World’s first USB edge AI Accelerator for both classic and generative AI features built-in USB Type-C connectivity, AI processor, and 8GB LPDDR4 memory
- Unmatched edge Gen AI performance: Built-in Hailo-10H AI processor delivers 40 AI TOPS to support LLMs, VLMs, and more
- Dedicated memory: Empowers host systems that lack AI capabilities to run AI tools effectively
- Plug-and-play convenience: USB-C® interface for Windows, Linux, Android compatibility; also supports AI frameworks like TensorFlow, PyTorch, and ONNX
TAIPEI, Taiwan, April 1, 2026 — ASUS today announced the UGen300 USB AI Accelerator — the first AI USB device from ASUS, bringing inference performance directly to any device. An M.2 version is also available.
This slim AI accelerator measures 105 x 50 x 18mm and features the Hailo-10H AI processor that delivers 40 AI TOPS of dedicated power to support large language models such as LLMs, VLMs, and more. UGen300 includes 8GB LPDDR4 dedicated memory and connects to other devices via a USB-C® interface, consuming just 2.5 watts of power under typical workloads. The convenient plug-and-play design ensures cross-platform compatibility with Windows[i], Linux, and Android[ii]. UGen300 also supports major AI frameworks like TensorFlow, PyTorch, and ONNX — right out of the box.
“By integrating the Hailo-10H into a ubiquitous USB device, ASUS brings the full power of AI and generative AI to everyone” said Max Glover, Chief Revenue Officer of Hailo. “We’re excited to see how our developer community will use this plug-and-play accelerator to push the boundaries of on-device AI. This is exactly how Hailo envisions the future of AI: accessible, affordable, and designed for anyone to build with.”
Unmatched edge Gen AI performance
The ASUS UGen300 USB AI Accelerator has a built-in Hailo-10H AI processor delivering 40 AI TOPS of dedicated performance and is optimized for generative AI workloads, including LLM inference and vision-language tasks. It supplements the host device’s CPU and NPUs and provides AI acceleration to free up system resources, enabling on-device generative AI inference without dependence on cloud computing.
Compared to cloud-based AI, UGen300 users pay no monthly subscription, experience no latency, and enjoy unmatched reliability and privacy. The UGen300 can run demanding Gen AI applications such as text generation, video summarization, event triggering, voice to action, and real-time perception. Its low-power design consumes just 2.5 watts of power to enable efficient, high-performance AI at the edge.
