Edge AI solutions provider, Aetina, launches its next-generation platform – the AIB-AT series. It is powered by NVIDIA Jetson Thor engines, to meet the demand for high-performance, low-latency edge intelligence in robots and Physical AI applications such as machines.
“Generative physical AI is transforming how machines perceive, decide and act in the real world. From autonomous mobile robots navigating dynamic warehouses to robotic arms adapting grasping in real time, and surgical and humanoid robots performing precise, complex tasks, physical AI enables unprecedented adaptability. Our AIB-AT series is purpose-built to deliver the compute density, industrial reliability, and flexible connectivity required for these applications,” said Richard Hung, Vice President of Product Division at Aetina.
The AIB-AT series includes the AIB-AT78 powered by NVIDIA Jetson T5000 module and the AIB-AT68 powered by NVIDIA Jetson T4000 module. Built on the NVIDIA Blackwell GPU architecture, both platforms deliver exceptional AI compute density in a compact 140 mm × 165 mm form factor. This design enables deployment in space-constrained systems such as humanoid robots, autonomous warehouse vehicles, surgical robots, agricultural drones, and utility inspection robots.
With up to 2070 FP4 TFLOPS of AI compute and 128 GB of memory, the platform delivers up to 7.5X higher AI performance compared to the previous generation, supporting complex workloads such as multimodal perception, sensor fusion, and generative AI inference. In addition, it is engineered with dual RJ45 10GbE ports, enabling high-throughput and low-latency AI data exchange, and with one dedicated RJ45 1GbE EtherCAT support for deterministic motion control. This enables robotics developers to consolidate perception, decision-making, and control into one compact, production-ready system.
Beyond control convergence, the AIB-AT series delivers comprehensive sensor and expansion support tailored for physical AI workloads. High-speed QSFP28 connectivity, optimized for NVIDIA Holoscan, enables stable, low-latency sensor streaming, while a 120-pin MIPI interface supports up to 8x GMSL cameras for advanced multi-view perception. Additional M.2 expansion slots provide PCIe Gen5 NVMe storage, wireless connectivity, and optional 4G/5G with 1PPS function for precision navigation and global time synchronization.
NVIDIA Jetson Thor represents a new class of robotic computing platforms, purpose-built for generative reasoning and multimodal sensor processing. The AIB-AT series supports the full NVIDIA AI’s open models, open-source frameworks and libraries, including NVIDIA Isaac, NVIDIA Metropolis, and NVIDIA Holoscan, as well as agentic AI workflows such as NVIDIA Blueprint for video search and summarization (VSS). From Vision Language Action (VLA) models like the open NVIDIA Isaac GR00T N robot foundation models to open, reasoning VLMs like NVIDIA Cosmos Reason, the AIB-AT series provides developers with a powerful platform foundation, enabling a seamless cloud-to-edge deployment experience and accelerating the implementation of physical AI innovations.
The AIB-AT series is scheduled to be available by the end of Q1 2026.
