Why does this Tsinghua professor say | Domestically produced AI chips are scarce | 'Sacrificial' | Opportunities

'There are a lot of people who can design a good chip, but the market does not give domestic chips an opportunity to iterate. Domestic companies can't give engineers the opportunity to 'sacrifice'. This is where our problem lies.'

Reporter | Cai Haoshuang Editor | Zhao Li

In the days when the chip, especially the 'SMIC' occupied the headline of public opinion, anxiety, panic, reflection, questioning... all sorts of emotions have emerged.

However, in the long and cumbersome reports, it is easy to clearly and clearly explain the complex 'core' without losing many professional articles.

Starting on Wednesday, looking for Chinese makers (ID: xjbmaker) began to launch a series of planning reports centering on “creating the core business.” At present, there are already “Zhongxing banned,” “SMIC” and “What Ali should do”. "China Core" is not the one that can save China's "core", "Several cold knowledge you should know in the chip industry", "This chip is not an accident, you may not know how hot the AI ​​chip is" and other reports In the coming days, we will also present you more professional and professional 'create customer core' reports from the perspective of entrepreneurship. Please do your best.

In 2 to 3 years (AI chip industry), there will be a low tide. Today, even most of the entrepreneurs will become the martyrs of technological change. ' In early March of this year, the ZTE event had not yet occurred before Tsinghua University micronano Professor Wei Shaojun, director of the Department of Electronics, director of the Microelectronics Institute, boldly predicted at an AI chip summit.

Although artificial intelligence has not yet appeared as a killer application, the field of domestic AI chips is just like a piece of the Red Sea. Internet giants such as Baidu, Ali, Tencent, terminal vendors such as Huawei, DJI, Hikvision, startup companies such as the Cambrian, Horizon, Shen Jian, etc., have joined the AI ​​chip.

China lacks players and capital in the AI ​​chip field. Talent shortages in the design can also be solved in a short period of time through high salaries. Before the outbreak of the ZTE event, Wei Shaojun believed that the AI ​​chip lacked the opportunity to go from 0 to 1 iterations.

Because foreign countries already have relatively mature chip products, domestic terminal manufacturers are accustomed to ignoring domestic brands when selecting chips, and domestic chips lack iterative opportunities.

The ZTE incident once again placed the plight of China's 'coreless souls' in front of the world. But fortunately, the crisis is also a turning point. Song Hui Innovation Fund Investment Director Liang Hui believes that the biggest significance of this chip crisis is to wake up the downstream terminal manufacturers. Let them consciously train domestic chip suppliers.

Difficulties: AI chip lacks opportunities from 0 to 1

Algorithms, data, and computing power are the three key factors that drive the development of artificial intelligence. If big data is the oil of the information age, then the chip that determines the computing power can be regarded as the engine of the information age.

From the PC era to the mobile Internet era to the artificial intelligence era, China’s advantages have always been reflected in big data based on the population base. As the world’s largest semiconductor market, China’s semiconductors rely on imports more than 80%. From 2015, continuous In three years, China’s imports in semiconductors were as high as US$200 billion.

The gap between China and the United States in the field of general-purpose chips is more than ten years, and AI chips are rarely in the same starting line. However, terminal manufacturers selectively ignore domestic products, causing chips to be tested even if they are designed and taped, and it is difficult to verify the products. Iterative upgrade opportunities.

'A lot of people could actually design a good chip, but because the market ecology does not give domestic chips an opportunity to iterate and ignore domestically, domestic companies can't, nor give engineers the opportunity to 'sacrifice'. This is where our problem lies. 'Wei Shaojun once said in front of the media that 'The big companies like Intel and ARM, they designed the first generation of chips are very difficult to use, but they can be iterated, and finally it will be useful. Chip. '

'Kirin's first-generation chip is also not easy to use, but because of its back-to-back Huawei, Huawei's design accommodates many of Kirin's shortcomings. Kirin's chip is also used on Huawei mobile phones. It also has the opportunity to improve iterative. 'A close to Huawei's Hass Insiders told to find Chinese founders.

In the case of the Cambrian era, only chip companies that do not end up as terminals, the current industry is more familiar with the application case, that is, the Cambrian AI chip is integrated into the Huawei mobile phone. There is a view that this is because of the Cambrian period. It is a national team with the background of the Chinese Academy of Sciences.

The above-mentioned insiders told reporters that Huawei's integration of the Cambrian technique is also an expedient measure. When the Huawei masters the core technology, the Cambrian is likely to be knocked out.

'Larger downstream terminal manufacturers, such as Midea, OPPO, Vivo, etc., in the final selection of such components, they still tend to choose foreign, because foreign quality is stable and reliable. ' Liang Hui said.

Before he was an investor, Liang Hui was engaged in chip-related business at Huawei. In his opinion, the chip has a long development cycle. After the first generation of products is designed, it takes a lot of time to iterate. In contrast, foreign countries have already had superior performance and acceptable prices. Mature products, manufacturers did not wait for you to mature patience.

'Foreign OEMs (Original Equipment Manufacturers) can't choose chips made by domestic startups. Can these domestic startups survive? Mainly the attitude of domestic OEMs.'

The ZTE event undoubtedly gave domestic OEMs a blow.

Turning: Terminal manufacturers awaken in pain

'Everyone has recently said that domestic attention to high-end chips is insufficient, but does China really do not pay attention to high-end chips? In fact, large funds (national integrated circuit industry investment funds) have also been engaged in one or two periods, and the country has been investing in this one block. However, the problem is that we have invested a lot in the upstream chip foundry, packaging and testing, and design, but the downstream terminal vendors can't keep up. ' Xu Hui said.

Xu Hui believes that for the domestic semiconductor industry, the most prominent problem at present is not upstream design, manufacturing, and testing, but downstream.

Anhui Chizhou: The company introduced the “Core” team to activate the semiconductor industry/ Vision China Anhui Chizhou: The company introduced the “Core” team to activate the semiconductor industry/ Vision China’s national team to invest more semiconductors in the upstream, if not downstream, the chip is A very complex, very professional project, iterative takes a certain amount of time and accumulation. Downstream terminal manufacturers should consciously adopt, nurture domestic suppliers. '

According to Xu Hui, the biggest significance of this ZTE incident is the awakening of downstream terminal manufacturers, which can give new chip companies the opportunity to try and make mistakes, and provide the chip industry with support in the entire industry chain. 'Because nobody can stand it. This attack by ZTE, including Huawei, will not stand. '

The chip is not only a technical problem, but also an ecological problem. Intel, ARM has established a complete ecosystem from hardware to software for several decades. It is not easy to surpass the domestic market in a short time. Even Huawei, from the beginning of research and development of Kirin chips to The iterative unicorn 970 took nearly 20 years.

The ZTE incident once again pushed AI to the air. This is an opportunity for the Chinese chip industry to suffer. However, Horizon CEO Yu Kai also warned at today’s event: We must also clearly understand that chip R&D is a complex project that requires scientific rigor. Attitude. China's chip should be self-reliant, but it must also be calm, and it should not be a big leap forward for all people to make core.

2016 GoodChinaBrand | ICP: 12011751 | China Exports