Memristive Synaptic Crossbar Array Patents: Leaders & Filing Trends 2026
A patent landscape review of memristive synaptic crossbar array technology: who leads the 359-record field, how filings have moved since the 2019 peak, and where IPC composition points to open ground.
Filing growth = 2021 (49 records) → 2024 (15); 2024 is the last year we treat as complete. Top-5 share = the 5 largest assignees ÷ all 359 records in scope (CR5), not the ranked leaders only.
What this landscape covers
Memristive synaptic crossbar arrays sit at the intersection of resistive-memory device physics and neural-network hardware architecture: a crossbar of memristor or RRAM cells performing vector-matrix multiplication in place, meant to stand in for the multiply-accumulate step that dominates neural-network compute. The 359 records in scope span filings from 2015 through the current data cut-off, drawn from both device-level claims on resistive switching elements and system-level claims on how those elements are wired and trained as synapse arrays. Publication lags filing by roughly 18 months, so the most recent one to two years in any trend understate actual filing activity.
The receiving-office spread is heavily US-weighted, with the United States, Europe and the WIPO PCT route together accounting for the bulk of filings, and India, the United Kingdom and South Korea forming a smaller but active second tier. That pattern points to a technology still being filed defensively and broadly rather than narrowed to a handful of jurisdictions.
Let an AI agent run this analysis on your own technology
Pick a task. Every answer cites the patents behind it.
Filing trend and technology composition
Two views of the same 359-record set: how filing volume has moved year over year, and which IPC subclasses the claims actually sit in.
A sharp pullback from a 2019 peak
Publications rose from 12 in 2017 to a peak of 61 in 2019, then fell substantially; the most recent complete comparison shows 49 records in 2021 declining to 15 in 2024, a 69% drop over that three-year span. Because publication lags filing by around 18 months, 2025 and 2026 figures are still filling in and should not be read as a continued decline.
Publication lags filing by roughly 18 months, so 2025 onwards are still filling in. Growth rates on this page therefore end at 2024; running them to the last bar would understate the field.
Compute classes outweigh memory classes
G06N (AI-model computing) appears in 82.5% of the 359 records, far ahead of G11C (static and digital memories) at 32.6% and G06F (electric digital data processing) at 25.1%. Semiconductor-device classes H01L, H10N and H10B each cover a smaller slice, and analogue-computing class G06G appears in just 2.5% of records — since a record can carry multiple classes, these shares add up to more than 100%.
Shares are the percentage of the 359 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Go deeper on Neuromorphic Computing: Memristive Synaptic Crossbar Array Patent Landscape with Eureka
This page is one run against one query. Ask Eureka your own question about neuromorphic computing: memristive synaptic crossbar array patent landscape and every answer comes back with the patent numbers behind it.
Try EurekaThe most-cited prior art in this space
Performing dot product operations using a memristive crossbar array
A method, computer system, and computer program product of performing a matrix convolution on a multidimensional input matrix for obtaining a multidimensional output matrix. The matrix convolution may include a set of dot product operations for obtaining all elements of the output matrix. Each dot product operation of the set of dot product operations may include an input submatrix of the input matrix and at least one convolution matrix. The method may include providing a memristive crossbar array configured to perform a vector matrix multiplication. A subset of the set of dot product operations may be computed by storing the convolution matrices of the subset of dot product operations in the array.Filed by IBM, published 2021-03-11.
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US20140172937A1 | Apparatus for performing matrix vector multiplication approximation using crossbar arrays of resistive memory… | 183 |
| 2 | US20180364785A1 | Memristor crossbar arrays to activate processors | 163 |
| 3 | US20110004579A1 | Neuromorphic Circuit | 154 |
| 4 | US20150170025A1 | Method and apparatus for performing close-loop programming of resistive memory devices in crossbar array base… | 127 |
| 5 | US20170017879A1 | Memristive neuromorphic circuit and method for training the memristive neuromorphic circuit | 106 |
| 6 | US20170200078A1 | Convolutional neural network | 91 |
| 7 | US20200342301A1 | Convolutional neural network on-chip learning system based on non-volatile memory | 83 |
| 8 | US20170083810A1 | Electronic Neuromorphic System, Synaptic Circuit With Resistive Switching Memory And Method Of Performing Spi… | 82 |
| 9 | US20190035154A1 | Sensor system based on stacked sensor layers | 81 |
| 10 | US20170330070A1 | Spin orbit torque based electronic neuron | 81 |
Citation counts inside a searched corpus favour older records; treat them as a signal of influence on subsequent filings, not of current commercial relevance.
Each row carries its publication number; clicking a row searches Eureka by that number.
Put your own technology through the same analysis
Eureka on the web
When you want the answer in the next five minutes.
The agent works the prompt against patents and technical literature, citing every source.
Run your analysis now →MCP server & REST API
When it has to run inside your own pipeline.
