SiC Power AI/ML Patent Landscape 2026
The application of AI and machine learning to SiC power semiconductor design, crystal growth, and device characterisation is an early-stage but rapidly expanding field, with annual filings still rising and the patent corpus still small enough for a well-targeted programme to establish a credible position. China accounts for the large majority of disclosures and the applicant base is highly fragmented, with no single organisation holding a dominant share.
A fragmented, China-centric field with no dominant assignee
The corpus spans 68 patent families. The leading applicant, Rohankar Ketaki Nilesh, holds only 2 patent records, a position shared by eleven other filers, making this one of the most evenly distributed early-stage technology landscapes observable at this scale.
The top five filers account for just 12% of the combined output of the hundred largest filers, confirming that no player has yet established a controlling portfolio. The field sits in a first-mover window where concentration is low and entrenchment is minimal.
| # | Applicant | Patent records | Share |
|---|---|---|---|
| 1 | ROHANKAR KETAKI NILESH | 2 | |
| 2 | CHINA UNIV OF MINING & TECH | 2 | |
| 3 | PLA Strategic Support Force Information Engineering University | 2 | |
| 4 | NANJING UNIV OF AERONAUTICS & ASTRONAUTICS | 2 | |
| 5 | Vishay Siliconix LLC | 2 | |
| 6 | ROHANKAR RUGVED N | 2 | |
| 7 | Ewision Semiconductor Technology (Shanghai) Co Ltd | 2 | |
| 8 | Anhui University | 2 | |
| 9 | ROHANKAR NILESH B | 2 | |
| 10 | Wolfspeed Inc | 2 |
| # | Applicant | Patent records | Share |
|---|---|---|---|
| 11 | Beihang University | 2 | |
| 12 | SICC Co Ltd | 2 | |
| 13 | Tianjin Polytechnic University | 1 | |
| 14 | Chongqing University of Science and Technology | 1 | |
| 15 | DR PUTTAMADAPPA C | 1 | |
| 16 | Wuhan University | 1 | |
| 17 | Suzhou University | 1 | |
| 18 | DR RANJITHA S | 1 | |
| 19 | Shenzhen Jinyu Semiconductor Co Ltd | 1 | |
| 20 | CHONGQING UNIV OF POSTS & TELECOMM | 1 |
The even distribution across academic institutions, semiconductor companies, and independent inventors signals that this technology is being explored in parallel by many actors rather than being driven by a single incumbent roadmap. That creates both freedom to operate and competitive uncertainty.
Filings from 2025 and 2026 are under-counted because of the standard patent-publication lag; the apparent plateau in those years should not be read as a slowdown. Longer-window growth, applicant concentration, and technology-route coverage are therefore more reliable signals than the latest-year bar alone.
Strong recent acceleration dominated by AI-model and data-processing classifications
Annual filing activity and the technology class mix together reveal a field that began gaining serious traction around 2020 and accelerated sharply in 2024–2025, with AI-model computing forming the clear technical core.
Annual filing trend
Filings were negligible before 2020, reached an initial plateau of seven records per year in 2020–2021, dipped to four to five in 2022–2023, then surged to twenty records in 2024 and nineteen in 2025. The 81% recent-window growth rate reflects a genuine acceleration. The 2025–2026 bars are suppressed by publication lag and will grow as disclosures publish.
↗ Hover for values · click a bar to ask EurekaTechnology composition
G06N (computing based on AI models) accounts for the plurality of records, confirming that neural-network and machine-learning methods are the dominant technical route. G06F (electric digital data processing) and H01L (semiconductor devices) follow, while power-conversion (H02M) and crystal-growth (C30B) classes reflect the underlying SiC application domains. The breadth of IPC classes — spanning measurement, image recognition, computational chemistry, and even additive manufacturing — points to wide exploratory activity rather than a focused application niche.
↗ Hover for values · click a bar to ask EurekaHighly cited patent families surfaced by the query
Citation-heavy patent families returned by the query. Use this section as citation context, not as a curated list of the most topic-specific patents.
