Published in


A Look at China’s Growing Semiconductor Industry

In February Chinese AI semiconductor maker Horizon Robotics raised US$600 million in Series B funding from SK China, SK Hynix and big auto groups, making it the first in its field to reach the staggering valuation of US$3 billion. Company CEO Kai Yu, a former Dean of Baidu Institute of Deep Learning, says Horizon’s next goal is to become the “Intel of edge computing” in areas such as intelligent driving and AIoT.

Market intelligence company Tractica forecasts the global deep learning chipset market will surge from US$1.6 billion in 2017 to US$66.3 billion by 2025. While many Chinese startup companies focus on AI applications, Horizon Robotics is among the few tackling hardware infrastructure, where China currently falls short.

Most Chinese semiconductor startups are operating on a net loss, and analysts expect this may last into the next two years. Further, the United States has placed export controls on close to a dozen AI and machine learning technologies. Chinese semiconductor companies are however catching up, propelled by multiple government policies, a firm push in R&D, and massive capital injection.

There are two primary application fields Chinese chip companies cater to: Security and Smartphones. China will have over 600 million security cameras by 2020. The country’s compound annual growth rate for Security from 2018–2022 is estimated at 22.6 percent, with wide-range applications in public security, urban transportation, eco-architecture and industrial parks.

Chips are the most expensive component in video surveillance. A typical device set requires four types of chips: two types of SoC accelerators, one for deep learning, and an ISP chip. Most current AI chip solutions use GPUs, but more FPGA/ASIC solutions emerging for example from Cambricon and DeePhi, while Huawei Hisilicon has become a big player in specialized SoC.

The other market worth watching is smartphones, where AI is the new buzzword. The Huawei Mate 10 and Glory V10 for example have the cutting-edge Huawei Neural Processing Unit (NPU) Kirin 970. Marketing gimmick or not, AI integration has dramatically increased in Chinese smartphone applications since 2017. Facial recognition, beauty cams, VR applications and voice assistants are now common smartphone features. Strategy Analytics predicts 90 percent of smartphones will be equipped with AI assistants by 2023.

Deep learning requires massive computing resources, and CPUs are at a natural disadvantage when it comes to images and video processing. At the moment the most compatible chips for deep learning training are GPUs produced by NVIDIA and AMD. The chips are however quite expensive and power consuming. The next best option is FPGAs (Field-Programmable Gate Arrays), which are commonly used for deep learning inferencing and manufactured by companies such as Xinlinx, Lattice, and Intel Altera. These adaptive, semi-customized chips have lower consumption and latency, but fall short with floating point speed and can have complicated wiring issues.

An emerging option is ASICs (Application-Specific Integrated Circuits), which are are fully customizable, but whose initial implementation is also relatively costly. ASICs are being developed by multiple tech companies such as Google TPU, Huawei Hisilicon, and Amazon’s ASIC. Meanwhile, the Google third generation TPUs that power TensorFlow enable more advanced ML instructions per watt than commercial GPUs or FPGAs.

Tractica’s projected market share for AI semiconductors

The edge computing market is expected to represent more than three-quarters of the total market opportunity, aside from cloud and data centre environments. Chinese startups lack the experience of large semiconductor companies with CPU, GPU and FPGA designs, therefore ASICs are seen as a catch-up opportunity. Current Chinese companies in this field include:

  • Bitmain (Series B+ $440 million, 2018–08)
  • Cambricon (Series B Hundreds of million approx., 2018–06)
  • Horizon Robotics (Series B $600 million, 2019–02)
  • Kneron (Series A+ $18 million, 2018–05)
  • NextVPU (Series A $28.82 million, 2018–10)
  • Easytech (Series A Undisclosed, 2019–03)
  • DeePhi Tech (Acquired by Xinlinx, 2018)
  • Huawei HiSilicon (Established in 1991)
  • Baidu
  • Alibaba
  • Hikvision

With AI deployment rapidly spreading to cars, phones, drones and robots, China recently launched a Science and Technology Innovation Board (STIB) on the Shanghai Stock Exchange (SSE) — something like Nasdaq in the US. As the SSE explains on its official website, “establishing a sci-tech innovation board and piloting the registration system is a key reform initiative for the capital market.” With markets open and ready for the next mega chip company to emerge, Chinese hopefuls will need to withstand foreign competition while developing their own advantages and demonstrating profitability.

Localization: Meghan Han | Editor: Michael Sarazen

2018 Fortune Global 500 Public Company AI Adaptivity Report is out!
Purchase a Kindle-formatted report on Amazon.
Apply for Insight Partner Program to get a complimentary full PDF report.

Follow us on Twitter @Synced_Global for daily AI news!

We know you don’t want to miss any stories. Subscribe to our popular Synced Global AI Weekly to get weekly AI updates.



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store