To build ultra-large-scale intelligent computing power, the industrial scale exceeds 5 trillion, and the market turnaround is in this direction
The relevant departments of Shanghai issued the Action Plan for Further Promoting New Infrastructure Construction in Shanghai (2023-2026), in which it was mentioned to build an ultra-large-scale autonomous and controllable intelligent computing power infrastructure. Support relevant innovation platforms to take the lead in building major scientific and technological infrastructure with independent and controllable intelligent computing power.
Build an ultra-large-scale intelligent computing power cluster of independent and controllable general artificial intelligence chips, independent controllable optoelectronic hybrid computing chips, independent controllable training frameworks, and independent controllable all-optical switching networks, and take the lead in forming independent and controllable intelligent computing capabilities that support the training of large-scale models with trillion-level parameters, and serve the large model training needs of key enterprises.
First, model training puts forward higher requirements for computing power
According to OpenAI's research, the growth rate of computing power required for AI training has exceeded the growth rate of Moore's Law of hardware, which doubles in 18-24 months, and the computing power of AI is growing exponentially. From 2012 to 2018, the computing power required for training AI models doubled about 3-4 months apart.
From 2012 to 2018, the computing power consumed to train AI models increased by 300,000 times, while Moore's Law increased only 7 times in the same amount of time.
Second, the scale of China's core computing power industry exceeded 1.5 trillion yuan by the end of 2021
According to the "2022-2023 Chinese Intelligent Computing Power Development Assessment Report" jointly released by IDC and Inspur, China's AI computing power will maintain a rapid growth trend, and the scale of intelligent computing power will reach 268 billion billion times per second in 2022. It is predicted that by 2026, China's AI computing power will reach 127.14 billion billion times per second.
According to statistics from the China Academy of Information and Communications Technology, the scale of China's core computing power industry exceeded 1.5 trillion yuan at the end of 2021, of which the scale of related industries exceeded 8 trillion yuan. Under the background of this explosive growth of computing power, the computing power model puts forward higher requirements for the data transmission rate of data centers, and distributed IDC and edge computing are expected to enter the channel of high prosperity.
I carefully selected four leading companies for you, and the market turnaround is here!
The first one: Cambrian
The company focuses on the research and development of artificial intelligence chips and technological innovation, is committed to creating core processor chips in the field of artificial intelligence, capable of vision, speech, natural language processing, the Cambrian 1A processor developed by the company is the world's first commercial terminal intelligent processor IP products, Siyuan 100 (MLU100) chip is China's first high-peak cloud intelligent chip.
The second family: Jing Jiawei
Domestic GPU faucet; In the field of graphics processing chips, the company has successfully independently developed a series of GPU chips with independent intellectual property rights; JM9 series graphics processing chip products meet the high-performance display needs and artificial intelligence computing needs such as geographic information system, media processing, CAD-aided design, games, and virtualization.
The third one: Rockchip
Leading AIoT chip supplier in China; The company has rich experience and technical reserves in large-scale SoC chip design, digital-analog hybrid chip design, image signal processing, high-definition video coding and decoding, artificial intelligence and system software development, the company's main product intelligent application processor chip, is a kind of SoC chip, is a system-level ultra-large-scale digital IC.
The fourth one: polar rotation information
The company plans to invest 300 million yuan in Zhejiang warp speed, which is mainly engaged in the research and development, design and sales of GPU-like chips (VPU chips), AI computing chips and accelerator cards, and its products are mainly suitable for AI computing and video processing.
This content only represents personal views, for reference only, not recommended, the market is risky, investment needs to be cautious! #股票#