Anaqua: AI Semiconductor Patent Filings Surge, Emerging as Key Force Reshaping Semiconductor R&D Investment
TECHWORLD ·
✦ AI Summary
Anaqua said in the 2026 Semiconductor Industry Patent Report that total semiconductor-related patent filings rose 78% over the past 5 years, while patents in AI-semiconductor convergence areas increased 114%.
The report said AI is affecting nearly every area of chip innovation and has become a core force shaping semiconductor R&D and patent filing directions.
The report also presented patent trends and rankings by area for IBM, Qualcomm, Samsung, China government-affiliated institutions, Intel, Alphabet and Google, and Microsoft.
Anaqua released the key findings of the 2026 Semiconductor Industry Patent Report: AI & Innovation Trends. Based on an analysis of global semiconductor patent activity, the report found that total semiconductor-related patent filings rose 78% over the past 5 years, while patents in AI-semiconductor convergence areas increased 114%.
Anaqua said its comparison of semiconductor patent trends showed that patent growth in AI-semiconductor convergence areas was about 40% faster than growth in total semiconductor-related patents. The findings suggest that filings in technology areas combined with AI have been rising faster than the overall average.
Anaqua said the analysis confirms a structural shift in the semiconductor market driven by AI demand. The report concluded that AI has moved beyond merely influencing semiconductor innovation and has become a core factor shaping the direction of innovation across the technology stack.
Changes are being seen across the semiconductor sector, including chip architectures, graphics processing units (GPU), inference processors, and custom accelerators for hyperscalers.
Tony Nam, CPO at Anaqua, said semiconductors play a foundational role in the technology industry and that advances in semiconductor technology have opened up new areas of innovation.
He said the company sought to measure how much AI drives the development of foundational technologies, and explained that AI goes beyond influencing semiconductor innovation and is defining innovation itself. He added that the scale of its impact is very large.
The report backed up that view. Using the patent analytics platform AcclaimIP, it analyzed more than 700,000 semiconductor patent filings and found that AI is affecting nearly every area of chip innovation.
The report said AI-semiconductor convergence patents rose 114%, far outpacing the 78% increase in total semiconductor-related patents. It also said a review of major AI-related semiconductor subareas showed AI's influence on the chip market was far greater than that of other technologies.
The report said this phenomenon is not limited to a few niche areas. It also noted that AI's influence does not remain confined to specific subfields.
It also described AI as a core force behind decisions on semiconductor research and development (R&D) investment directions. At the same time, it said AI is the core force behind decisions on semiconductor patent filing directions.
Nvidia holds 49 semiconductor patents with GPU in the name, ranking 7th in semiconductor patents with GPU in the name and 15th in inference chips. However, the analysis added that the source of Nvidia's sustained competitive advantage is not hardware itself, but the CUDA (Compute Unified Device Architecture) software ecosystem introduced in 2006.
IBM ranked 1st with 794 patent filings in the AI-semiconductor intersection and led the analog AI category with 389 filings. IBM's patent portfolio is focused on quantum computing, neuromorphic chips, and phase-change memory.
Among these, phase-change memory is closely tied to technology that stores neural network weights directly on chips. The technology has shown about a 14-fold improvement in energy efficiency.
Phase-change memory technology has entered the commercialization stage through IBM's Spyre accelerator. IBM is building a patent portfolio centered on advanced areas including this technology.
By contrast, Qualcomm's traditional core areas were on-device AI and edge AI, but it has recently moved to rapidly expand into the data center market. Qualcomm recorded 655 patent filings in inference, ranking 9th, showing a steep rise in patent activity compared with zero registered patents in inference 5 years ago.
The report noted that patent filings in AI architecture increased 186% to 252, the fastest growth in the report. It also said the GPU category ranked 4th with 134 filings.
It then presented Qualcomm's analog AI-related technology spread, ranging from neuromorphic mobile circuits to AI200 and AI250 inference accelerators for data centers. It also identified power efficiency and memory capacity as differentiating factors for AI200 and AI250.
Samsung ranked 1st in AI architecture patent filings with 1,194 filings. It also led the HBM category with 107 filings and was ranked 2nd in inference and 2nd in analog AI.
Samsung's technology range was said to include neural processing units (NPU), processing-in-memory (PIM), and the Exynos system-on-chip (SoC) product line. The report described this as a representative example of a memory maker building end-to-end AI chip competitiveness based on its existing strengths.
Over the past 5 years, patent filings for inference and accelerators by China government-affiliated institutions totaled 5,387. During the same period, Samsung filed 1,897 inference and accelerator patents, indicating that China government-affiliated institutions had an advantage over Samsung in the inference and accelerator segment.
The holders of related patents were found to be distributed across universities, national laboratories, and state-backed research institutes. These institutions ranked 3rd in GPU patent filings.
In China's GPU sector, emerging startups such as Moore Threads and Muxi are also on the rise.
The report analyzed this as the result of China's semiconductor self-reliance strategy in response to U.S. export controls.
For comparison, Intel ranked 1st in GPU semiconductor patents with 188 filings. Intel also ranked 2nd in AI-IC architecture convergence patents with 967 filings, 3rd in HBM-related patents with 23 filings, 5th in AI-analog IC convergence registered patents with 225 filings, and 8th in inference semiconductor patents with 792 filings.
The report said that patents related to Alphabet and Google TPU, as well as Microsoft Maia accelerators, are appearing across nearly all areas of AI-semiconductor convergence. Based on this, it said cloud giants are not limiting themselves to buying off-the-shelf semiconductors but are directly designing custom chips optimized for their own workloads.
It also emphasized that this movement is tied to the race to secure leadership in AI hardware. It added that the trend shows vertical integration is rapidly advancing in the competition for AI hardware leadership.
Source: TECHWORLD · Lee Gwang-jae
Original: https://www.epnc.co.kr/news/articleView.html?idxno=406424
References
This article was produced with the help of an automated content generation algorithm.
Source: TECHWORLD
View originalThis article was summarized and organized by BizCrush based on the original article from TECHWORLD. For exact quotations and full details, please refer to the original article.