Talent Shortage in the Booming Large Model Sector: Investment Strategy and Planning Analysis
-
What does it take to excel in large models?
Lin Dahua: The most critical factor is talent. Training top-tier professionals for large models is extremely costly, akin to training pilots—requiring extensive flight hours to become qualified experts. The cost of nurturing a large model researcher is no less than that of a pilot, and such talent remains highly scarce today.
Xu Li: We aim to create a 'full-course banquet,' where chefs represent talent, ingredients are data, kitchen tools are computing power, and recipes are algorithms. From an industry perspective, ingredients (data) are more crucial. High-quality, refined industry data becomes the key.
Wang Jian: Scientists should be trained like pilots, emphasizing the importance of experience—mere theoretical knowledge is insufficient. To truly excel in large models, craftsmanship is essential, which sets this industry apart. Scientists trained like pilots are what we lack the most.
In the open-source era, how should China approach large models?
Wang Jian: Open source is a vital tool. Today, open source means finding ways to reduce the cost of technological innovation globally, thereby benefiting everyone.
Lin Dahua: Some say open source is like reinventing the wheel, but attempting to do so is an important innovative approach. If the effort to create a new wheel yields significant value, it’s worth it.
Xu Li: Open source will undoubtedly drive industry growth. True open source involves complex systems and multi-party collaboration—not just horizontal knowledge sharing but vertically advancing innovation across the chain. When cars emerged, real estate developers built garages, and the entire supply chain rapidly iterated. Such collaborative innovation holds immense value for the industry.
Zhao Zhiyun: Large models especially require an open-source and open innovation ecosystem. They must empower diverse industries while involving small and medium-sized innovators in their application and development. In such an ecosystem, the roles of governments, enterprises, research institutions, universities, and innovators are clearly defined.
Large model technology is leading AI to new heights, garnering global attention. For enterprises and AI providers, large models represent a significant developmental opportunity.
In the early stages of large model deployment, commercialization primarily stems from the model layer, with 'large models + computing power' being the most common monetization method. As the ecosystem matures, the application layer will take the lead in commercialization.
The adoption speed of large models varies across industries, with energy and finance being the most proactive due to their strong data, computing power, and AI foundations, largely driven by state-owned enterprises.
2023 China Large Model Market Prospects and Investment Strategy Analysis
As large model deployment deepens, application scenarios and ecosystems grow rapidly, spawning new demands. Market research has revealed early trends, including LLMOps and large model all-in-one machines.
AI’s significant value lies not only in cutting-edge science but also in its deep integration with industries. Only with a robust industrial ecosystem, business models, and a closed-loop supply chain can AI be widely adopted by consumers. Public data shows China’s AI industry is thriving, with a core market size of 500 billion yuan and continuous innovation breakthroughs.
Data from the Ministry of Industry and Information Technology shows that China's artificial intelligence industry is flourishing, with the core industry scale reaching 500 billion yuan and the number of enterprises exceeding 4,300. Innovations such as AI chips, development frameworks, and general large models continue to emerge.
Supercomputing, intelligent computing, and cloud computing are working in synergy, with China ranking second globally in computing power scale. The deep integration of AI with manufacturing has led to the establishment of over 2,500 digital workshops and smart factories, driving the digital, intelligent, and green transformation of the real economy.
Chinese tech companies' AI strategies primarily revolve around "Five New Areas": 1) "New Infrastructure" (AI chips, intelligent computing centers, foundational large models), 2) "New Terminals" (AR glasses, autonomous vehicles, robots), 3) "New Interaction" (AR metaverse), 4) "New Services" (AIGC, ChatGPT-like products), and 5) "New Ethics" (large model ethics, generative AI governance).
"The next step is to strengthen the integration of AI large models with manufacturing technology, industry, and applications, and to develop manufacturing-oriented large models more deeply, providing solid support for the digital transformation and intelligent upgrading of China's manufacturing industry," said Li Bohu, academician of the Chinese Academy of Engineering and expert in system simulation and manufacturing informatization.
According to Tianyancha Research Institute data, AI industry development is closely linked with regional technological innovation. From the perspective of financing events' geographical distribution, Beijing, Shanghai, Guangdong, Zhejiang, and Jiangsu lead with 76, 59, 53, 37, and 30 financing events respectively.
Since being included in national planning, AI technology has become a crucial battleground for future technological innovation. Currently, multiple cities are accelerating in this new field. Since late 2022, many provinces and regions have issued policies and measures related to AI development, breaking barriers from the "top-level design" stage to promote original innovation and data circulation, enabling more effective data utilization and driving healthy industry development.
On July 6, the Ministry of Industry and Information Technology announced at the 2023 World AI Conference that China's AI core industry has reached 500 billion yuan with over 4,300 enterprises, ranking second globally in computing power. Over 2,500 digital workshops and smart factories have been established, reducing R&D cycles by about 20.7%, improving production efficiency by 34.8%, lowering defect rates by 27.4%, and cutting carbon emissions by 21.2% through intelligent transformation.
The "leading goose effect" of AI empowering various industries is accelerating, with foundational technological innovations as the base. Huawei has built an AI computing cloud platform based on Kunpeng and Ascend, along with the heterogeneous computing architecture CANN, full-scenario AI framework MindSpore, and AI development pipeline ModelArts, providing key capabilities for large model development and operation. Based on Huawei's AI root technologies, large model training efficiency can be optimized to 1.1 times that of mainstream GPUs.
In 2023, investment institutions such as MiraclePlus, Innovation Works, and Gaorong Capital led with 8 and 5 investments respectively. The surge in AI computing demand has significantly increased the need for communication products like optical modules, accelerating the iteration to 800G and above products, with high-end products like 800G gradually increasing their shipment share.
In recent years, China's AI industry has flourished, with accelerated infrastructure deployment and deep integration applications. Data from the 2023 AI Conference shows that China's supercomputing, intelligent computing, and cloud computing are working together, ranking second globally in computing power. The deep integration of AI with manufacturing has established over 2,500 digital workshops and smart factories, strongly promoting the digital, intelligent, and green transformation of the real economy.
In the future, as AI development accelerates, industry enterprises can foster ecological prosperity through systematic innovation and open-source practices, supporting high-quality AI development with robust computing power.
The rise of China's large model industry signifies a new chapter in the field of artificial intelligence. Fueled by relentless innovation, it is swiftly integrating into various industries, presenting unprecedented opportunities for the business world. Looking forward, we can confidently anticipate that China's large models will continue to spearhead the tide of innovation.
In the future, we can expect the emergence of more innovative application scenarios, deeper commercial collaborations, and the continuous evolution of technology. Additionally, international expansion and policy support will provide a broader stage for China's large model industry.
We firmly believe that China's large model industry will continue to contribute to global technological progress and commercialization, becoming one of the key forces shaping the future. Let us collectively anticipate and actively participate in this exciting era, contributing to the future development of China's large models.