Dislocation distribution for silicon carbide cryst…
Silicon carbide (SiC) wafers, SiC boules, and related methods are disclosed that provide improved dislocation distributions. SiC boules are provided that demonstrate reduced dislocation densities and improved dislocation uniformity across longer boule lengths. Corresponding SiC wafers include reduced total dislocation density (TDD) values and improved TDD… (excerpt from the patent abstract)
| # | Patent | Citations |
|---|---|---|
| 1 | 一种预测SiC单晶炉内整体温度场的方法及设备 | 50 |
| 2 | 一种基于神经网络的碳化硅MOS器件结温在线测量方法 | 12 |
| 3 | 一种基于神经网络的碳化硅场效应管模型 | 11 |
| 4 | Dislocation distribution for silicon carbide cryst… | 10 |
| 5 | 一种电致发光半导体板材表面缺陷AI检测方法 | 9 |
| 6 | Dislocation distribution for silicon carbide cryst… | 8 |
| 7 | 离子注入碳化硅半导体掺杂过程的模拟仿真方法与装置 | 7 |
| 8 | 一种基于碳化硅的智能功率模块及应用 | 7 |
Ranked by total forward citations. Citation counts favour older and broadly cited patent families, and broad or adjacent patents may appear when they match the search scope. Treat this section as citation context, not as a curated list of the most topic-specific patents. Some patent titles may be shown in their original, non-English language where an accurate translation could not be guaranteed.
What the competitive structure means for R&D investment decisions
China-heavy jurisdiction profile shapes a specific set of opportunities and risks for non-Chinese entrants and academic collaborators alike.
Early Growth — annual filings still rising
The lifecycle assessment places this field firmly in the Growth stage, with annual filings still rising and an 81% increase in the recent measurement window. The corpus of 68 patent families is small enough that a focused filing programme initiated now could reach the top tier within two to three years. The absence of a dominant incumbency position means freedom-to-operate risk is currently lower than in adjacent mature SiC-device classes.
Growth stageExtremely fragmented — top five hold 12% of the top-100 share
The top five filers collectively account for only 12% of the combined output of the hundred largest filers, and the leading applicant holds just 2 patent records. This degree of fragmentation is unusual even for an emerging field and suggests that no player has yet committed the resources necessary to build a defensive portfolio. A sustained filing programme over 24–36 months could realistically capture a material share of the active landscape.
Low concentrationOne confirmed co-filing pair: Nanjing University of Aeronautics & Astronautics with Suzhou Nuclear Power Research Institute
The evidence shows a single documented co-application between Nanjing University of Aeronautics & Astronautics and Suzhou Nuclear Power Research Institute (Suzhou Thermal Engineering Research Institute). This pairing of an aerospace-oriented university with a nuclear-power-sector research organisation points to reliability and condition-monitoring applications of SiC AI/ML as one collaboration axis. The scarcity of co-filings overall suggests that university–industry collaboration in this space is nascent and underexploited.
Nascent collaborationChina accounts for the large majority of filings; US and India are present but minor
China is the lead filing jurisdiction by a substantial margin, followed at a great distance by India, with Israel, the Netherlands, the United States, Germany, South Korea, and Taiwan each contributing a small number of records. This skew means that IP protection strategies outside China are largely unchallenged today, and a parallel filing programme targeting US, European, or Japanese offices would face minimal prior-art density from existing portfolio holders.
China-dominantGo beyond the landscape: Eureka’s TRIZ Solution agent breaks down an R&D problem and returns patented concept solutions, each with a technical approach and cited patent & literature evidence.
| Applicant | Collaborator | Co-filings |
|---|---|---|
| Nanjing University of Aeronautics and Astronautics | Suzhou Nuclear Power Research Institute Co Ltd | 1 |
Co-filing pairs, ranked by the number of jointly-filed patent families.
New entrants dominate the leaderboard; device specialists and AI-focused labs lead by route
Every ranked applicant with tracked momentum is classified as a new entrant, reflecting how recently this application domain crystallised. Technical emphasis divides broadly into AI-model methods applied to device physics versus power-conversion circuit optimisation.
Wolfspeed Inc
Wolfspeed holds 2 patent records in the corpus, focused on crystal-growth methods (C30B) and semiconductor devices (H01L), reflecting its position as the leading merchant SiC substrate and device supplier applying machine-learning methods to crystal-quality and defect characterisation. Its momentum is that of a new entrant in this specific AI/ML application sub-field, consistent with the overall nascency of the corpus.
patent records: 2Ewision Semiconductor Technology (Shanghai)
Ewision Semiconductor Technology (Shanghai), listed in the evidence as, holds 2 patent records with technical emphasis on digital data processing (G06F) and AI-model computing (G06N), complemented by electrical measurement (G01R). Its entry as a new entrant suggests recent strategic commitment to integrating machine-learning diagnostics with SiC device characterisation workflows.
patent records: 2| Applicant | Recent (3 yrs) | Trend |
|---|---|---|
| Vishay Siliconix LLC | 2 | ▲ new entrant |
| Anhui University | 1 | ▲ new entrant |
| Nanjing University of Aeronautics and Astronautics | 2 | ▲ new entrant |
| China University of Mining and Technology | 2 | ▲ new entrant |
| PLA Strategic Support Force Information Engineering University | 2 | ▲ new entrant |
| MR RUGVED N ROHANKAR | 2 | ▲ new entrant |
Under-served IPC branches adjacent to the AI/ML-SiC core
Several IPC classes within the corpus carry low record counts relative to the dominant G06N cluster, representing areas where AI/ML methods touch SiC power technology but have received limited dedicated patent attention to date.
