ASUS has officially announced the UGen300 USB AI Accelerator, its first AI USB device designed to deliver inference performance directly to any connected system. This slim accelerator houses the Hailo-10H AI processor, providing 40 AI TOPS of dedicated power for running large language models and other generative AI tasks.
The UGen300 features 8GB of dedicated LPDDR4 memory and connects via a USB-C interface, operating at a low power draw of just 2.5W under typical loads. Its plug-and-play design ensures cross-platform compatibility across Windows, Linux, and Android, and it supports major AI frameworks like TensorFlow, PyTorch, and ONNX out of the box.
This dedicated hardware supplements the host device’s CPU and NPUs, enabling demanding on-device generative AI inference without needing constant cloud connections. Users benefit from no subscription fees, zero latency, and enhanced privacy compared to cloud-based AI solutions. The device can handle tasks like text generation, video summarization, and real-time perception.
The dedicated RAM architecture prevents memory bottlenecks common with shared system memory, ensuring consistent throughput for complex AI pipelines. Support is further bolstered by an upcoming UGen Utility that offers quick validation using over 100 pre-trained models, alongside resources from the Hailo Developer Community.
Availability
ASUS Malaysia mentioned that they will be bringing it in sometime during Q2 of this year. There’s also an M.2 version of the UGen300, although that one has yet to be confirmed.
