China AI – No impact.

Unlike Huawei, this one is not going to hurt.

  • The entry of Huawei onto the entity list did real damage,e but I think that SenseTime, Megvii and the six others that have been placed on the list will be able to brush this aside without too much difficulty.
  • The US has added eight new companies to the entity list as a result of allegations of their involvement in human rights abuses of Muslim minorities in the Xinjiang province in Western China.
  • 2 video surveillance companies Zhejiang Dahua and Hangzhou Hikvision are included who between them account for around 33% of the global video surveillance market.
  • Also added were SenseTime and Megvii who create facial recognition software and are at the forefront of China’s ambitions to become a global power in AI.
  • Unlike Huawei, neither SenseTime, Megvii or any of the other companies added to the list are dependent on US technology to sell their products and services outside of China.
  • Components to make surveillance cameras are widely available from many geographies and the AI that is being used for analysis is all homegrown.
  • The one potential chokepoint is the silicon upon which the training algorithms are run in the cloud where Nvidia still commands almost all the market.
  • However, I have long been of the opinion that there is not very much that is special about these chips and that Nvidia was lucky when it transpired that GPU’s, which are massively parallel, are also optimal for training AI.
  • Furthermore, at least 70 start-ups have sprung out of nowhere all offering chips that can run AI algorithms in edge devices implying that creating these chips is not very hard.
  • Hence, the Chinese should be able to create their own versions if needed without too much time or effort being required.
  • TSMC will be more than happy to give Chinese customers capacity in its cutting-edge factories meaning that China’s current limitation of 28nm for manufacturing is also not going to be a problem.
  • The end result is that should these companies lose access to Nvidia and other US companies that are supplying them with neural processing units (NPUs), replicating then will not be that difficult.
  • Finally, when it comes to AI itself, China is already running a close second to the US and the nature of deep learning means that Chinese algorithms have to be created at home with Chinese data in order to be effective.
  • Outside of raw computing power, there is very little hardware involvement in AI, which combined with the fact that AI is a very open and collaborative field means that there are very few if any barriers preventing China from becoming a world-class player in AI
  • Hence, RFM thinks that blocking Chinese AI in the USA will have very little if any effect.
  • This is because RFM has concluded that Chinese AI will not work well outside of its borders and the same can be said for American, British, Japanese, Indian and so on.
  • This is due to the inherent weakness of any AI that is based on deep learning which almost all of it is.
  • The only potential problem I can see for these companies is access to non-Chinese capital.
  • Megvii has filed to go public in Hong Kong and support from non-Chinese investors may now evaporate its entry onto the entity list.
  • Even in this situation, I really do not see a huge problem.
  • Both of these companies are backed by the BATMen (see here) and should they struggle to raise capital, I suspect that these companies would step in to support them.
  • Finally, AI is also an important focus for the Chinese government and as a last resort, I am sure that it would support the development of this sector if needed.
  • Hence, I see this most recent move as having no effect other than to put greater pressure on the Chinese negotiators in the upcoming trade talks.
  • China has already emerged as a leader in AI and because the algorithms are still very domain-specific, I think that there is very little that the USA can do to contain its continued development.
  • This is very different from established technologies such as smartphone software or semiconductor manufacturing where China remains completely dependent on US technology.
  • This is a good example of how it is the emerging technologies such as AI, 5G, quantum computing and so on where China will be able to give the US a run for its money and these are the areas to watch.
  • In the established technologies, the race has already been run and China will have to grin and bear it until they become obsolete.
  • For more details please see Reality Bytes – Clash of the Titans, 20th September 2019.

RICHARD WINDSOR

Richard is founder, owner of research company, Radio Free Mobile. He has 16 years of experience working in sell side equity research. During his 11 year tenure at Nomura Securities, he focused on the equity coverage of the Global Technology sector.