Jiwei Network News, the "2022 Chinese Intelligent Chip Industry Research Report" released by yiou think tank a few days ago predicts that by 2025, the market size of the core industry of Chinese intelligent industry will reach 400 billion yuan, of which the market size of basic layer chips and related technologies will be about 174 billion yuan. GPUs, ASICs, FGPA, brain-like chips and other four major types of artificial intelligence chips and system-level intelligent chips have different degrees of progress in China, and the development of large-scale computing power general chips such as training and inference for the cloud is relatively backward; terminal application chips suitable for more vertical industries such as automatic driving, intelligent security, robots and other special chips have developed rapidly.
Previously, Yiou Think Tank proposed that the development of artificial intelligence chips is divided into three stages, the first stage is due to insufficient chip computing power, neural network algorithms failed to land; the second stage of chip computing power increased, but still can not meet the needs of neural network algorithms; the third stage, GPU and new architecture OF chips to promote the landing of artificial intelligence. At present, with the emergence of the third generation of neural networks, bridging the barriers between neuroscience and machine learning, AI chips are developing closer to the human brain.
Under the guidance of government policies in recent years, the artificial intelligence chip market in mainland China has continued to develop rapidly, and at the same time, the digital transformation of various industries has accelerated, resulting in more diversified application data and more complex deep learning algorithm needs in the artificial intelligence industry, which has promoted the continuous growth of the edge/terminal chip market. Yiou Think Tank predicts that as the overall artificial intelligence market recovers from the epidemic, with the maturity of technology and the transformation and upgrading of digital intelligence, the internal demand will increase, and the market size of the core industry of Chinese intelligent industry will continue to grow steadily, and it is expected to reach about 400 billion yuan in 2025. At the same time, with the increase of large computing power centers and the gradual landing of terminal applications, China's DEMAND for AI chips has also continued to rise, and new chips such as brain-like chips are expected to enter mass production as early as 2023, and will begin to grow significantly in 2024 and 2025, when the market size is expected to reach 174 billion yuan.
In terms of technology trends, Yiou Think Tank pointed out that AI chips will develop in the direction of lower power consumption, closer to the human brain, and closer to the edge. As AI hardware acceleration technology matures, more innovations may come from the combination of circuit and device-level technologies, such as in-memory computing, brain-like computing, or for special computing modes or models, such as sparse computing and approximate computing. At the same time, if the algorithm does not change significantly, according to the current main methods of AI acceleration and the trend of semiconductor technology development, it may reach the limit of digital circuits in the near future (about 1 to 10TFlops/W), and then it will be close to similar computing, analog computing, and even innovation in materials or basic research.
At the application level, with the maturity of technology, the application scenarios of AI chips in addition to the cloud and big data centers, will also gradually move with the computing power to the edge end, gradually focusing on the optimization of special scenarios, and will also be deployed in smart phones, security cameras, autonomous vehicles and other intelligent terminals with increasingly rich types of smart products, and the variety of smart products is becoming increasingly rich, which has spawned a large number of smart chip needs. But for now, the cloud is still at the center of AI, and driven by the need for privacy, cybersecurity, and low latency, higher-performance computing chips and learning architectures will be key to solving the problem.
In terms of investment, compared with 2020, the number of investments in the field of artificial intelligence last year has decreased, but the scale of single investment has shown an upward trend, and the AI chip industry has continued to have capital to enter, with a single financing of more than 100 million yuan, and there are 92 investment events in artificial intelligence chip-related fields throughout the year, with a total amount of about 30 billion yuan. The highest of the disclosed funding amounts was Horizon's $1.5 billion Series C financing completed last year (in December 2020, Horizon announced the launch of a Series C financing with an estimated total of more than $700 million, and has since received $150 million in Series C1, $400 million in Series C2, $350 million in Series C3 financing, and Horizon Series C financing has also exceeded its intended target).
In terms of talents, domestic artificial intelligence enterprises have encountered many obstacles in talent recruitment, talent shortage, high cost is the main problem, the supply and demand ratio of talents in different technical directions is generally lower than 0.4, of which the talent supply-demand ratio of intelligent voice positions is only 0.08.
Looking forward to the future, Chinese smart chip industry challenges and opportunities coexist.
First of all, policy support and market demand are still the main driving forces for the development of artificial intelligence chips. In terms of algorithms, in addition to the research and development of innovative computing paradigms, data problems will also be solved under the guidance of policies, and the value of China's huge data sets will be brought into play, providing AI algorithms with a larger and more accurate data set for learning and training.
Secondly, with the gradual improvement of the underlying technology driven by China's digital transformation, the international influence in the field of AI has risen year by year, and it has gradually established a dominant position in big data, chip design and application landing, attracting more overseas talents to return to China to start a business and find employment. The state has also begun to attach importance to the cultivation of artificial intelligence-related talents, the central and local governments have issued a number of talent training and introduction of related policies; in 2018-2021, more than 300 colleges and universities have opened artificial intelligence majors; some enterprises have also begun to cooperate with universities to jointly cultivate high-quality talents with outstanding comprehensive capabilities in the industry-university-research cooperation teaching model, and further promote the development of artificial intelligence and chips.
Finally, a complete and smooth ecological supervision system will help the technology to fit with rich scenarios and accelerate the landing of AI. In the future, the industrial chain structure may be reconstructed, and more enterprises, universities, and organizations may form a synergy to form a complete technology and business ecology to jointly promote the sustainable development of artificial intelligence and chips.
It should be noted that in the field of chip design and manufacturing, China still lacks design software, there is still a gap between advanced processes and equipment and the world's leading level, and the bottleneck is difficult to break through, so there is still a gap in chip manufacturing and is very dependent on imports. (Proofreading/Mike)