Home About us

China has 70%! Global GenAI patent application ranking: Tencent ranks first, Huawei ranks 13th!

Xinzhixun 2024/07/05 12:40

Xinzhixun, China excludes 70%! Global GenAI patent application ranking: Tencent ranks first, Huawei ranks 13th!

The latest report released by the World Intellectual Property Organization (WIPO) on July 3 pointed out that with the continued popularity of generative artificial intelligence (Gen AI) in recent years, the number of global Gen AI-related patent applications has also shown explosive growth, with the cumulative number of patent applications reaching about 54,000 in the decade from 2014 to 2023, but about 25% of them were filed in 2023. For example, Tencent Holdings, Ping An Insurance, and Baidu, the world's top three companies, are all Chinese companies, which also drives China to rank first in the world in terms of the number of Gen AI-related patents, reaching 38,210, accounting for about 70% of the world's total, and six times that of the United States, which ranks second.

The field of Generative Adversarial Networks (GANs) has the highest number of patent applications

Specifically, WIPO invoked data from EconSight/IFI Claims to assign all identified GenAI patent families to five different model types based on information in patent abstracts, claims, or titles, including: GANs (generative adversarial network models), LLMs (large language models), VAE (variational autoencoder models), Diffusion models (mainly used in diffusion models used primarily for image and audio generation), Autoregressive models. However, many of the GenAI patent families are not suitable for any particular model type, and many GenAI patents do not contain keywords about the specific model used in the patent abstract, claims, or title, but instead focus on describing the use case of the patent, with only a general description of the GenAI process used. This makes it difficult to find patents and map them to the five core GenAI models, which also have some overlap in content. As a result, since 2014, only about 25% of all GenAI patent family publications can be mapped to one of these five models.

In GenAI's core models, many patent families fall into the GAN model category. Between 2014 and 2023, the number of patent family publications belonging to this model was close to 9,700, and nearly 2,400 patent families were published in 2023 alone. The VAE model and the LLM model are the second and third most filed patent families, with approximately 1,800 and 1,300 patent families filed between 2014 and 2023, respectively.

Xinzhixun, China excludes 70%! Global GenAI patent application ranking: Tencent ranks first, Huawei ranks 13th!

As can be seen from the chart above, in terms of patent growth, GAN-related patent families have shown the strongest growth over the past decade. However, in the last three years, the growth of patents in GAN-related patent families has slowed down and increased only slightly. Patent families for VAE and autoregressive models have seen a similar slowdown in patent growth. In contrast, the diffusion model and LLM have grown at a much higher rate over the past three years, with the number of patent families for the diffusion model increasing from 18 in 2020 to 441 in 2023 and LLMs increasing from 53 in 2020 to 881 in 2023. The GenAI boom sparked by AI chatbots like ChatGPT has clearly increased research interest in LLMs.

Xinzhixun, China excludes 70%! Global GenAI patent application ranking: Tencent ranks first, Huawei ranks 13th!

Over the past 10 years, the growth of patents for autoregressive models has been the weakest. However, it is important to point out that there is some overlap in the content of different GenAI models, and some patent families belong to multiple models. For example, there is an overlap between autoregressive models and large language models because LLMs are autoregressive by definition, but not all autoregressive models are LLMs. As patent terminology changes over time, it is likely that newer patent families are more likely to use terminology large language models than autoregressive models in patent titles, claims, and abstracts due to the significant increase in the popularity of large language models. This may explain the weak growth momentum of patents in the autoregressive model.

China has the world's largest number of generative AI patent applications

At the national level, China ranks first in the world in terms of the number of Gen AI-related patent applications, with 38,210, accounting for about 70% of the global share and six times that of the United States (6,276), which ranks second. It was followed by South Korea (4,155), Japan (3,409), India (1,350), the United Kingdom (714), and Germany (708).

And China dominates the number of patent family filings in all five GenAI models. China's leading position in diffusion models is particularly evident, with 500 patent family applications published in this area since 2014, more than 14 times that of the second-ranked United States (35). China's global share of patents in the field of autoregressive models is also very high.

