AI is Moving Forward

The epidemic is a challenge even an opportunity for AI, which plays a significant role in the fight against COVID-19 to disinfect, delivery, and stand guard at the highway at the front, decreasing the risk of the medical staff infected. AI has been invisibly accepted and trusted by the public through the epidemic. AI, an important point of the fourth industrial revolution, has been layout by countries around the world, which gradually permeated and creatively integrated with all industries.

Improved by the Policies, Technologies and Demand in China

There is a fantastic environment of policies, technologies， and demand to drive AI development. AI has been considered into the new infrastructures as an important grip of the information technology and adopted 3 steps by the government, thus, it ushers a new opportunity. And there is a significant breakthrough in big data and a basic algorithm and a maturity of vision algorithm and phonetic algorithm. With industrialization and informatization, as well as the deep application in the field of security and medical care, it exists vast demand to meet.

AI Industry Chain

The market scale of AI in 2019 reached 271 billion USD and it is estimated that it will be 914 billion USD in 2025 under the absolute drive by the world.

Through massive data input and simulation with the algorithm, AI is formed to deal with practical application problems. It consists of 3 layers of foundation, technology, and application. Foundation layer is to offer the data and compute ability. The technology layer refers to explore algorithm pattern and obtain the intelligent technology through the training of software framework. The application layer is that AI is utilized commercially and touched widely for us.

GPU Builds Core Computing Unit with FPGA and ASIC

GPU is a large-scale parallel computing architecture composed of a large number of cores, which is specially designed for processing multiple tasks at the same time. It has powerful computing power, more than 80% of which are computing units, and is good at large-scale parallel computing. GPU market is absolutely occupied by NVIDIA, followed by AMD and entered by Intel planning, but it is just in the initial stage in the field of GPU in China and Bitmain launched the first generation of products in 2019 with a quite small market.

FPGA integrates a large number of basic gate circuits and memory, with flexibility between CPU, GPU, and ASIC, performing many merits such as strong computing ability and low cost. At present, the global FPGA market is monopolized by Xilinx, Intel, and lattice with the market of the first two taking 70 percent of the whole share.

ASIC is a kind of customized chip designed for a specific purpose, which has the strengths of small size, low power consumption, and higher reliability. ASIC is mainly used for deep learning acceleration. According to Google data, TPU is 15-30 times faster than GPU or CPU at that time, whose ratio of energy consumption is higher than them over 30-80 times. Foreign plays are mainly dominated by Google, while domestic companies are mainly Cambrian. In recent years, the industry presented a state of blooming together due to some domestic AI algorithm companies have started to extend to the downstream, develop their own chips, or jointly research chips by their algorithm characteristics.

In the next few years, global chip companies, including Internet giants and start-ups, will compete in this market. It is estimated that the global market size will reach 32.3 billion USD by 2023 with a CAGR of about 50 percent.

The demand for AI infrastructure showed a stably increasing growth with an estimated market scale reaching 400 million USD with a GR of 30 percent between 2019 and 2022. For one thing, AI infrastructure is regarded as one and only solution to meet the huge amount of calculation. For another thing, there is a large growth of AI infrastructure demand in traditional industries including the traffic, administration, and others,promoting a quickly increasing demand for AI servers particularly.

At present, the AI infrastructure with GPU, FPGA, and ASIC as the core computing unit gradually replaces the previously main X86. Data shows that in 2018, the market sales of AI infrastructure in China reached 1.3 billion USD with a GR of 132%. The domestic AI infrastructure market is mainly occupied by server manufacturers such as Inspur, Huawei, Sugon, and H3C, in which GPU is the focal point.

With the implementation of the relevant AI development strategies in China, the AI application market is the most active, and domestic AI enterprises are mostly concentrated in this sector whose space is also gradually broader. At present, it is commonly known that AI is widely used in medical treatment, transportation, and intelligent manufacturing, and other fields.