DPU and SmartNIC Offload Patents: Leaders & Filing Trends 2026
- Filing has plateaued, not accelerated. annual filings peaked at 100 in 2022 and the 2026 count of 22 (partial year) sits well below that mid-decade high.
- Ownership is dispersed, not locked up. the top 5 assignees together hold 18.6% of all 999 records and the top 10 hold 27.7% — a long tail of single- and few-filing entrants fills the rest.
- Filing offices skew toward the US and China. United States (377) and China (206) receive roughly triple the filings of the next-largest office, EPO (112), shaping where freedom-to-operate work matters most.
Filing growth compares 2021 (53 records) with 2024 (76) — 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 999 records in scope (CR5), not by the ranked leaders only.
What this landscape covers
DPU and SmartNIC offload architectures move networking, security and storage workloads off the host CPU onto dedicated silicon. The 999 records in scope span filings from 2015 through the partial-year 2026 cut-off, covering offload engines, fixed-function versus programmable tradeoffs, isolation mechanisms and the infrastructure functions that let a host reclaim CPU cycles. Because publication lags filing by roughly eighteen months, the most recent filing years are undercounted and should be read as a floor, not a ceiling.
The dataset draws on assignee identity, IPC classification, filing date and citation counts to show where claim density is high, where it is thin, and which organisations are still actively filing versus those that have gone quiet.
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 999 records: how filing volume has moved year over year, and how those records distribute across IPC subclasses. Because a single record can carry multiple IPC classes, the composition shares are read against the full record total and will sum past 100%.
A flat-to-declining filing curve after a 2022 peak
Annual filings rose from 33 in 2017 to a peak of 100 in 2022, then eased back toward 22 in 2026 — a partial year. The 2022 peak sitting at the trend's midpoint signals the field has cooled from its most active filing window rather than continuing to accelerate.
Digital data processing and transmission dominate the classification mix
G06F (electric digital data processing) covers 33.4% of the 999 records and H04L (digital information transmission) covers 21.0%, together anchoring the core offload-engine and data-path claims. Wireless (H04W, 6.0%), AI-based computing (G06N, 4.7%) and control systems (G05B, 3.6%) appear at much lower density, marking adjacent areas that are referenced but not heavily claimed.
Shares are the percentage of the 999 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Go deeper on DPU and SmartNIC Offload Architectures with Eureka
This page is one run against one query. Ask Eureka your own question about dpu and smartnic offload architectures and every answer comes back with the patent numbers behind it.
Try EurekaRepresentative and most-cited filings
US9723518B2 — Network offload method and apparatus
Describes a method where a first network controller receives network state information for a terminal, generates an offload policy from that state, and sends the policy to the terminal and/or an access device — the mechanics of dynamically routing traffic off a given access network based on live terminal state.Filed by Huawei Technologies, published 2017-08-01.
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US4780599A | Apparatus for retrieving stored information about various items in response to coding on the items | 264 |
| 2 | US6233339B1 | Physical property based cryptographics | 173 |
| 3 | US20020076057A1 | Method and apparatus for determining in situ the acoustic seal provided by an in-ear device. | 151 |
| 4 | US20190012278A1 | Data processing unit for compute nodes and storage nodes | 132 |
| 5 | US20040199808A1 | State recovery and failover of intelligent network adapters | 129 |
| 6 | US5936913A | Acoustic formation logging system with improved acoustic receiver | 126 |
| 7 | US6687377B2 | Method and apparatus for determining in situ the acoustic seal provided by an in-ear device | 125 |
| 8 | US6458605B1 | Method and apparatus for controlling photolithography overlay registration | 113 |
| 9 | US20100179421A1 | System and method for non-invasive instantaneous and continuous measurement of cardiac chamber volume. | 102 |
| 10 | US20050119643A1 | Method and apparatus for opto-thermo-mechanical treatment of biological tissue | 102 |
Citation counts accumulate over time, so older filings are structurally favoured; treat these as markers of influence within the searched corpus rather than of current technical 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 say about the field
Three signals stand out once filing volume, assignee concentration and classification density are read together.
No single owner controls the field
The leader holds 60 records against a fifth-place holder at 23 and a tenth-place holder at 15 — a steep drop-off after the leader, then a long, shallow tail. Combined, the top 5 assignees account for 18.6% of all 999 records and the top 10 for 27.7%, leaving the majority of filings spread across the remaining ranked companies and unranked entrants.
Filing has cooled from its 2022 high
Volume grew from 33 filings in 2017 to a peak of 100 in 2022, then declined toward 22 in the partial 2026 year. Several previously active filers show flat or negative year-on-year momentum, consistent with a maturing claim landscape rather than one still in a land-grab phase.
