How AI Is Fueling China’s Microchip Industry

Ready for a revolution? AI is going to change the world in unprecedented ways. See how it could benefit your business right now. Check out our Free Machine Learning Whitepaper!

In order for China to accomplish its goal of becoming a world leader in AI (artificial intelligence) by 2030, it must manufacture its own microchips to fuel its innovation in the field. Historically, the country has never been a strong leader in the global semiconductor industry.

This will have to change if they want to accelerate AI application development and outpace competitors.

A Unique Opportunity

Although AI developers in China are creating their own AI algorithms, the hardware was designed and manufactured on foreign soil. Microchip manufacturing has always been a problem for China, causing it to rely on integrated circuits from other countries. But when considering variables like shipping time and back-and-forth on design specs, the costs of doing so are starting to greatly outweigh the benefits.

Enter AI. This new technology has undoubtedly opened up a plethora of opportunities. And for China, there’s one, in particular, that’s simply too good to pass up. AI requires microchips that vastly differ from regular computing microchips. If China can develop an expertise in manufacturing these, it will not only become an industry leader in AI algorithms and software but hardware as well.

On top of this, China’s probably the country in the best position to design AI-optimized chips because of its rapid progress in AI development. It would also allow the country to become more economically independent and competitive, something that has become more valuable in recent times thanks to increasing trade tensions.

A New Age for Semiconductors

Some Chinese companies aren’t waiting for a cue from the industry or other companies. Tsinghua Unigroup is a government-backed chip manufacturer that has a new factory the size of several football fields. The factory will start producing silicon wafers by the end of 2019.

The company’s vision is to expand the factory to triple its current size. Tsinghua Unigroup will also create two more factories, with each costing about $24 billion. This is a huge step in the right direction; the Chinese government has been pushing hard for domestic chipmaking in general since 2014. But pivoting these efforts more to AI could prove to be infinitely more fruitful for the country’s future.

There’s still a lot of work to be done. China’s market for semiconductors is the largest and fastest-growing in the world; yet,  no Chinese chipmakers have ever taken the Top 15 spot for sales. These top spots are held by companies in the U.S., Taiwan, Japan, Western Europe, and South Korea. The U.S. alone takes half of all global sales and accounts for half of China’s imported chips.

A Colorful History With Microchips

Mark Li is an analyst at Bernstein, a research firm headquartered in New York City. He tracks Asia’s chip industry, and he thinks even the most advanced Chinese chip manufacturers are behind the rest of the world by five years. With Moore’s Law stating that chip performance doubles every two years, China stands to fall behind even more in the next few years if it doesn’t change its strategy fast.

Although China was beginning to join the microchip race in the 80s, the government and civil unrest prevented progress for more than two decades. In those two decades, microchips advanced rapidly, and the newest chips have billions of transistors. It’s impossible to take apart a chip to reconstruct it; this would be a challenging task even for the most seasoned of experts.

Even if the chips were replicable, knowing how to improve and optimize the current generation of chips would be impossible without years of experience and knowledge. “Manufacturing involves hundreds, even thousands, of technical challenges. It will take a long time to catch up,” said Yungang Bao. Bao is the director of the Center for Advanced Computer Systems at the Chinese Academy of Sciences.

Besides academia and government-backed entities, other Chinese organizations are also putting in effort into this endeavor. Baidu is working on microchips made for deep-learning algorithms. Alibaba created a new company called Pingtouge, the Chinese name for the honey badger, dedicated to designing and manufacturing AI chips. These companies will help lead the way for China to become a true AI leader by 2030.

AI Is the Future of Computing

Kai Yu is an expert in neural networks. He studied AI in China and Germany before founding Baidu’s Institute of Deep Learning in 2013. He’s a vocal supporter of domestic chip hardware manufacturing, and even more so after working at Baidu on deep learning algorithms. In 2015, he recommended that Baidu start manufacturing their own chips specialized for AI.

Later in 2015, he left Baidu to start his own company, called Horizon Robotics. His new company focuses on “application-specific” microchips that can robustly run deep learning algorithms. Some applications include autonomous cars and smarter robots.

He’s very optimistic about AI-specific hardware: “If we look back in 10 years, more than half of the computations on a device will be AI-related.”

A Hopeful Industry

In the past, China’s chip-making goals have resulted in aggressive tactics and espionage. Consequently, the U.S. and other Western countries have tried to counteract any such attempts by actively blocking China from using their chips.

But a trade war is not the clear-cut answer everyone is looking for. Previously, when China faced a trade war, Bao says, “People in China realized that the US can easily stop their progress. It will probably speed things up.”

The truth is that China cannot become a true superpower or a real leader in AI without manufacturing their own hardware. It’s unknown how China’s hardware industry will evolve, but they’re on the right path to hardware independence with domestic companies building their own factories and developing their own chips.

What do you think of China’s AI ambitions? How can a country today secure themselves a spot as an AI leader in the future? Let us know in the comments!