How Alibaba’s HanGuang 800 AI Chip Redefines Cloud Computing Power

Alibaba Unveils Its First Self‑Developed AI Chip

At the September 25 Cloud Xi conference, Alibaba announced the launch of its first self‑designed chip, the HanGuang 800, hailed as one of the world’s most powerful AI processors.

The chip’s name, meaning “the hidden light,” reflects its invisible yet strong computing power. In tests, a single HanGuang 800 delivers the equivalent of ten GPUs in the Hangzhou City Brain workload.

In the industry‑standard ResNet‑50 benchmark, the HanGuang 800 achieves 78,563 IPS, four times the performance of the current best AI chip, and an energy efficiency of 500 IPS/W, 3.3 × higher than the runner‑up.

Hardware‑Software Co‑Design Drives Performance

The breakthrough stems from tight hardware‑software integration: a custom chip architecture combined with Damo Academy’s advanced algorithms, deeply optimized for CNN and vision tasks, enabling large models to run on a single NPU.

Real‑World Applications

HanGuang 800 is already deployed in Alibaba’s core services. In the City Brain demo, processing traffic video for the main urban area dropped from 40 traditional GPUs with 300 ms latency to just four HanGuang 800 chips with 150 ms latency. Product image recognition for the “Pailitao” catalog, which adds 1 billion new images daily, shrank from one hour on GPUs to five minutes.

The chip will be offered as an AI compute service on Alibaba Cloud, promising a 100 % cost‑performance improvement over conventional GPU solutions.

Damo Academy’s Two‑Year Milestones

Since its founding two years ago, Damo Academy has published over 450 papers at top international conferences and secured more than 40 world‑first achievements in natural language processing, speech, and visual computing.

The institute now hosts over ten IEEE Fellows, dozens of renowned professors, and a majority of scientists with PhDs from elite universities.

Youth Innovation – The “Qing‑Orange” Award

The second‑edition Qing‑Orange Award recognized ten young researchers, each receiving ¥1 million and full access to Alibaba’s global R&D facilities, data, scenarios, and computing resources. The youngest laureate is 28 years old.

Recipients gain dedicated technical and engineering support to transform scientific ideas into practical applications, exemplified by the 2018 winner’s fingerprint‑based data‑traceability platform.

Overall, the conference highlighted how cloud computing, big data, AIoT, and mobile integration are shaping the digital economy, with Alibaba’s AI chip and research initiatives at the forefront.