Chinese technology giant Huawei Technologies is expected to unveil a new artificial intelligence infrastructure technology on Friday that could double the utilisation efficiency of graphic processing units (GPUs), according to the state-owned Shanghai Securities News.

The report said the technology would be able to lift the utilisation rate of AI chips – including GPUs and neural processing units (NPUs) – to 70 per cent, up from the current 30 to 40 per cent, in another major advance in “using software improvements to make up hardware backwardness”.

The new technology, which was scheduled to be launched at an AI industry conference by Huawei, would enable “unified management” of computing resources from Huawei, Nvidia and third-party GPUs, according to the report.

Huawei did not immediately respond to a request for comment.

If confirmed, the technology could boost sales of Huawei’s Ascend AI chips in China and further reduce the country’s reliance on Nvidia chips.

The report said Huawei’s new technology would be similar to a product from Run:ai, a Tel Aviv-founded software firm that Nvidia bought in 2024 for US$700 million, which can orchestrate large-scale AI workloads across GPU clusters.