NewsCompany newsIndustry News



Artificial Intelligence Chips Falling into the

MetInfo enterprise content manager system | MetInfo CMS

In the third wave of AI, in 2018, the new word "AI chip" has become a big fire. Domestic and foreign technology giants, traditional chip giants and start-ups have laid out the AI chip market one after another. It seems appropriate to describe the current AI chip market with "the brutal growth of grass-roots warblers".

But for AI chips, 2019 should be a critical year. In terms of capital operation, international trade environment and global semiconductor market, after the "100 schools of thought contention" in 2018, it is time for some new things to emerge when AI chips reach 2019.

Artificial Intelligence Chips Falling into the "Homogeneous Growth Circle"

In the past hundred days of 2019, many technology companies at home and abroad have started a lot of big moves in the AI chip market: Yingweida released a new AI chip Jetson AGX Xavier, Yunzhisheng announced AI chip strategy, Intel jointly developed AI chip with Facebook, Huawei bought 500 acres of self-built chip factory in the UK...

In addition to the production of giant AI chips, the huge financing news of AI start-ups driven by capital operation is constantly updated. At least 45 start-ups around the world have developed AI chips (voice interaction and auto-driving), and five enterprises have raised more than $100 million. According to CITIC Securities, the market size of AI chips will reach 14.616 billion US dollars by 2020.

This seems to imply a well-known fact: AI chips are becoming the next trillion market, the next market new blue sea. But unfortunately, the current AI chip market can only be called the "Red Sea" market at best. Like the Red Sea market of AI industry, AI chips have also come into the "homogeneous growth cycle".

At present, AI chips are mainly based on ASIC, FPGA, CPU, GPU and DSP, and these AI chips have a common rule that the universality of the chips is inversely proportional to the performance and energy consumption used to run AI. That is to say, the more versatile AI chips are, the more disappointing their performance and power consumption will be. Each enterprise is producing corresponding AI chips, but these chips can only be called accelerator at present, only in terms of computing power and algorithm has been improved, and the application has not yet been solved. That is to say, most AI chip companies in the market can only be called chip companies at best.

In addition, in the current market of AI chips, which is dominated by ASIC, FPGA, CPU, GPU and DSP, there are four small dragons in China, namely, Bitland, Horizon, Huawei and Cambrian. They basically focus on ASIC schemes, while the schemes of FPGA, GPU and CPU are basically in the hands of foreign enterprises, and the ideas of these companies are almost the same in terms of technology and application. There is no more splendid place.

It should be mentioned that the problem of "AI chip homogeneity" may not be too big in 2018, because AI chips only broke out in 2018. The "wild growth of grass warblers" depends not only on the promotion of market level, but also on the deep demand of AI application landing. But after a year of groping and growing, AI chips have gradually realized that the disadvantages of this "homogeneous growth" must be greater than the advantages.

In order to solve the problem of homogenization, it requires AI chip companies to get out of the differentiation from homogenization. From a technical point of view, enterprises should find their own competitiveness now that there is no big difference in arithmetic and calculating power; and from an application point of view, although "artificial intelligence +" is covering the whole industry chain, so is the artificial intelligence chip, but it should be noted that a single enterprise in the application of artificial intelligence chip, should "penetrate" a single scene before "touch". "Angle" extends to other fields, and this single scene must be more familiar with the enterprise, the field of certainty, bear in mind that we can not blindly follow the trend.

Artificial Intelligence Chip Living in the Period of Publicity

In addition to falling into the vicious circle of homogenization, "living in the propaganda period" is also the confused reality that the current AI chips must look directly at.

Despite the foreign AI chip manufacturers, it seems that domestic manufacturers have been exaggerating since the beginning of propaganda. Because AI has become very popular, many companies have begun to play with concepts, and some companies that make bitcoin miners or DSPs have transformed into an AI chip company. And under the propaganda of the media, there are also many loyal fans.

In the media reports, "China's first", "global leadership" and "breaking foreign monopolies" have been used badly, but enterprises still spare no effort to advertise themselves in the publicity, while rubbing hot spots, hope in the capital market, win the favor of capital. Without hesitation, there must be a lot of moisture in these propaganda. For example, the slogan of an AI chip product of a domestic company says "China's first global leading embedded AI processor", but in fact, two months ago, domestic companies launched similar products.

