Stanford AI Index 2026: China Holds 74 Percent of Global AI Patents While the U.S. Falls to 12 Percent

The AI Index Report 2026, published in April 2026 by Stanford University’s Human-Centered Artificial Intelligence Institute (Stanford HAI), is the most comprehensive annual report on AI research, development, and societal impact. Among its most striking findings for the intellectual property community is the geographic distribution of AI patents: of the 131,121 AI patents granted worldwide in 2024, China accounted for 97,206 (74.2%), while the United States’ share fell to just 12.1%.

A Decade of Reversal

According to Stanford HAI’s 12 Key Takeaways, this landscape has completely reversed from ten years ago. In 2015, the United States held 42.8% of all AI patent grants. China’s share began its rapid expansion around 2017, following the State Council’s release of the “New Generation AI Development Plan,” after which filing volumes surged dramatically.

SiliconAngle reports that China leads not only in AI patents but also in academic publications, citations, and industrial robot installations. The United States, however, maintains advantages in private AI investment, AI chip design and manufacturing infrastructure, and the development of high-performance AI models.

Stanford HAI’s composite “Global Vibrancy Tool” shows the overall score gap separating the U.S. from China has narrowed to just 2.7 points. As The Next Web notes, given that China’s AI investment is approximately one-twenty-third of the U.S. level, this suggests China is achieving remarkably high returns on its AI investment.

Quantity vs. Quality: The Reality Behind the Numbers

The 74.2% figure might suggest that China exercises overwhelming dominance in AI intellectual property. However, multiple analyses caution against taking this number at face value.

First, according to Humai Blog’s detailed analysis, the grant rate for Chinese AI patents (the ratio of grants to applications) is approximately 32%. This means the volume of applications is far larger than the already impressive grant numbers, suggesting that many applications are filtered out during quality screening. Patent offices in the U.S. or Europe maintain higher grant rates, indicating relatively higher qualitative standards in applications.

Second, when focusing on high-impact patents—those with significant commercial influence measured by citation counts, technology transfer records, and commercialization outcomes—the United States retains its lead. American patent portfolios demonstrate qualitative superiority on these metrics.

Third, China’s surge in patent filings is partly driven by government incentive policies. Application subsidies from local governments or universities, evaluation systems using patent counts as KPIs, and other policy mechanisms have fueled quantitative expansion. In recent years, the Chinese government itself has acknowledged this problem, promoting a quality-first approach, but the effects have yet to fully materialize in the data.

Patent Trends by Sector

AI patents are not a monolithic category. Examining the Stanford AI Index data by sector reveals that Chinese strengths differ significantly from American strengths in terms of domain concentration.

China is particularly strong in computer vision, the application layer of natural language processing (NLP), and robotics-related patents. According to IEEE Spectrum, China is rapidly accumulating patents in “physical AI,” encompassing industrial robots and autonomous systems, reinforcing its position as the world’s largest adopter of industrial robots.

The United States, meanwhile, leads in foundation models and generative AI patents. Companies developing cutting-edge AI models, including OpenAI, Google DeepMind, Anthropic, and Meta, are predominantly U.S.-based. Fundamental patents covering model architectures and training methodologies are held by American firms.

Japan’s Position

Amid the U.S.-China bipolarization in AI patents, Japan’s positioning merits attention. According to the Stanford AI Index 2026, Japan maintains its third-place ranking in AI patent applications and grants, though its share remains in the single digits, far behind the two leaders.

However, Japanese companies possess distinctive strengths. Toyota has built a globally significant portfolio of AI patents related to autonomous driving. Panasonic and Sony have accumulated patents in edge AI (on-device AI processing) and image sensors. Canon maintains a strong patent position in image processing AI.

South Korea also warrants attention. As Crypto News reports, South Korea leads the world in AI patent filings per capita (innovation density), driven by active patent prosecution from semiconductor and electronics giants including Samsung Electronics, LG, and SK Hynix.

Structural Factors in the U.S.-China AI IP Competition

Understanding the U.S.-China gap in AI patents requires grasping the structural differences in each country’s innovation ecosystem.

The U.S. AI research model predominantly involves innovations originating from universities and startups that are scaled up through private investment. According to Fortune, U.S. private AI investment in 2025 was estimated at approximately $120 billion, vastly exceeding China’s approximately $9.6 billion. This investment tends to flow toward product development and market deployment rather than patent filing.

China’s AI research operates on a model where state-led top-down planning runs parallel with large-scale applied research by private companies. Patent applications are valued as quantitative indicators of research output, with universities and research institutions accounting for a significant share of total filings. While tech giants such as Baidu, Huawei, and Tencent are active filers, the high proportion of academic filings is a distinctive Chinese characteristic.

These structural differences carry important implications for interpreting patent data. U.S. companies tend to protect and deploy AI technologies through trade secrets or open-source strategies rather than patents, meaning that patent filing counts alone may underestimate American AI innovation capacity.

Rethinking What “Patent Dominance” Means

China’s 74% share of AI patents is undeniably a striking figure. However, patents are not an end in themselves but a means of realizing the fruits of innovation.

AIBase’s analysis points out that quantitative patent dominance does not necessarily equate to technological superiority or market power. What matters is whether patents are generating licensing revenue, being implemented in products, and effectively blocking competitor entry—in other words, their practical effectiveness.

From this perspective, the current AI patent competition can be summarized as China dominating in quantity while the U.S. competes on quality, but there is no guarantee this dynamic will persist. If the Chinese government’s quality-first transition succeeds, China could close the qualitative gap as well.

For Japanese companies, the lesson is the importance of concentrating patent resources in areas of core competence—such as autonomous driving, edge AI, image processing, and semiconductor-related technologies—to build qualitatively superior portfolios. Rather than chasing global ranking positions in patent volume, formulating IP strategies aligned with business strategy is what this era of U.S.-China bipolarization demands.

The data presented in the Stanford AI Index 2026 confirms that the geopolitical reorganization of AI intellectual property is accelerating. Reading the strategic intent behind each country’s patent data and accurately assessing one’s own positioning will be indispensable for IP management going forward.