• United States

Huawei targets Nvidia, Intel, Qualcomm with new AI chips

News Analysis
Oct 15, 20182 mins
Computers and Peripherals

Huawei's artificial intelligence (AI) strategy is not just chips, but a full-stack portfolio, including chips, cloud services, and products.

mwc huawei
Credit: Peter Sayer/IDG

Chinese smartphone giant Huawei Technologies Co. announced at its Huawei Connect 2018 show in Shanghai an update to its Ascend artificial intelligence (AI) chips with a new set of cloud services, software, tools, training, and framework.

The company is putting itself in direct competition with the main AI chip developers in the U.S., namely Nvidia, Intel, and Qualcomm, but also ARM, IBM, to some degree Google, and even fellow Chinese tech giant Alibaba.

Chairman Eric Xu introduced the Ascend 910 and Ascend 310 chips along with the Compute Architecture for Neural Networks (CANN), a chip operator’s library and automated operators development toolkit, and MindSpore, an inference framework for devices, edge networks, and cloud training.

That includes a full-pipeline of services (ModelArts), hierarchical APIs, and pre-integrated solutions, Huawei said, with plans to expand its AI stack to include an AI acceleration card, AI server, AI appliance, and other AI products.

Huawei already has an AI-capable line of processors called Kirin for its own smartphones. And they are pretty ambitious. Its RoadReader project uses the Kirin-powered smartphone to conduct intelligent object recognition to distinguish between thousands of different objects a car may encounter, such as dogs, cats, balls, and bicycles, and “learn to take the most appropriate course of action.”

Huawei said its new AI research will focus on natural language processing, computer vision, decision/interference, and machine learning. It will use the AI internally, as well as offer deployments to businesses across public and private cloud, edge computing, industrial Internet of Things (IoT) devices, and consumer devices.

This is a massively ambitious project, and it is unclear whether the technology will ever make its way out of China. Most of these company-driven AI projects are for the company that made them. Google doesn’t sell its TensorFlow processor, after all. Intel and Nvidia do sell their processors, but they are chip makers, so holding on to them wouldn’t make sense.

Not helping things at all is the fact that trust between the U.S. and China has been frayed due to the Bloomberg/Super Micro issue. Despite the strong denials all around, it still has an air of believability and has rapidly become one of those cases where believability is killing reality.

Andy Patrizio is a freelance journalist based in southern California who has covered the computer industry for 20 years and has built every x86 PC he’s ever owned, laptops not included.

The opinions expressed in this blog are those of the author and do not necessarily represent those of ITworld, Network World, its parent, subsidiary or affiliated companies.