Chinese tech giant Huawei, in partnership with universities, unveiled and open-sourced an innovative AI technology on Friday (November 21), designed to dramatically boost the utilisation of computing resources.

The framework, called Flex:ai, partitions a single AI chip into multiple virtual units with a granularity as fine as 10 per cent. This “one-card-performing-multiple-tasks” capability, enabled by flexible resource isolation, can boost average compute utilisation by 30 per cent in these scenarios.

The rapid expansion of the AI industry is fueling massive demand for computing power, while low utilisation rates of these resources have become a major bottleneck, with significant wastage hindering progress.

By open-sourcing Flex:ai, Huawei is set to grant global access to the full suite of its core technologies, according to Zhou Yuefeng, Huawei’s vice president, who introduced the framework at a Shanghai forum.