To some extent, AI chips "live in the propaganda period", and the media should also take part in the responsibility. The media over-hyped the broad prospects of AI chips in China and advocated the omnipotence of AI and AI chips, but in fact AI chips are only chips for specific scenarios and solving specific problems, and can not replace CPU, GPU, FPGA, DSP, NAND Flash, DRAM and other types of chips; some media claimed that China should vigorously develop AI cores. Film, rely on artificial intelligence to break the monopoly of X86 and ARM. This statement sounds exciting, but it is clear that this argument lacks basic industry knowledge. After all, relying on artificial intelligence to overtake in bends, China should still be difficult to achieve in recent years.

And the shortage of AI talents

Artificial intelligence (AI) is becoming a global topic, and the competition for AI talents in various countries is becoming more and more fierce.

Contrary to media hype about the bright future of AI chips in China, there are still many people who are worried about the development of AI chips in China. In addition to the bottlenecks in technology and application, the development of AI and even AI chips in China is still short of talents.

In a comparative study of the current situation of AI between China and the United States, researchers at Oxford University pointed out that in terms of talents, there are about 50,000 AI talents in China and about 830,000 in the United States. The number of AI talents in the United States is not only 16 times that in China, but also more senior. People who have worked in the field of artificial intelligence for more than 10 years account for 70% of the American artificial intelligence talents, but only 40% of the Chinese artificial intelligence talents.

In addition, AI companies in China do not have much financial advantages. Data show that the number of AI companies in the U.S. is more than three times as many as that in China.

However, in such a situation of lack of talents, the tendency of integration of AI in various vertical industries is constantly strengthened. Although the AI talent pool is small, the demand for talent in China's AI field doubled in 2018. As far as the current situation is concerned, the gap of AI talents in China is at least more than 1 million, and because AI requires very high technology, the training time of AI talents is much longer than that of general IT talents. Within a short period of time, it is difficult to fill the gap in China.

Fortunately, with the rapid development of artificial intelligence, big data, robotics and other technology research and industrial landing applications, higher education is speeding up the follow-up of relevant professional settings. Up to now, 101 colleges and universities in China have been approved as "Robot Engineering", 96 as "Intelligent Science and Technology", 203 as "Data Science and Big Data Technology", 25 as "Big Data Management and Application", 25 as "Network Space Security" and 14 as "Internet of Things Engineering".

In addition, although the industry has been boasting that the rise of artificial intelligence will enter or even invade and replace many human occupations, it seems that at present, thanks to the wide application of artificial intelligence, human beings have more platforms and opportunities to show. On April 1, this year, the Ministry of Human Resources and Social Security, the General Administration of Market Supervision and Statistics officially released to the society artificial intelligence engineering technicians, Internet of Things engineering technicians, big data engineering technicians, cloud computing engineering technicians, digital managers, building information model technicians, electronic competitive operators, electronic competitive athletes, UAV pilots and farmers. Industry managers, Internet of Things installation debuggers, industrial robotic system operators, industrial robotic system operators and other 13 new occupational information. This is the first batch of new occupations since the promulgation of the National Classification of Occupations in 2015.

That is to say, in this new round of AI competition, whether it is the training of innovative talents at the national level or the competition for talents of science and technology companies. China is making great efforts to gnaw down the hard bone of "artificial intelligence talent shortage".


As AI enters the era of "great navigation", the outbreak of AI chips has become a natural trend of technology.

In the future era of intelligence, with the continuous renewal of demand, the pattern of AI may undergo tremendous changes, and a new AI chip market is expected to emerge. At present, Zhonggu has the largest artificial intelligence market in the world, and artificial intelligence talents and technology are constantly supplying and developing. In the foreseeable future, the development of AI and AI chips in China will also be able to get rid of the state of being subject to human beings. Only in the past two years, this vision has been basically difficult to achieve.

How does AI continue to penetrate Ping'an City? Why do security enterprises "march into" business one after another? What else should intelligent transportation pay attention to besides "brain"? How to grasp the fragmented market of family and community safety?

On May 23/24, 2019, Euro 100 million will hold GIIS2019 China Smart City Summit, which will continue the theme of the previous two sessions, inviting well-known experts and scholars, leading enterprises in the industry, benchmarking start-ups, well-known investors, etc. to focus on the application status and future development of technology in the field of smart city (safe city, smart business, smart transportation, family community safety).