ASUS Launches Portable USB AI Accelerator UGen300
ASUS has unveiled a portable neural network accelerator that fits in the palm of your hand and looks like an ordinary USB flash drive. The new UGen300 connects via USB Type-C and instantly adds full local artificial intelligence capabilities to any laptop, smartphone or tablet — no cloud services or subscriptions required.
The device is designed for users who need fast, private and independent AI performance directly on their everyday gadgets. By offloading heavy neural network tasks, the UGen300 frees up the host device’s CPU and built-in NPU, allowing large language models (LLM), vision-language models (VLM) and other generative AI workloads to run locally in real time.
Beyond AI acceleration, ASUS consistently delivers high-performance solutions across its product lines, including essential components like the Asus ROG Equalizer 12V-2x6 cable unveiled with enhanced heat protection, designed to keep your powerful GPUs running smoothly.
Pocket-sized device in the form of a regular flash drive
The UGen300 features an ultra-compact design that slips easily into a pocket or bag. It connects through a standard USB Type-C port, making it compatible with virtually any modern laptop, tablet or smartphone. No external power supply or complicated setup is needed — plug it in and start working with AI immediately.
Hailo-10H chip delivers up to 40 AI TOPS
At the heart of the UGen300 is the specialized Hailo-10H AI processor. According to ASUS, the chip provides performance of up to 40 AI TOPS — sufficient for smooth local execution of large language models, vision-language models and a wide range of generative AI tasks.
The accelerator takes on the bulk of the computational load itself, significantly reducing strain on the host device’s CPU and integrated NPU while enabling direct on-device model inference.
8 GB of isolated LPDDR4 memory
The gadget comes equipped with 8 GB of LPDDR4 memory. Unlike many competing solutions where AI resources are shared with the main system, the UGen300 keeps all its memory fully isolated. This dedicated architecture delivers noticeably more stable and predictable performance.
Just 2.5 W power draw enables smartphone use
Power consumption is kept to approximately 2.5 watts, low enough that the UGen300 can be used safely even with smartphones without causing rapid battery drain. The modest energy requirements make the device truly mobile and practical for on-the-go AI tasks.
Full compatibility with Windows, Linux and Android
ASUS has ensured broad platform support. The UGen300 works natively with Windows, Linux and Android operating systems, extending its usefulness from desktop and laptop environments to mobile devices.
Supports TensorFlow, PyTorch and ONNX with zero extra software
The accelerator is compatible with the most popular AI frameworks, including TensorFlow, PyTorch and ONNX. No additional drivers or software installation is required — simply connect the device via USB Type-C and it is immediately ready for use.
Positioned as a ready-to-go solution, the ASUS UGen300 allows users to run modern neural network models locally, preserving data privacy and eliminating dependence on an internet connection.
