Heterogeneous Computing Patents: Top Companies & Filing Trends 2026
- 40.8% concentration at the top. The five most active filers account for 1,604 of 3,930 records in scope, but the ranked field runs 100 companies deep, so a long tail still files against the leaders.
- Filings grew 73% in three years. From 249 records in 2021 to 431 in 2024 — the last year the dataset treats as complete, since publication lags filing by roughly 18 months.
- G06F carries three-quarters of the field. 2,979 of 3,930 records sit in electric digital data processing, while AI-model computing and interconnect/transmission classes trail well behind at 13-14% each.
Filing growth compares 2021 (249 records) with 2024 (431) — a three-year span. 2024 is the most recent year we treat as complete: publication lags filing by roughly 18 months, so 2025 onwards are still filling in and any growth rate that ends there would understate the field. Top-5 share is the combined record count of the five largest assignees divided by all 3,930 records in scope (CR5), not by the ranked leaders only.
What this landscape covers
Heterogeneous computing spans systems that combine dissimilar processing elements — CPUs, GPUs, accelerators — under a shared memory or interconnect model. This landscape draws on 3,930 published records filed or published between 2015 and the 2026 data cut-off, searched against terms spanning cache coherence, compute kernels, memory access, task parallelism, interconnect fabric and task scheduling. The scope favours records that couple the architectural claim (heterogeneous computing) to a specific mechanism, rather than every general-purpose processor filing.
Because the same invention can generate multiple family members across jurisdictions, the assignee ranking below counts patent families rather than raw document counts, which keeps aggressive continuation filers from crowding out single-filing entrants in the read.
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 3,930-record dataset: activity over time, and where that activity sits across the IPC classification.
Filing activity, 2017-2026
Filings ran from 161 records in 2017 to a peak of 545 in 2025, with 2021-2024 alone showing 73% growth (249 to 431 records). Treat 2025 and 2026 as still filling in — publication lag means the most recent years always understate true filing volume.
Where the claims sit
G06F (electric digital data processing) covers 75.8% of the 3,930 records, with G06N (AI-model computing) and H04L (digital transmission) each near 13-14%. Because records can carry several IPC classes, these shares add up to more than 100% and should be read against the full record total, not against each other.
Shares are the percentage of the 3,930 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Go deeper on Heterogeneous Computing Patent Landscape with Eureka
This page is one run against one query. Ask Eureka your own question about heterogeneous computing patent landscape and every answer comes back with the patent numbers behind it.
Try EurekaMost-cited records in the corpus
Heterogeneous computing terminal for task scheduling
A heterogeneous computing terminal assigns a task scheduling strategy across multiple computing units, each of which can acquire and apply a target upgrade program before processing its share of a task and returning results to a processing unit that recombines them into a completed output.Filed by Shanghai Makesens Energy Storage Technology; published 2024-05-02 as US20240143394A1.
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US6105122A | I/O protocol for highly configurable multi-node processing system | 424 |
| 2 | US8032586B2 | Method and system for caching message fragments using an expansion attribute in a fragment link tag | 400 |
| 3 | US20190220703A1 | Technologies for distributing iterative computations in heterogeneous computing environments | 370 |
| 4 | US20150134824A1 | Constructing virtual motherboards and virtual storage devices | 341 |
| 5 | US20030187935A1 | Method and system for fragment linking and fragment caching | 333 |
| 6 | US20030191812A1 | Method and system for caching role-specific fragments | 290 |
| 7 | US20070244964A1 | Method and system for caching message fragments using an expansion attribute in a fragment link tag | 272 |
| 8 | US6148349A | Dynamic and consistent naming of fabric attached storage by a file system on a compute node storing informati… | 265 |
| 9 | US6836839B2 | Adaptive integrated circuitry with heterogeneous and reconfigurable matrices of diverse and adaptive computat… | 253 |
| 10 | US6078990A | Volume set configuration using a single operational view | 253 |
Citation counts inside a searched corpus favour older filings that have had more time to accumulate references — read them as a signal of influence on later filings, not as a measure 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 data implies for R&D and IP strategy
Three read-throughs from the filing pattern, citation pattern and technology split above.
The top of the field is dense but not closed
Five assignees hold 1,604 of the 3,930 records in scope, yet the ranked list runs 100 companies deep with a fifth-place holder at 146 records and a tenth-place holder at 100 — well below the leader's 551. That gap between rank one and rank five signals room to file adjacent to the leaders without immediately colliding with their core claims.
Growth is real but recent-year figures understate it further
The 2021-to-2024 span shows genuine acceleration in filing volume. 2025's apparent peak and 2026's partial count should not be read as deceleration; they reflect the roughly 18-month gap between filing and publication rather than a slowdown in inventive activity.
