NXP Launches Unified Robotics Platform for Physical AI, Starting With Drones
NXP Semiconductors launched a unified robotics platform for physical AI on 9 November 2026, combining a modular open-source software development kit with ready-to-use reference designs, starting with drones that the company says can reach initial functionality within 30 minutes out of the box. The platform and the robotics SDK are available now, while enablement for service robots and autonomous mobile robots is expected only in the second half of 2027.
Physical AI is AI that acts in the real world through machines, rather than only producing text or images on a screen. A physical AI system must sense its surroundings, decide what to do, and move safely, all in real time. That is harder than chatbot AI, because a mistake can crash a drone, damage property or hurt a person, so the hardware and software must be reliable, secure and predictable.
NXP built the platform on its Neural Axis Architecture, a distributed foundation that securely connects sensing, real-time control, AI processing and communications across robotic systems. The architecture is inspired by nature, splitting robotic autonomy into reasoning, coordination and reflex layers, so that fast reactions happen locally while higher-level planning happens at a slower pace. The platform brings together NXP secure, safety-enabled technologies for AI, connectivity, motor control, security and energy management in one development path.
The promise is speed for developers. Using a single validated development experience, developers can scale from single functions to complete trusted physical AI systems, according to NXP. Instead of choosing a processor, finding drivers, wiring together middleware and building a board from scratch, a developer starts from a reference design that already combines those pieces, and adds their own application on top.
 |
| Illustration: A drone flying over sunlit green fields with wildflowers and trees in the background, illustrating the landscape where NXP's drone platform operates |
What the drone platform includes
The first reference designs under the unified platform support drones, and NXP divides the main hardware functions across four boards. Autonomy compute and flight control run on the NavQ95, based on the i.MX 95 applications processor. Flight management runs on the Tropic VMU and FMU, based on the i.MX RT1064 crossover microcontroller. Intelligent sensor networking runs on the 100Base-T1 Sensor Node, based on the MCX N947 microcontroller. Motor control runs on an AM32-compliant electronic speed controller, based on the MCX A13 microcontroller.
That division is the point of the design. A drone needs a powerful processor for AI and autonomy, a fast microcontroller for flight control, a separate path for sensors, and precise motor control, and each of those jobs has different timing and safety needs. By splitting them across boards that are designed to work together, NXP lets a developer replace one part without rebuilding the whole drone, and lets different teams work on different functions at the same time.
Supported by the robotics SDK, these reference designs allow developers to deliver an initial drone functionality within 30 minutes out of the box, according to NXP. That is a vendor claim about a first, basic function, not a claim that a finished commercial drone can be built in half an hour. The 30-minute figure is useful as a signal of how much setup work NXP has removed, but a real product still needs testing, safety checks, regulatory approval and a great deal of application software.
There is an important boundary that buyers should notice. The Modular Drone Platform and its four associated boards carry preproduction labels, and NXP states that the drone platform specifications and other information may change without notice. The platform, drone reference designs and SDK are available now, but the hardware is not yet final. Developers who build on it should expect changes and should not treat a preproduction board as a finished product.
 |
| Illustration: Drones, motors, circuit boards, screws and tools on a workbench, illustrating drone hardware assembly for NXP's robotics platform |
The open-source software stack
The robotics SDK is modular and open-source, and it provides supported drivers, middleware integration, reference applications, configuration tools and reusable software components. It is designed to work with industry-standard robotics frameworks, and it includes integrations with established open-source technologies including PX4, Zephyr and ROS 2. ROS 2 Jazzy, PX4 autopilot, CogniPilot Cerebri, Zephyr RTOS, NuttX and AM32 electronic-speed-controller firmware are listed in the software stack, with eIQ Toolkit and Ara SDK for edge-AI work.
ROS 2 is an open-source framework for writing robot software, widely used in research and industry. PX4 is an open-source autopilot system for drones and other vehicles. Zephyr is a small real-time operating system for microcontrollers, designed for devices that must respond quickly and reliably. By supporting all three, NXP is betting that developers want to use the tools they already know, rather than learning a closed NXP-only system.