Xinzhixun, China excludes 70%! Global GenAI patent application ranking: Tencent ranks first, Huawei ranks 13th!

However, the United States is also relatively strong in the field of GAN model, VAE model and LLM model. A large percentage of GenAI patents in South Korea fall into the GAN model category. For Japan, it's worth noting that the vast majority of its GenAI patent families can't be associated with any of the five core GenAI models. The Indian GAN patent family accounts for a relatively high proportion of its total GenAI patents. Many patent families in the UK fall under the GAN model and VAE model categories. Germany, on the other hand, prefers the GAN model and the VAE model. It should be noted that the number of GenAI patent applications in India is increasing at an average annual rate of 56%, which is the fastest growth rate among the top 5 countries.

Chinese companies ranked first: Tencent Holdings, Ping An Insurance, and Baidu took the top three, and Huawei ranked thirteenth

When global companies are ranked in terms of the number of patent applications involved in the five key GenAI models identified, companies from China are in the lead. The top 20 companies are Tencent Holdings (2,074), Ping An Insurance (1,564), Baidu (1,234), IBM (601), Alibaba Group (571), Samsung Electronics (468), Google's parent company Alphabet (443), ByteDance (418), Microsoft (377), BBK Electronics (377), NetEase (337), NTT (330), Huawei (328), China Mobile (300), State Grid of China (291), Adobe (257), Sony Group (218), Siemens (208), Ant Group (202), Industrial and Commercial Bank of China (191).

Xinzhixun, China excludes 70%! Global GenAI patent application ranking: Tencent ranks first, Huawei ranks 13th!

It can be seen that among the top 20 companies in the world, 12 are from China, 4 are from the United States, 2 are from Japan, 1 is from South Korea, and 1 is from Germany.

Looking at the number of patent applications involved in the breakdown of the five key GenAI models, Tencent ranks first in terms of decoder-based LLM patent families, followed by Baidu. Baidu and Tencent are also leaders in diffusion models among global companies. Ping An Insurance Group has a comprehensive GenAI patent portfolio, with many patent families across all five GenAI models. China's State Grid is leading the way in the GAN patent family.

Google's parent company, Alphabet, has recently significantly increased the number of its LLM-related patent families and ranks second in the world for VAE models. IBM is the leader in the VAE family of patents, ranking second in terms of GAN models, behind Baidu.

While most companies have patents in one of the major GenAI models, there are some exceptions, such as Tencent and Alphabet, which have filed a large number of patent applications across multiple model types.

The number of patent applications filed by China's scientific research institutions is the first: the Chinese Academy of Sciences is the first, and Tsinghua University is the second

In terms of the ranking of global research institutions in terms of the number of patent applications involved in the above five key GenAI models, it is also China's research institutions that rank high. The top 10 research institutions were Chinese Academy of Sciences (607), Tsinghua University (321), Zhejiang University (320), Zhejiang University of Technology (190), National Science and Technology Research Commission of Korea (89), Nanjing University of Aeronautics and Astronautics (80), Dalian University of Technology (75), China University of Petroleum (68), China University of Mining and Technology (35), and University of California (34). It can be seen that among the top 10 scientific research institutions, 8 are from China, 1 is from South Korea, and 1 is from the United States.

Xinzhixun, China excludes 70%! Global GenAI patent application ranking: Tencent ranks first, Huawei ranks 13th!

From the perspective of specific models, many scientific research institutions mainly apply for patents for one model, of which the GAN model is the most popular type. The Chinese Academy of Sciences is leading the way in the number of applications for GAN model, VAE model and autoregressive model. Zhejiang University leads the way in the number of patent applications in diffusion models; Tsinghua University has the largest number of LLM patent family applications.

Editor: Xinzhixun-Rogue Sword

This article is from Xinzhi self-media and does not represent the views and positions of Business Xinzhi.If there is any suspicion of infringement, please contact the administrator of the Business News Platform.Contact: system@shangyexinzhi.com