Core architecture claims dominate; AI and transmission trail
G06N (AI-model computing) and H04L (digital transmission) each sit near 13-14% of records, well behind G06F. Image processing (G06T), video recognition (G06V) and bioinformatics (G16B) each hold single-digit shares, suggesting these are still secondary claim surfaces layered onto core heterogeneous-architecture filings rather than standalone filing centres.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to heterogeneous computing patent landscape, with the prior art for and against each one.
| Assignee | Co-assignee | Shared families |
|---|---|---|
| Advanced Micro Devices, Inc. | ATI Technologies ULC | 15 |
| International Business Machines Corporation | IBM UNITED KINGDOM LTD INTELLECTUAL PROPERTY DEPARTMENT | 5 |
| Intel Corp | WEISSMANN ELIEZER | 4 |
| Intel Corp | SUBBAREDDY DHEERAJ R | 4 |
| Intel Corp | SRINIVASA GANAPATI N | 4 |
| Intel Corp | NAVEH ALON | 4 |
| Intel Corp | NARVAEZ PAOLO | 4 |
| Intel Corp | NAIK MISHALI | 4 |
Ten co-assignee pairs appear in the dataset; the strongest link ties two units of the same corporate group together, and joint filings otherwise remain rare relative to the 3,930-record total.
Who is filing, and how their pace is changing
The ranking spans 100 companies. The leader's recent-year momentum has fallen sharply, which is common when a filer front-loaded a platform's core architecture claims years ago and is now filing selectively.
The leader's pace has slowed from its own base
The top-ranked assignee's latest-year count is down sharply year-on-year, consistent with a filer that built out its core claim position early and is now filing more selectively rather than exiting the space.
A visible drop-off after the top few
Fifth place sits at 146 records against the leader's 551 — a wide gap that leaves room for a mid-field filer to build a denser position in a specific mechanism (scheduling, coherence, interconnect) without directly overlapping the leader's broadest claims.
A long tail files steadily below the top ten
Tenth place holds 100 records, and the ranking extends to 100 companies total — this is not a top-50 or top-100 cut, it is the full ranked field the dataset returns, and it shows sustained single- and low-digit filing activity well outside the leaders.
| Assignee | Recent year | YoY |
|---|---|---|
| NVIDIA Corp | 17 | -78% |
| Huawei Technologies Co., Ltd. | 2 | -75% |
| Intel Corp | 1 | -96% |
| Qualcomm Inc | 1 | -80% |
| Advanced Micro Devices, Inc. | 1 | 0% |
| Dell Products L.P. | 0 | -100% |
| International Business Machines Corporation | 0 | — |
| Samsung Electronics Co., Ltd. | 0 | -100% |
Where to take this analysis
The dataset points to specific follow-up work rather than a single answer.
Map the mid-field's specific claim mechanisms
The gap between the leader's 551 records and fifth place's 146 suggests distinct filing strategies rather than a single race. Pulling claim language from the mid-field ranks would show whether they cluster around scheduling, coherence, or interconnect specifically.
Explore assignee claims in EurekaTrack the 2025-2026 filings as they resolve
Because publication lags filing by roughly 18 months, the 2025 peak and 2026 partial-year count will keep revising upward. Re-running this scope in six to twelve months will sharpen the recent-momentum picture considerably.
Set up a monitoring search in EurekaTest claim drafts against the under-claimed branches
Bioinformatics-specific kernels and mixed-precision interconnect fabric hold single-digit record shares against the 3,930 total. A first claim there faces materially less dense prior art than a general G06F architecture claim.
Draft and check claims in EurekaCommon questions on heterogeneous computing patents
The ranked field covers 100 companies, with the leading assignee holding 551 of the 3,930 records in scope — well ahead of fifth place at 146. That said, the top five combined account for 40.8% of all records, meaning well over half the filing activity sits with companies outside the top five. Any answer naming a single leader should be read alongside that long tail, since low-volume filers can still hold specific, hard-to-design-around claims.
Filings grew 73% between 2021 and 2024, rising from 249 to 431 records in the dataset — 2024 is the most recent year that can be treated as complete. 2025 shows a higher raw count of 545, but because publication lags filing by roughly 18 months, both 2025 and 2026 figures will keep revising upward as more records publish. Growth should be read from the 2021-2024 span, not from the tail years.
G06F (electric digital data processing) is the dominant class, appearing on 75.8% of the 3,930 records in scope. G06N (AI-model computing) and H04L (digital information transmission) each cover roughly 13-14% of records, with image processing, video recognition, printed-circuit assembly and bioinformatics classes trailing in the single digits. Because a single record can carry several IPC classes, these percentages add up to more than 100% and shouldn't be summed together.
The smaller IPC classes point to it: bioinformatics-specific heterogeneous kernels sit at 2.0% of records and video/image recognition at 2.6%, both far below the 75.8% carried by core digital-data-processing claims. These are branches where the architecture claim exists but the specific mechanism — scheduling, coherence, or interconnect tuned to that domain — is comparatively under-filed. A first claim narrowly targeting one of these mechanisms faces materially less dense prior art than a general-purpose architecture claim.
They almost certainly didn't; publication lags filing by roughly 18 months, so the most recent one to two years in any patent dataset always look artificially low. The 2025 figure of 545 records is already the dataset's peak year despite this lag, and 2026 is a partial year by definition since the cut-off is mid-2026. Treat 2024 as the last complete year for any trend statement and expect 2025-2026 numbers to rise as more records publish.
Research Heterogeneous Computing Patent Landscape in depth with Eureka
Go past this page: query the whole heterogeneous computing 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.