AI patent publications totalled 378,118 in the 12 months to May 2026, up from 368,191 in the prior year, a 2.7% increase. Grant ratio (the share of applications that convert to granted patents) rose from 0.465 to 0.505. Volume is growing and quality is improving, but the companies and segments building within that expansion are not yet visible in any public disclosure.
A growing market filing better quality IP
AI patent activity peaked at 110,941 publications in Q3 2025 and has since moderated, settling at 56,585 in Q2 2026. The TTM grant ratio of 0.505, up from 0.465 in the prior year, signals a field where the quality of what is being filed is improving faster than the volume. Microsoft Corp (NASDAQ: MSFT), accelerating at 8.3% H2 vs H1 to 1,360 publications, and Huawei Investment & Holding Co Ltd (unlisted), up 7.1% to 1,877, are the two named commercial operators building IP quality and volume simultaneously in a moderated market.
Figure 1: AI Patent Publications and Grant Ratio — Quarterly, Q1 2023 to Q2 2026
Source: GlobalData Patent Analytics
AI platforms up. MOS microprocessors and ML down
Sector-level data across the 12 months to May 2026 shows a cross-sectional range that the aggregate growth figure completely conceals. Machine Learning, the largest sector by volume at 107,700 TTM publications, fell 24.1% from H1 to H2. MOS Microprocessors fell 38.4%, Image Management fell 29.0%, and IT Services fell 31.6%.
Against that broad contraction, Artificial Intelligence Platforms grew 18.6%, from 26,135 in H1 to 30,985 in H2, the only segment with meaningful positive momentum across the tracked universe. For an investor building a view on which AI technology layers are attracting durable investment commitment rather than legacy filing activity, this divergence is the signal. The companies concentrating IP activity in AI Platforms are making a specific architectural bet on software-defined AI infrastructure. CAD, CAM and CAE also grew modestly at 3.0%.
Figure 2: AI Patent Publications by Sector — H1 vs H2, TTM Jun 2025 to May 2026
Source: GlobalData Patent Analytics
Who’s accelerating, declining, and retreating
Assignee data shows a dispersion that the sector-level numbers do not. Microsoft Corp (NASDAQ: MSFT) grew H2 publications 8.3% to 1,360. Huawei Investment & Holding Co Ltd (unlisted) grew 7.1% to 1,877. International Business Machines Corp (NYSE: IBM) was broadly flat at -4.7%. Samsung Electronics Co Ltd (KRX: 005930) fell 4.3% and Alphabet Inc (NASDAQ: GOOGL) fell 8.8%.
The largest single-name decline in the top ten is Tencent Holdings Ltd (HKEx: 700), which fell 40.6% from H1 to H2, from 2,231 to 1,325 publications. State Grid Corporation of China (unlisted, state-owned enterprise) fell 5.2%. For an investor with exposure to Chinese internet or technology names, Tencent’s retreat is not a data artefact. It reflects a reallocation of IP investment activity that will take months to appear in any public strategic communication.
Zhejiang University (unlisted) and Nanjing University (unlisted) fell 34.0% and 35.4% respectively. The top ten individual inventors by publication volume are all Chinese and collectively average a grant ratio below 0.02, confirming that the dominant source of Chinese AI patent volume is academic filing programmes calibrated to publication targets rather than commercial protection. The commercial-academic split in AI is more extreme than in any other sector in this dataset.
Figure 3: AI Patent Publications by Assignee — H1 vs H2, TTM Jun 2025 to May 2026
Source: GlobalData Patent Analytics
Volume and quality are different rankings
China leads by volume at 264,843 publications, 70.0% of the TTM total, at a grant ratio of 0.414. South Korea converts at 1.042: 10.4 granted patents per 10 applications. Taiwan converts at 1.098. The United States accounts for 71,630 publications at a grant ratio of 0.787. Japan converts at 0.755.
For an investor sizing exposure to AI across geographies, China’s 70% volume share overstates its commercial IP position by a factor of roughly two relative to its grant ratio. Taiwan and South Korea are building IP estates with conversion rates above 1.0, meaning prior-period filings are generating grants faster than new applications are arriving. That is a direct signal of commercial IP maturity: the defensible positions have already been staked. WIPO carries 12,212 publications pending national phase entry, representing a pipeline of international filings whose eventual conversion will determine their commercial value.
Figure 4: AI Patent Publications by Patent Authority with Grant Ratio — TTM Jun 2025 to May 2026
Source: GlobalData Patent Analytics
Image analysis and vision are settled
Technology classification data shows where the AI IP race is resolving and where it remains open. Image analysis techniques convert at 0.774, image processing at 0.745, and image analysis and machine vision at 0.667: all well above the sector average of 0.505. These are the technology categories where granted protection is accumulating and late entrants face meaningful barriers. The computer vision IP race is largely settled.
Neural network architectures, the largest classification at 174,689 publications, convert at 0.557, above average. Machine learning methods convert at 0.587. Both are in the building phase: above-average quality and still growing. Expert systems convert at 0.327 and business process AI at 0.374, both below the sector average, indicating active competition with no single filer or geography yet dominant. For an investor evaluating AI platform technology exposure, the distinction between settled IP such as vision and contested IP such as expert systems is a more precise input than revenue forecasts.
Figure 5: AI Patent Publications by CPC Classification with Grant Ratio — TTM Jun 2025 to May 2026
Source: GlobalData Patent Analytics
The evidence that this leads to outperformance
AI Platforms is the one segment building IP momentum while the broader technology stack cools. The interval between these patent signals and their appearance in product launches or earnings guidance is where position sizing decisions get made. Companies identified as innovation leaders through patent indicators show 1.4 times higher revenue growth than the broader market. The full evidence for how patent indicators convert into systematic outperformance across sectors and cycles is set out in GlobalData’s ‘Extracting Innovation Alpha Using Patents’ report, which is free to download below.
