ASUS UGen300 Βrings Hailo-10H Εdge AI Αcceleration Οver USB-C
The ASUS UGen300 is a USB AI accelerator with Hailo-10H, 40 TOPS performance, and 8GB memory, enabling plug-and-play edge inference on Linux, Windows, and Android systems.
ASUS UGen300
ASUS has announced the UGen300 USB AI Accelerator, a compact external device that brings hardware-accelerated AI inference to existing systems through a USB connection. The device targets developers and system integratorσ who need local AI processing without modifying internal hardware or adding PCIe expansion cards.
The UGen300 focuses on edge inference workloads rather than AI training. Users can connect the device to laptops, mini PCs, industrial computers, or embedded platforms and enable AI acceleration instantly. ASUS positions the USB AI Accelerator as a simple way to deploy generative and vision-based models on systems that lack dedicated AI hardware.
ASUS UGen300 USB AI Accelerator Inside
The accelerator uses the Hailo-10H neural network processor, which delivers up to 40 TOPS (INT4) of inference performance. ASUS integrates 8 GB of LPDDR4 memory directly into the device, allowing AI models to execute locally on the accelerator. This design avoids loading models into host system memory and reduces CPU utilization during inference.
The UGen300 connects through a USB-C port using a USB 3.1 Gen 2 interface with bandwidth up to 10 Gbps. This approach enables plug-and-play operation across x86 and Arm platforms without requiring internal expansion slots. ASUS confirms compatibility with Linux, Windows, and Android, which supports a wide range of development and deployment scenarios.
Hailo-10H Software Stack
Key features of the ASUS UGen300 USB AI Accelerator:
- Hailo-10H AI accelerator delivering up to 40 TOPS (INT4)
- Integrated 8 GB LPDDR4 memory for on-device inference
- USB-C interface with USB 3.1 Gen 2 (10 Gbps)
- Typical power consumption of approximately 2.5 W
- Designed specifically for edge AI inference workloads
- Supports LLMs, VLMs, Whisper, and vision networks
- Compatible with x86 and Arm host systems
- External enclosure for easy deployment and portability
The USB Accelerator uses the Hailo software stack and supports common machine learning frameworks, including TensorFlow, TensorFlow Lite, Keras, PyTorch, and ONNX. ASUS also provides access to an online model zoo with more than 100 pre-trained models covering generative AI, speech recognition, and classical computer vision tasks. This software ecosystem allows developers to deploy and evaluate models quickly without extensive host-side optimization.
ASUS UGen300 USB AI Accelerator Front
Previously, we covered the Axelera AI Metis M.2 AI Accelerator, the Axelera AI Metis Compute Board, and the Radxa AICore AX-M1 M.2 module. You can check those products if you are interested in similar AI acceleration solutions.
ASUS has announced the UGen300 USB AI Accelerator but has not yet disclosed pricing or availability. Interested users can monitor the official ASUS product listings and distribution partners for updates. Once released, the device should appeal to developers seeking a low-power, external AI accelerator that works over USB without requiring PCIe or M.2 hardware upgrades.
Images used courtesy of ASUS