Claims cluster on data processing and transmission
G06F and H04L together anchor the bulk of offload-engine and data-path claims, while adjacent classes like G06N (AI-based computing, 4.7%) and G05B (control systems, 3.6%) show far lower density — these are areas where the core offload mechanics are claimed heavily but the surrounding application layers are comparatively open.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to dpu and smartnic offload architectures, with the prior art for and against each one.
Who is filing, and who has slowed down
Recent-year momentum diverges sharply from historical filing volume: several assignees with substantial cumulative counts show zero or negative filings in the latest year, while the field's under-claimed branches remain open to new entrants.
A clear leader, but not a lock
The top-ranked assignee holds 60 of the 999 records, well ahead of fifth place at 23 and tenth place at 15. That gap suggests an early-mover advantage in core claim space rather than an unassailable position, since the combined top 10 still account for only 27.7% of all records.
Established filers have gone quiet
Multiple assignees with meaningful historical filing counts — including several ranked in the top tier — show zero filings in the latest tracked year, with year-on-year changes as steep as -100%. This does not mean these organisations have exited the space; it more likely reflects publication lag on recent applications not yet in the public record.
Filing activity is US- and China-centric
The United States (377 records) and China (206) together receive far more filings than EPO (112), WIPO/PCT (86), India (57) or Austria (22). Freedom-to-operate diligence for commercial deployment should weight these two jurisdictions first.
| Assignee | Recent year | YoY |
|---|---|---|
| VMware, Inc. | 1 | -50% |
| Mellanox Technologies, Ltd. | 1 | -67% |
| Fungible, Inc. | 0 | — |
| Dell Products L.P. | 0 | -100% |
| International Business Machines Corporation | 0 | — |
| Huawei Technologies Co., Ltd. | 0 | -100% |
| Nozomi Networks Sagl | 0 | — |
| Xilinx, Inc. | 0 | -100% |
Where to take this analysis
The dataset points to specific next steps depending on whether the goal is freedom-to-operate, competitive tracking, or identifying open claim space.
Map claims against your own architecture
Run your specific offload mechanism — fixed-function block, programmable pipeline, or isolation scheme — against the densest IPC clusters (G06F, H04L) to find where existing claims sit closest to your design.
Explore in Patsnap EurekaTrack the quiet filers
Several assignees with large historical footprints show no recent-year filings in this dataset; monitoring their continuation and PCT activity over the next 12-18 months will clarify whether this reflects publication lag or an actual pullback.
Set up assignee monitoring in Patsnap EurekaScope the under-claimed branches
The gaps in AI-integrated offload policy and power-budget-aware scheduling are candidates for a first-mover filing strategy, but require a targeted prior-art search before drafting.
Run a white-space search in Patsnap EurekaCommon questions about this landscape
The leading assignee in this dataset holds 60 of the 999 records in scope, notably ahead of the fifth-ranked company at 23 and the tenth-ranked at 15. That said, the top 10 assignees combined account for only 27.7% of all records, so ownership is concentrated at the very top but not consolidated across the field. Anyone assessing competitive position should look at both the leader's count and the long tail of single-filing entrants behind it.
Filing volume rose from 33 records in 2017 to a peak of 100 in 2022, then declined toward 22 in 2026, though 2026 is a partial year. Because publication typically lags actual filing by around 18 months, the true 2025-2026 volume is understated in any dataset pulled today. Taken together, the trend looks flat to declining rather than accelerating, which is a different signal than a field still in its land-grab phase.
G06F (electric digital data processing) covers 33.4% of the 999 records and H04L (digital information transmission) covers 21.0%, making them the two densest classes by a wide margin. Adjacent classes such as A61B, H04W, G01R and G06N appear at much lower density, in the 4-7% range. Because records can carry multiple IPC codes, these shares sum to more than 100% and should be read individually against the full record count, not against each other.
By receiving office, the United States leads with 377 records and China follows with 206, together far outpacing EPO (112), WIPO/PCT (86), India (57) and Austria (22). A freedom-to-operate review with limited budget should prioritize US and Chinese filings before expanding to the smaller offices. The classification density in G06F and H04L also indicates where claim language is most likely to overlap with a new offload design.
US9723518B2, assigned to Huawei Technologies and published in 2017, covers a method where a network controller receives a terminal's network state information, generates an offload policy from that state, and transmits the policy to the terminal or an access device. It is narrowly scoped to controller-driven, state-based offload policy generation and distribution, not to offload hardware architecture generally. A new design that offloads workloads through a different triggering mechanism, or that does not rely on a network controller generating and pushing a policy in this way, has room to design around it, though a full claim chart review is the only reliable way to confirm that.
Research DPU and SmartNIC Offload Architectures in depth with Eureka
Go past this page: query the whole dpu and smartnic offload architectures 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.