H02M · Power conversion (AC/DC etc.)
H02M covers the inverter, converter, and gate-driver circuits where SiC devices deliver their primary performance advantage. With only 7 patent records in this branch and a 4% share of the top-100 filers’ output, AI/ML methods for real-time converter control, loss optimisation, and predictive gate-drive tuning are sparsely covered. The technical value is high — converter efficiency and reliability are commercial differentiators — and an entry path exists through combining existing SiC circuit expertise with neural-network control literature. This is an under-served adjacent branch with plausible technical value.
Search this in Eureka →G16C · Computational chemistry / materials informatics
G16C (computational chemistry) holds 7 patent records in the corpus, covering AI-assisted simulation of SiC crystal defects, doping profiles, and epitaxial process parameters. Given the critical role of material defect density in SiC device yield, this branch sits at the intersection of process engineering and AI-driven materials informatics — an area where pharmaceutical and materials-science AI methods are increasingly being transferred. The branch is sparse (4% share of top-100 output) but technically adjacent to the most-cited patent cluster, which centres on neural-network prediction of SiC furnace temperature fields. Research groups with computational chemistry capability have a realistic entry path.
Search this in Eureka →Frequently asked questions
The corpus covers 68 patent families in scope. This is a small but rapidly growing body of IP, reflecting the early-stage commercialisation of AI and machine-learning methods applied specifically to SiC power semiconductor design and manufacturing.
No single organisation dominates. Twelve applicants are tied at the top of the ranking with 2 patent records each, including Rohankar Ketaki Nilesh, China University of Mining and Technology, Nanjing University of Aeronautics and Astronautics, Vishay Siliconix LLC, Wolfspeed Inc, Beihang University, and SICC Co Ltd, among others. The top five filers account for only 12% of the combined output of the hundred largest filers.
China is the dominant filing jurisdiction, accounting for 52 of the patent records tracked in this corpus. India is the second jurisdiction by record count, followed at distance by Israel, the Netherlands, the United States, Germany, South Korea, and Taiwan.
The lifecycle assessment places the field in the Growth stage. The recent-window growth rate is 81%, with annual filings rising from near-zero before 2020 to a peak of 20 records in 2024. The 2025 and 2026 figures are under-counted due to publication lag and will increase as applications publish.
G06N (computing based on AI models, primarily neural networks) is the dominant IPC class, appearing in 58 patent records. Applications include predicting temperature fields in SiC crystal-growth furnaces, measuring junction temperature in SiC MOS devices, modelling SiC FET characteristics, and detecting surface defects in electroluminescent semiconductor wafers — all evidenced in the most-cited patents in the corpus.
The evidence shows one confirmed co-application: between Nanjing University of Aeronautics and Astronautics and Suzhou Nuclear Power Research Institute (Suzhou Thermal Engineering Research Institute). Broader co-filing activity is not yet evident in the dataset, indicating that university–industry collaboration in SiC Power AI/ML is nascent.
Built on Patsnap Open Platform
This report’s underlying patent dataset — filings, assignees, technology clusters — is open for developers via MCP and REST API. Free to start, 10,000 credits, no credit card required.
Disclaimer. This page is generated from Patsnap Eureka data drawn from a limited snapshot of global patent and scientific-literature records, and is provided for general information and reference only.
Patent data carries inherent limitations: recent filings (typically the most recent 18–24 months) are under-counted due to standard publication lag; counts may be reported at either a patent-family or a patent-record basis and are not always directly comparable; classification, applicant-name, and citation data may contain errors, duplicates, or omissions; and the underlying search query defines and constrains the scope shown. As a result, the analysis may be incomplete or inaccurate and may not reflect the full technology landscape.
Nothing on this page constitutes an exhaustive prior-art, novelty, freedom-to-operate, or validity search, nor does it constitute legal, financial, investment, or professional advice, and it should not be relied upon as such. Any patent, commercial, or strategic decision should be verified independently and reviewed with qualified patent, legal, and domain professionals. Patsnap makes no warranties, express or implied, as to the accuracy, completeness, or fitness for any particular purpose of the information presented.