ZED Box 2 Series
Supercomputers for Physical AI and Robotics
Powered by NVIDIA® Jetson Thor™ and Orin™ NX platforms, the NEW ZED Box 2 is a powerful computer designed for Perception.
Powered by NVIDIA® Jetson Thor™ and Orin™ NX platforms, the NEW ZED Box 2 is a powerful computer designed for Perception.
The new ZED Box 2 series are designed for AMRs, industrial automation and autonomous robots. It delivers onboard AI processing, multi-camera synchronization and sensor connectivity in a compact, robot-ready platform.
The new ZED Box 2 series offer a range of embedded computers optimized for vision applications. Designed for robotics, automation, and industrial applications, each model provides a different balance of performance, efficiency, and integration flexibility.
Edge AI computer built for multi-camera robotic vision. Plug in up to 8x ZED X cameras, run the full ZED SDK at up to 157 TOPS, and mount it anywhere, it's only 55 mm tall !
8x GMSL2 camera inputs
Up to 157 TOPS of AI compute.
Full I/O support for LiDAR, GPS, IMU, and other external sensors
Innovative passive cooling. Wide-range power input.
Built-in ESD protection and resistance,CE and FCC certified.
Edge AI compute for the next generation of robots. ZED Box 2 Max is an AI computer built on NVIDIA® Jetson™ Thor for autonomous machines and multi-camera perception at scale.
Jetson Thor inside up to 1200 FP4 TFLOPS
8x GMSL2 camera inputs
Full I/O support for LiDAR, GPS, IMU, and other external sensors
9 –36 V DC wide-range power input.
Built-in ESD protection and resistance,CE and FCC certified.
Edge AI compute for the next generation of robots. ZED Box 2 Max is an industrial AI computer built on NVIDIA® Jetson™ Thor for autonomous machines and multi-camera perception at scale.
Jetson Thor inside up to 2070 FP4 TFLOPS
8x GMSL2 camera inputs
Full I/O support for LiDAR, GPS, IMU, and other external sensors
9 –36 V DC wide-range power input.
Built-in ESD protection and resistance,CE and FCC certified.
ZED Box 2 connects up to 8 GMSL2 ZED cameras to a single unit, giving your robot the coverage it needs. Mix and match ZED stereo and monocular cameras to fit your application: stereo cameras where your robot needs depth, and monocular cameras where you need extra coverage. Whether you need full 360° awareness or targeted coverage of critical zones, you can configure the setup to fit your robot, and adapt it as your needs evolve.
Robotics computing is only as useful as its ability to communicate with the rest of the machine. ZED Box 2 series connect to the systems that make a robot work, including motor controllers, lidars, cameras and network infrastructure. CAN and CAN FD support communication with robot hardware, while ZED Box Max adds 40 GbE, four PoE+ ports and 4G/5G connectivity for demanding, sensor-rich and networked applications.
Compact, Efficient Thermal Design. The ZED Box 2 series feature aluminum enclosure with an integrated heatsink, engineered to dissipate heat efficiently and maintain reliable performance during continuous AI workloads. With no moving parts, the passive cooling system minimizes maintenance, reduces potential points of failure, and ensures silent operation. For more demanding applications, an optional fan kit provides additional cooling capacity when needed.
Every robotic application has different computing requirements. With ZED Box 2 and ZED Box 2 Max, choose from a broader range of NVIDIA® platforms, from Jetson Orin™ NX 8GB and 16GB to Jetson Thor™ T4000 and T5000. Whether you're building a compact perception system or developing more demanding AI-powered robots, find the right balance of computing power and integration needs for your application.
ZED Box combines onboard AI computing with Stereolabs' depth perception technology, enabling robots to process spatial information and run AI workloads directly on the machine.
Robots need more than computing power. They need intelligence that fits within their physical constraints, connects seamlessly to their hardware and performs reliably in real-world conditions. ZED Box 2 and ZED Box Max bring AI computing directly onboard, from real-time perception to Vision-Language-Action (VLA) models.
Robots need more than images to understand their surroundings. They need synchronized, real-time 3D perception across multiple viewpoints. Built around Stereolabs' vision technology, ZED Box 2 and ZED Box Max bring together multi-camera acquisition, hardware synchronization and sensor fusion in a single embedded computing platform.
The ZED SDK is one of the industry’s most advanced, scalable and customizable AI Vision framework, used by 100,00+ users around the world. It provides multiple features allowing you to build any vision-based applications very easily.
The ZED Box 2 is available now and will be dispatched within 4 – 5 business days.
ZED Box 2 Max without GPS will be dispatched end of November 2026.
ZED Box 2 Max with GPS will be dispatched end of December 2026.
Both boxes share the same 8x GMSL2 camera foundation; the choice depends on the workload.
ZED Box 2 — efficient multi-camera perception
ZED Box 2 Max — full edge AI platform
Choose the ZED Box 2 to deploy vision AI efficiently at scale; choose the ZED Box 2 Max when compute headroom and I/O density drive the design.
Up to 8x GMSL2 stereo cameras @HD1200 30fps, or 4x HD1200 60fps
Up to 8x GMSL2 monocular cameras @HD1200 60fps
UP to 6x ZED X One 4K @4K 15fps
Yes. Depending on the application, you will be able to choose between two models based on the compute load, the response time you need and the accuracy the task demands. The ZED Box 2 is well suited to multi-camera perception and navigation on mobile robots, where efficiency and a compact footprint matter. The ZED Box 2 Max is the choice for humanoids and advanced platforms running larger models, sensor fusion and control loops where response time is critical. Both integrate natively into humanoid and mobile robotics workflows, with 8x GMSL2 cameras, CAN and Ethernet out of the box
Yes. All ZED Box models ship with NVIDIA JetPack and the ZED SDK preinstalled, allowing it to be used directly upon reception.
Yes. The ZED Box runs any model compatible with the NVIDIA Jetson platform (TensorRT, CUDA, PyTorch), and the ZED SDK lets you feed your own custom 2D object detector (YOLO/Ultralytics models, in ONNX, TensorRT or OpenCV DNN format) into it, which then computes 3D localization, bounding boxes and tracking on top of your detections.
CE, REACH, RoHS, EMC, FCC, IEC 62368-1, Certificate of Origin
Start building exciting new applications that recognize and understand your environment.
Discover our range of compact, powerful computing solutions designed to bring advanced perception and AI to robotics.