Patent search, landscape analysis and assignee resolution as MCP tools. Drop them into any agent framework, or call REST directly.
Browse MCP servers →What the numbers mean for strategy
Three findings a claims search alone would not surface.
The top of the field is dense, the rest is a long tail
Five assignees out of 91 ranked hold 44.0% of all 359 records, and the top 10 hold 58.5%. That leaves a wide tail of single- or few-filing entrants working on narrower device or integration claims — exactly where a freedom-to-operate search needs the most care, since these filers are easy to miss in a leader-focused review.
A cooling filing curve after a 2019 peak
Publications peaked at 61 in 2019 and fell to 15 by 2024, a 69% decline from the 2021 level of 49. That pattern is consistent with a technology whose foundational claim space filled quickly and is now seeing fewer new entrants rather than sustained growth in filing volume.
Claims skew toward compute architecture, not device physics
G06N (AI-model computing) appears in 82.5% of the 359 records, well above memory-class G11C at 32.6% and semiconductor-device class H01L at 9.2%. Most filers are claiming the neural-network application layer built on top of the crossbar rather than the underlying resistive-switching material itself.
Corporate-academic pairing is limited but visible
Only 10 co-assignee pairs appear across the dataset, the strongest being an intra-corporate IBM pairing and a Samsung-university collaboration. This is a field where most records are filed by a single assignee rather than jointly, so co-filing signals are a minor part of the competitive picture.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to neuromorphic computing: memristive synaptic crossbar array patent landscape, with the prior art for and against each one.
Where to take this analysis
The dataset points to a few concrete next steps depending on what you are trying to decide.
Map the long tail behind the leaders
44.0% concentration in the top 5 still leaves most of the 359 records spread across 86 other assignees. A targeted review of that tail surfaces niche device or fabrication claims that a leader-only search would miss.
Explore the assignee landscape in EurekaTrack the compute-versus-device split
With G06N claims running well ahead of G11C and H01L, the open question is whether device-physics improvements to the memristive cell itself are being under-claimed relative to system-level architecture.
Run a class-level gap search in EurekaWatch for the filing lag to resolve
2025-2026 volumes will keep revising upward as publication catches up with filing. Re-running this trend in six to twelve months will show whether the 2021-2024 decline is a genuine slowdown or a temporary trough.
Set up a trend alert in EurekaCommon questions on this landscape
The assignee ranking covers 91 companies across the 359 records in scope, with the leading assignee holding 64 records outright. The field is concentrated at the top: the five largest filers together hold 44.0% of all records, and the top ten hold 58.5%. Beyond that group there is a long tail of assignees with only a handful of filings each, so a competitive review needs to look past the leaders to catch narrower device-level claims.
Filing peaked at 61 records in 2019 and has fallen substantially since, with the 2021 to 2024 window showing a 69% drop from 49 records to 15. That said, publication typically lags actual filing by around 18 months, so the 2025-2026 figures in any trend chart are still incomplete and should not yet be read as confirmation of a continued decline. The honest read is that the initial claim-filing wave has passed its peak, not that new activity has stopped.
The dominant IPC class is G06N, covering AI-model computing, which appears in 82.5% of the 359 records — meaning most filings claim how the crossbar is used in a neural-network system rather than the resistive-memory device itself. Memory-focused class G11C appears in 32.6% of records and general digital-processing class G06F in 25.1%. Semiconductor-device and memory-manufacture classes (H01L, H10N, H10B) each cover a smaller slice, suggesting device-fabrication claims are comparatively less crowded than system-architecture claims.
US20210073317A1, filed by IBM and published in March 2021, claims a method and system for performing matrix convolution using a memristive crossbar array configured for vector-matrix multiplication, including how dot-product subsets are computed and stored across convolution matrices. It sits inside the G06N-heavy portion of this landscape that ties crossbar hardware directly to neural-network compute operations. Anyone building a convolution-capable crossbar accelerator should read its claims closely before assuming a design-around is straightforward.
The United States leads by a wide margin with 212 filings, followed by the European Patent Office at 43 and the WIPO PCT route at 34. India, the United Kingdom and South Korea form a smaller second tier at 25, 14 and 8 respectively. The heavy US and PCT presence indicates filers are seeking broad, defensible protection rather than concentrating on a single regional market.
Research Neuromorphic Computing: Memristive Synaptic Crossbar Array Patent Landscape in depth with Eureka
Go past this page: query the whole neuromorphic computing: memristive synaptic crossbar array patent landscape corpus yourself, in your own scope.
Every answer comes back with patent numbers you can open.
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.
Machine translation. Assignee and organisation names originally recorded in Chinese, Japanese or Korean have been rendered into English by an AI translation step so that the tables stay readable. These renderings are best-effort and may not match a company’s registered English name; the original name is what the underlying patent record carries, and it is what any Eureka query launched from this page uses.