Policy-Driven AI Chip Industry Accelerates as Institutions Rush to Research Related Concept Stocks
-
With the explosion of ChatGPT, a global wave of AI (Artificial Intelligence) large models has emerged. 'Generative AI is one of the biggest events the semiconductor industry has ever experienced,' said Handel Jones, CEO of IBS, a U.S. electronic industry strategic consulting firm.
Due to the massive computational demands of large model training, GPU (Graphics Processing Unit)-centric computing power has become a critical infrastructure for the development of the large model industry. 'Computing power, as the 'engine' of AI, is the core foundation for AI development. A global 'race' for computing power is underway, with data centers, AI chips, servers, and other components being highly prioritized by governments worldwide,' Wu Qihong, chief researcher at Guangzhou Wanlong Securities Consulting, told Securities Daily.
Supportive Policies Rolled Out Across Regions
AI chips, also known as AI accelerators or computing cards, are modules specifically designed to handle large-scale computational tasks in AI applications. Currently, AI chips are mainly categorized into GPUs, FPGAs (Field-Programmable Gate Arrays), and ASICs (Application-Specific Integrated Circuits).
According to estimates by Morgan Stanley analysts, the AI chip market generated approximately $43 billion in annual sales in 2022, accounting for about 8% of the total semiconductor industry sales. Within four years, the share of AI chips in the industry is expected to nearly double, with revenues reaching $125 billion.
Regarding the current development of the global AI chip industry, Sun Qian, co-chief of the computer industry at Industrial Securities, stated, 'Since the breakthrough in large model innovation in the second half of last year, demand for AI chips has surged, but supply remains concentrated in the hands of a few companies.'
From a market share perspective, a Citigroup research report released on July 10 stated that NVIDIA, the global leader in GPUs, would occupy 'at least 90%' of the AI chip market.
To promote the development of the AI chip industry, governments worldwide have placed great emphasis on it, and China is no exception. Recently, key cities such as Beijing, Shanghai, and Shenzhen have successively introduced policies to support the rapid development of the AI industry.
On May 30, Beijing released the 'Implementation Plan for Building a Globally Influential AI Innovation Hub (2023-2025)' and 'Several Measures to Promote General AI Innovation in Beijing.' On May 25, Shanghai issued the 'Several Policy Measures to Support Private Investment Development.' On May 31, Shenzhen released the 'Action Plan for Accelerating High-Quality AI Development and Application (2023-2024).'
Sun Qian commented, 'Local policies have taken the lead, and national-level policies are also being rolled out.'
Huge Market Potential for Domestic Chips
Currently, China still faces a shortage of AI chips. Zheng Xiaoxi, manager of the Southern Military-Civilian Mixed Fund, noted that the delivery time from overseas suppliers is at least 40 weeks. While many large companies still have stockpiles of computing chips, new entrants to the industry, such as vertical application companies, are already encountering supply issues.
The domestic chip industry has enormous growth potential. The 'Metaverse Blue Book: China Metaverse Development Report (2023)' released by the Beijing Blockchain Technology Application Association and Social Sciences Academic Press shows that from 2020 to 2023, the compound annual growth rate of domestic AI chips reached 95.86%, with the market size expected to exceed 130 billion yuan.
Of course, China's AI chip industry in its infancy also faces 'growing pains.' Zheng Lei, Chief Economist at Samoyed Cloud Technology Group, told Securities Daily, 'Some Chinese companies already possess AI chip design capabilities, but the advanced manufacturing processes required for such chips cannot yet be produced domestically.'
Blue Whale Private Fund Manager Wei Yu shared with Securities Daily, 'First, there is a technological gap between China and global leaders in this field, requiring increased R&D investment to narrow it. Second, the domestic AI chip market is highly fragmented, indicating fierce competition. Third, international chip giants hold advantages in talent aggregation and mergers. In contrast, Chinese firms need to focus more on talent development to enhance competitiveness.'
Despite challenges, domestic AI chip companies have achieved notable progress. For example, Hygon Information's 'ShenSuan No.1' has demonstrated performance comparable to global counterparts in specific applications, according to disclosed technical parameters.
Startups in the AI chip sector are also flourishing. Public data shows a surge in GPU and AI chip ventures from 2019 to 2020, including companies like Biren Technology, Enflame Technology, Moore Threads, and MetaX Integrated Circuits, founded by former AMD director Chen Weiliang. Additionally, Innosilicon is dedicated to developing general-purpose GPU chips.
Wu Qihong noted, 'Domestic innovators in the AI chip industry enjoy fertile ground for growth, presenting rare opportunities for investors. Both semiconductor manufacturers and internet giants are actively enhancing their AI chip R&D capabilities to share in the industry's growth dividends.'
Supply Chain Poised for Rapid Development
In the secondary market, the AI chip sector has performed strongly this year, with the Tonghuashun AI Chip Index rising 38.38% as of July 28, outperforming the Shanghai Composite Index (up 6.04% year-to-date). Among 22 AI chip stocks, 12 have gained over 20% this year, with Cambricon and Sugon more than doubling in value. Companies like Huaxi Shares, Hangyu Micro, and Omnivision have seen gains exceeding 50%.
The sector's growth is driven by three factors, according to Liu Yan, Chairman of Anjue Asset: 'Market demand, rapid technological advancements, and policy support have all significantly boosted the AI industry, raising expectations for AI chips' future development.'
Notably, Tonghuashun data shows that 14 AI chip companies received institutional research visits in June and July, with 11 attracting over 20 institutions each. Omnivision led with 176 visits, followed by Goke Microelectronics, Jingjia Micro, and Hikvision, each with over 50 visits.
Sun Qian remarked, 'While the AI chip industry is currently oligopolistic, its long-term prospects are promising. The next 3–5 years will be a period of sustained high growth.'
Wu Qihong added, 'Global demand for AI chips is rising, with TSMC's 5/7nm process orders extending into 2024, representing at least a 20% increase over 2023. Edge computing complements traditional cloud computing, with Huawei and Qualcomm exploring AI model deployment on mobile devices, positioning them well in edge AI. Investors should monitor domestic GPU chip developers and IoT module suppliers.'
Zheng Lei stated, "AI chips and AI servers represent a sector with long-term growth potential. Whether it's graphics processing chips or solutions and products following other technological paths, there are both opportunities and challenges. A diversified portfolio can be formed to reduce investment risks."
Multiple securities firms have recently noted in research reports that domestically developed large AI models have entered a "blooming" phase, further igniting massive demand for computing power. As technology companies worldwide continue to increase investments in the field of large AI models, the chip industry chain related to AI is expected to experience rapid development.