The documented build environment uses Linux with Docker, the west meta-tool, the Arm GNU toolchain, CMake and Ninja, while prebuilt artifacts require a Nexus user token. That detail matters for teams that want to try the platform, because it shows the setup is aimed at professional embedded developers, not at hobbyists. A team that already builds Linux-based robotics software will find the path familiar, while a team new to embedded work will face a learning curve.
Security and safety are central to how NXP presents the platform. The reference designs incorporate secure processing, deterministic real-time control and safety-ready system architectures built from NXP broad hardware portfolio. Deterministic real-time control means the system responds within a known time, every time, which is essential for a drone that must correct its flight path hundreds of times a second. Safety-ready means the design is intended to support safety certification, not that it is already certified for every use.
 |
| Illustration: A person in a laboratory looking at a drone on a workbench, with a 3D printer, tools and equipment in the background, illustrating hands-on hardware development |
What comes next, and when
While the first designs support drones, the platform will extend across a multitude of autonomous systems requiring perception, navigation, motor control, connectivity and safe real-time operation. NXP says additional enablement targeting other autonomous systems is expected to be available in the second half of 2027. That includes service robots, autonomous mobile robots and, in the longer term, humanoid robots, but those are future enablement only and should not be presented as available now.
That timeline is the most important fact for anyone planning a product. A team building a drone can start today, with the caveat that the boards are preproduction. A team building a warehouse robot or a service robot must wait until the second half of 2027 for NXP reference designs for that class of machine, although the SDK and the Neural Axis Architecture are meant to carry over. The platform is therefore a starting point, not a complete robotics catalogue.
For the robotics industry, the launch fits a wider shift. Chip companies are moving from selling single processors to selling whole platforms, with software, reference hardware and safety features bundled together. That shift helps smaller teams build robots without a large hardware department.
For drone makers, the immediate question is whether the 30-minute claim holds up in practice and whether the NavQ95 delivers enough AI performance for autonomy tasks such as obstacle avoidance and visual navigation. Those questions will be answered by developers who try the platform, not by a press release.
| Function | Reference design | NXP processor |
|---|---|---|
| Autonomy compute and flight control | NavQ95 | i.MX 95 applications processor |
| Flight management | Tropic VMU and FMU | i.MX RT1064 crossover microcontroller |
| Intelligent sensor networking | 100Base-T1 Sensor Node | MCX N947 microcontroller |
| Motor control | AM32-compliant electronic speed controller | MCX A13 microcontroller |
The agentic AI context for this launch is covered in our Claude Duet is here report, the hub of this cluster. The hardware race is covered in our Apple M6 Macs are here article, and the wearable and edge-device side is covered in our Reported Samsung intelligent eyewear report.
Key Takeaways
- NXP launched a unified robotics platform for physical AI on 9 November 2026, built on its Neural Axis Architecture, with a modular open-source SDK and reference designs.
- The first reference designs support drones, with the NavQ95 on the i.MX 95 for autonomy compute and flight control, and NXP claims initial drone functionality within 30 minutes out of the box.
- The SDK integrates ROS 2, PX4 and Zephyr, plus CogniPilot Cerebri, NuttX and AM32 firmware, with eIQ Toolkit and Ara SDK for edge AI.
- The drone boards carry preproduction labels and specifications may change without notice, so the hardware is not yet final.
- Service robots and autonomous mobile robots are future enablement only, expected in the second half of 2027, and should not be presented as available now.
Conclusion
NXP new platform is a practical step for physical AI, not a robot itself. It gives drone developers a validated starting point, with open-source software and a clear path from a single function to a full system. The test will be whether teams can actually reach that first flight in 30 minutes, and whether the platform grows into service robots and mobile robots on the promised 2027 timeline.
Sources
Unite.AI, NXP Launches Modular Robotics Platform for Physical AI Development, 9 November 2026
Comments
Post a Comment