AI Supremacy - A quantitative patenting landscape analysis and econometric study investigating the leadership of the 21st century’s most powerful technology

Sammanfattning

In the last decade, China has become the global leader in the quantity of Artificial Intelligence (AI) patents, overtaking the US. While there at first glance appears to have been a catch-up effect in terms of quality, it is unclear as to whether these patents are internationally valuable. The purpose of this study is to compare the two world’s two AI superpowers, China and the US, and answer the question of how the quality of their AI patents differ in terms of international recognition, and how this is impacted by the ownership of the assignee, contrasting public with private actors. Through an analysis of 915,891 AI patents gathered from Google Patents Public Datasets(2000–2023), we perform multiple ordinary least squares (OLS) regressions with interaction terms to isolate the effects of assignee ownership, contrasting results between total forward citations and the stricter quality measure of foreign citations. Our results show that Chinese patents exhibit a pronounced relative disadvantage when evaluated through foreign citations rather than total forward citations. While no statistical difference from baseline could be detected between Chinese universities and the baseline of US corporations using total forward citations (β = 0.01, not significant), it becomes evident that Chinese universities are responsible for the lowest quality when evaluated strictly through foreign citations (β = -0.59, p < 0.001), corresponding to 45% fewer citations relative to the baseline. While Chinese firms exhibit a smaller negative effect (β = -0.29, p < 0.001) with only 25% fewer citations, this is contrasted by all other economies showcasing the opposite pattern, with results ranging from 27% to 35% higher foreign citations than the baseline (βKR = 0.24, p < 0.001; βEPO = 0.30, p < 0.001). Our results indicate that there is indeed a citation inflation in China, also within AI, rendering traditional patent citation metrics obsolete. Additionally, we can confirm that the suggested patent bubble driven by institutional incentives applies deeply within the AI sector. Finally, despite what superficial measures indicate, our findings show that the US is still clearly leading the AI race in terms of genuine technological influence, highlighting the necessity for future research to account for domestic policy distortions when evaluating global capabilities.