufacture in Europe also puts configuration control and supply continuity under greater scrutiny than a development platform assembled in small numbers.
European production links the embedded hardware design directly to the robot platform’s manufacturing plan. The partners expect production deployments to generate data that can be fed back into later automation systems, while the compute modules themselves will have to accommodate controlled revisions to processors, memory, networking, and sensor interfaces. Hardware changes therefore become part of a managed product platform rather than ad hoc upgrades to a prototype.
Module-level production also creates a clearer qualification boundary than a collection of development boards. Each compute module can be tested against defined electrical, thermal, and interface requirements before installation in the robot, while firmware and hardware revisions can be tracked against a controlled configuration. That discipline becomes increasingly important as the same architecture is reused across multiple robot variants and industrial applications.
The partnership has not yet disclosed module specifications, processor variants, memory configurations, interfaces, environmental ratings, or a timetable for the first production hardware. Those details will determine how much processing is distributed into limbs and subsystems, how much remains centralised, and what the power and cooling budget looks like across the machine. For now, the engineering programme establishes a clear division of responsibility: NEURA defines the distributed robot architecture, while SECO is tasked with turning that architecture into compute modules that can be manufactured repeatedly in Europe.