Richard Windsor
8 September 2017

Huawei - The AI of others

Huawei needs its own algorithms to succeed in AI

. Huawei abandoned its habit of launching a new phone at IFA 2017 and instead focused on a new chipset called the Kirin 970 that promises all usual the bells and whistles as well as artificial intelligence. Huawei made some bold claims regarding its hardware performance as well as power efficiency thanks to its 10nm geometry but we get the impression that it intends to drive differentiation through its embedded neural processing unit (NPU). This is a part of the chipset that has been specifically designed to run AI algorithms more quickly and more efficiently than running them on the CPU or in the cloud. The result should be faster processing of AI tasks resulting in better services that drain the battery less.

This is all well and good but what really matters is what users of Huawei devices will notice, to whom they will attribute the value created and for what they will pay. The Kirin 970 NPU supports Huawei’s own APIs as well as Google’s TensorFlow and Facebook’s Caffe 2 meaning that AI created by these two ecosystems will also run optimally on the NPU. The idea is that the algorithms are created in the cloud, downloaded to the device where they run locally improving both speed as well as privacy as the data will not leave the device. We have long believed that this type of AI will be limited to functions where the algorithms are very well established. In the early days this is likely to be image processing such as facial recognition or computer vision. This is where Huawei will begin to struggle as it has very little AI of its own meaning that the Kirin 970 will spend almost all of its time processing the AI of others. The AI of others will be running on the devices of all of Huawei’s competitors meaning that Huawei will be competing purely in hardware performance.

When other chipmakers come to market with their own NPUs, it will then be a straight fight based on hardware performance. When it comes to AI, users are going to place value of the depth, richness and intuitiveness of the services themselves meaning that to improve its differentiation, this is where Huawei needs to compete. Of this there is no sign meaning that while the Kirin 970 may help Huawei increase market share, it will do nothing to enable it to increase the prices of its phones. The net result is that until Huawei can outsell Samsung by a factor of 2 to 1 in terms of volume, it will really struggle to increase its margins beyond the 2-4% that everyone else (except Samsung) is stuck with. Google is the only company that really makes money from Android but we continue to be cautious as its valuation is already pretty full. Tencent, Baidu and Microsoft remain our top choices.

Disclaimer - Past performance is no guarantee of future results. Inherent in any investment is the potential for loss. This material is being provided for informational purposes only and nothing herein constitutes investment, legal, accounting or tax advice, or a recommendation to buy, sell or hold a security. This document may contain materials from third parties, which are supplied by companies that are not affiliated with Edison Investment Research. Edison Investment Research has not been involved in the preparation, adoption or editing of such third-party materials and does not explicitly or implicitly endorse or approve such content. No recommendation or advice is being given as to whether any investment is suitable for a particular investor. It should not be assumed that any investments in securities, companies, sectors or markets identified and described were or will be profitable. All information is current as of the date of publication and is subject to change without notice. While based on sources believed reliable, we do not represent this material as accurate or complete. Any views or opinions expressed may not reflect those of the firm as a whole. Edison Investment Research does not engage in investment banking, market making or asset management activities of any securities. The material has not been prepared in accordance with the legal requirements designed to promote the independence or objectivity of investment research.