News Center



Home News Center News

Software Industry Boom Serves as a Barometer for the Outbreak of AI Large Model Applications

Jun 06-2025   



*This article is based on a keynote speech delivered by An Xiaopeng, Vice President of AliCloud Intelligence and Executive Committee Member of the China Informatization Society.

 

Three Revolutions Corresponding to AI Large Models

In my opinion, AI is indeed opening up a brand new era of intelligence, which contains three revolutions:

The first is the computing paradigm revolution. The computing unit has shifted from the CPU to the GPU. The second is the cognitive collaboration revolution. The Internet's purpose is to provide accurate information, while AI aims to find the most relevant answers or solutions to problems. The third is the human-computer interaction revolution. Nowadays, AI technology enables people to interact with machines through natural language, fundamentally changing the way humans interact with computers.

 

Classification of AI Large Models & Their Characteristics

There are many current AI Large models, which can be roughly classified into 5 categories:

1. Language models, such as ChatGPT and Tongyi, which are similar to humanities students and are good at language processing;

2. Reasoning Large models: OpenAI GPT-3, Tongyi QWQ32B, etc. They are based on linguistic large models and are trained by science students, reaching the level of PhD students.

3. Multimodal Large Model can generate and understand pictures and text, and are similar to literature students.

4. Models that can enter the physical world, such as robots and self-driving models based on Transformer technology. These can be regarded as athletic learners with both ability and intelligence.

5. Scientific large models differ from ordinary scientific research in that they can achieve the level of the Nobel Prize.

Currently, the AI Large Model is the world's 25th general-purpose technology because it meets four major conditions: extensive integration into various fields; improvement of production efficiency; complementarity with existing technologies; and transformation of production relations and organiational methods.

From the perspective of global industrial development, today's AI Large Model has two significant features. The first is the 'universal effect' at the application level. Secondly, the 'polarisation effect' at the supply level. This is especially evident in the highly concentrated investment in talent, models and algorithms.

 

How Can China Break out of the Large Model Dilemma?

Firstly, this iteration of products is centered on the idea that all intelligent hardware, software and data will be driven by AI. Secondly, the focus has shifted from a 'revolution of tools' to a 'revolution of productivity'. This means that AI will not only improve efficiency, but also prompt enterprises to redefine their markets, customers, needs, business processes and business models, and even organizational forms.

Two judgements can be made about the current AI application:

1. The gap between China and the United States in terms of supply capacity for large models is narrowing, but the gap at the application level is widening. Revenue data shows that Microsoft Copilot's annualized revenue has reached $13 billion, marking a 175% year-over-year growth. Several AI large model companies in the United States are expected to generate $7 billion in revenue in 2024. Meanwhile, China's AI large model revenue is estimated to be 1% of that of the U.S. Investment data indicates that the United States will attract $100 billion in private investment for AI in 2024, while China's investment is expected to reach $9 billion. Company data indicates that 1,000 new AI companies have been established in the US, compared to 98 in China. Additionally, the US boasts 27 AI unicorn companies, each with an average valuation of $2 billion, whereas China has a smaller number.

2. Although all kinds of large model ranking lists are popular, the market is the ultimate standard by which the value of technology and product competitiveness can be tested. China's AI Large Model faces challenges due to the highly fragmented To B market, the lack of a unified big market and enterprises' tendency towards private deployment, which leads to a lack of competitiveness in AI applications.

Is China's AI Large Model industry heading towards the mobile internet development model or the SaaS industry dilemma? Only time will tell how the industry will develop, depending on the degree of effort and the right approach.

Software Industry Boom Serves as a Barometer for the Outbreak of AI Large Model Applications

Jun 06-2025   



*This article is based on a keynote speech delivered by An Xiaopeng, Vice President of AliCloud Intelligence and Executive Committee Member of the China Informatization Society.

 

Three Revolutions Corresponding to AI Large Models

In my opinion, AI is indeed opening up a brand new era of intelligence, which contains three revolutions:

The first is the computing paradigm revolution. The computing unit has shifted from the CPU to the GPU. The second is the cognitive collaboration revolution. The Internet's purpose is to provide accurate information, while AI aims to find the most relevant answers or solutions to problems. The third is the human-computer interaction revolution. Nowadays, AI technology enables people to interact with machines through natural language, fundamentally changing the way humans interact with computers.

 

Classification of AI Large Models & Their Characteristics

There are many current AI Large models, which can be roughly classified into 5 categories:

1. Language models, such as ChatGPT and Tongyi, which are similar to humanities students and are good at language processing;

2. Reasoning Large models: OpenAI GPT-3, Tongyi QWQ32B, etc. They are based on linguistic large models and are trained by science students, reaching the level of PhD students.

3. Multimodal Large Model can generate and understand pictures and text, and are similar to literature students.

4. Models that can enter the physical world, such as robots and self-driving models based on Transformer technology. These can be regarded as athletic learners with both ability and intelligence.

5. Scientific large models differ from ordinary scientific research in that they can achieve the level of the Nobel Prize.

Currently, the AI Large Model is the world's 25th general-purpose technology because it meets four major conditions: extensive integration into various fields; improvement of production efficiency; complementarity with existing technologies; and transformation of production relations and organiational methods.

From the perspective of global industrial development, today's AI Large Model has two significant features. The first is the 'universal effect' at the application level. Secondly, the 'polarisation effect' at the supply level. This is especially evident in the highly concentrated investment in talent, models and algorithms.

 

How Can China Break out of the Large Model Dilemma?

Firstly, this iteration of products is centered on the idea that all intelligent hardware, software and data will be driven by AI. Secondly, the focus has shifted from a 'revolution of tools' to a 'revolution of productivity'. This means that AI will not only improve efficiency, but also prompt enterprises to redefine their markets, customers, needs, business processes and business models, and even organizational forms.

Two judgements can be made about the current AI application:

1. The gap between China and the United States in terms of supply capacity for large models is narrowing, but the gap at the application level is widening. Revenue data shows that Microsoft Copilot's annualized revenue has reached $13 billion, marking a 175% year-over-year growth. Several AI large model companies in the United States are expected to generate $7 billion in revenue in 2024. Meanwhile, China's AI large model revenue is estimated to be 1% of that of the U.S. Investment data indicates that the United States will attract $100 billion in private investment for AI in 2024, while China's investment is expected to reach $9 billion. Company data indicates that 1,000 new AI companies have been established in the US, compared to 98 in China. Additionally, the US boasts 27 AI unicorn companies, each with an average valuation of $2 billion, whereas China has a smaller number.

2. Although all kinds of large model ranking lists are popular, the market is the ultimate standard by which the value of technology and product competitiveness can be tested. China's AI Large Model faces challenges due to the highly fragmented To B market, the lack of a unified big market and enterprises' tendency towards private deployment, which leads to a lack of competitiveness in AI applications.

Is China's AI Large Model industry heading towards the mobile internet development model or the SaaS industry dilemma? Only time will tell how the industry will develop, depending on the degree of effort and the right approach.