According to Caifeng News, UBS stated in a recent report that the technology industry has just begun a large-scale growth cycle. It is projected that by 2027, with the widespread application of artificial intelligence in various economies, AI models and applications will become a $225 billion sub-market. From $2.2 billion in 2022, it will be a mere fraction of this figure.
01
AI Large Models Spark a Massive Surge in Demand for Intelligent Computing Power
Similar to the role of water resources in the agricultural era and electricity in the industrial era, computing power is the key productive force in the digital economic era. With AI entering the era of “big models,” the improvement of computing power is also being accelerated.
This is because AI heavily relies on related infrastructure, including computing, storage, and networks. As people’s demands for AI performance increase, the data required for training AI continues to grow, and the complexity of algorithms is constantly increasing, placing higher demands on computing power.
As of the end of June 2023, China’s total number of data center racks exceeded 7.6 million standard racks, and the total computing power reached 197 EFLOPS, ranking second globally. The total computing power has seen an average annual growth rate of nearly 30% over the past five years.
As AI advances to high-application stages such as multi-scenario, large-scale, and integration, the corresponding data volume, algorithm model parameters, and the scale of computing centers focusing on accelerated computing will continue to increase.
AI chips, as chips in the field of artificial intelligence, are modules specifically designed to handle large amounts of computational tasks in AI applications. “No chip, no AI” reflects the high demand for computing power in AI. The computing power achieved through AI chips is an important measure of the level of development in artificial intelligence. Therefore, the development of China’s AI chip industry is closely watched, with new production designers constantly emerging, and the market size is continuously expanding.
02
Large Models Aren’t a Panacea – Computing Power Emerges as the Key Factor
Dr. Wang Shaojun, Partner and COO of Kuncloud Technology, believes that in 2023, there will be an explosive growth of domestic large models in China. However, realizing the practical implementation in the industry will still take time. Currently, many industries still mainly use traditional deep learning methods, such as computer vision (CV) and natural language processing (NLP) algorithms. As new application scenarios become more specific, AI technology needs to be more closely integrated with specific business requirements.
He points out that AI companies not only need to have technical capabilities but also need to collaborate with various industries, gaining an in-depth understanding of business scenarios to truly unleash the practical impact of AI technology.
In the development of AI, how to implement products in real-world scenarios has always been a challenge. Can large models change this situation? According to Wang Shaojun, for regular scenarios, large models are expected to address this issue, but for some long-tail, scarce scenarios, large models still have limitations. In his view, computing power is the most fundamental and highest-priority issue in the era of large models. Without sufficient computing power, the foundation for the effective application of other technologies does not exist.
The competition in AI chip development is becoming increasingly intense. Relying on the existing technical route and advanced processes to enhance chip performance is getting more challenging both from a technological development perspective and in the current international context.
03
AI Chip Industry Poised for Accelerated Embrace of Localization
Despite GPU chip giant NVIDIA’s monopoly advantage in the global AI chip market, the domestication process of Chinese AI chips will accelerate due to the U.S. “chokehold.” In recent years, the United States has gradually tightened China’s ability to access advanced international chips, imposing restrictions not only on chip imports but also on the acquisition of chip production tools.
At the same time, there is a significant talent gap in China, with talent supply and demand ratios for various AI technical positions being below 0.4, including an AI chip position ratio of only 0.32. Therefore, we are continuously increasing investment in education and industrial development, as well as attracting overseas talent for entrepreneurship and employment.
In recent years, under these restrictions, China has witnessed the rise of a batch of AI chip startups, accelerating the development of the computing power industry chain. Examples include domestic CPU+DPU leader Haiguang Information, ASIC route pioneer Cambricon, Jingjia Micro, and Biren Technology. In December 2023, Moore’s Thread held the inaugural ceremony of the first nationally produced thousand-calorie billion-model training platform—Moore’s Thread KUAE Smart Computing Center in Beijing, announcing the official landing of the first large-scale computing cluster based on domestically produced full-function GPU.
Disclaimer:
- This channel does not make any representations or warranties regarding the availability, accuracy, timeliness, effectiveness, or completeness of any information posted. It hereby disclaims any liability or consequences arising from the use of the information.
- This channel is non-commercial and non-profit. The re-posted content does not signify endorsement of its views or responsibility for its authenticity. It does not intend to constitute any other guidance. This channel is not liable for any inaccuracies or errors in the re-posted or published information, directly or indirectly.
- Some data, materials, text, images, etc., used in this channel are sourced from the internet, and all reposts are duly credited to their sources. If you discover any work that infringes on your intellectual property rights or personal legal interests, please contact us, and we will promptly modify or remove it.