Fastest
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 →https://www.patsnap.com/resources/blog/rd-blog/distributed-inference-partitioning-patent-landscape/ · Patsnap · data cut-off 2026-07-31 · downloaded from the live page
Patent Landscape · AI Inference ServingDistributed Inference Partitioning Patents: Who Leads and Where the Claims Sit
- Filing jumped 525% from 2021 to 2024, rising from 8 to 50 records a year before publication lag starts understating the most recent filings.
- One assignee holds 41 of 193 records, and the top 5 combined account for 48.2% of the entire dataset - concentration sits hard at the top.
- G06F and G06N each cover roughly half the field, while H04L digital-transmission claims trail at 27.5%, marking where interconnect-level claims are thinner.
193
Published Records
48%
Top-5 Share of All Records
+525%
Filing Growth 2021→2024
US
Leading Jurisdiction
Filing growth compares 2021 (8 records) with 2024 (50) — 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 193 records in scope (CR5), not by the ranked leaders only.
Published byPatsnap Research··7 min readSourced from Patsnap Eureka
Overview
What this patent landscape covers
Distributed inference partitioning describes the mechanics of splitting a trained model's execution across multiple compute nodes or pipeline stages, then managing where each piece runs as load, latency and cold-start conditions change. This dataset draws together 193 published records filed between 2015 and mid-2026 that combine partitioning or offloading language with the operational concerns that make it hard in practice - interconnect bandwidth, replica placement, load balancing and cold-start latency. That combination narrows the field to filings that treat partitioning as a systems problem rather than a purely algorithmic one.
Filing activity was negligible before 2017 and has accelerated sharply since, tracking the broader build-out of large-model serving infrastructure. Because publication typically lags filing by around 18 months, the most recent one to two years in any trend chart are undercounts rather than a sign of cooling demand.
Filing activity by year, 2017-2026
Source: Patsnap Eureka. Assignee ranking and totals. Derived from a Patsnap search on Distributed Inference Partitioning covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
The Numbers
Filing trends and technology composition
Two views of the same 193-record dataset: how filing volume has moved year over year, and which IPC subclasses the claims actually sit in.
A five-year climb, still understated at the edge
Annual filings moved from 0 in 2017 to a peak of 55 in 2025, with growth from 2021's 8 records to 2024's 50 marking a +525% rise over that span. 2025 and 2026 figures will keep rising as publication catches up, so treat the most recent two years as a floor, not a ceiling.
Compute and AI classes dominate, transmission trails
G06F (electric digital data processing) and G06N (AI-model computing) each appear in roughly half the 193 records, confirming that most filings frame partitioning as a compute or model-architecture problem. H04L (digital information transmission) covers 27.5% of records, a smaller share that suggests interconnect and network-layer claims are comparatively less crowded than compute-side ones.
Shares are the percentage of the 193 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Source: Patsnap Eureka. Filing trend and technology composition. Derived from a Patsnap search on Distributed Inference Partitioning covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.
Go deeper on Distributed Inference Partitioning with Eureka
This page is one run against one query. Ask Eureka your own question about distributed inference partitioning and every answer comes back with the patent numbers behind it.
Try EurekaKey Patents
Representative and most-cited filings
Representative Filing
Serving distributed inference deep learning (DL) models in serverless computing
META PLATFORMS, INC.
The filing describes a system that receives a request to initialize a container and then requests candidate servers from both an available-resource finder and a resource optimizer in parallel. A server allocator prioritizes one of the two candidates and feeds back the outcome of that choice, giving the system a mechanism to reconcile fast availability against optimized placement when serving a distributed DL model under serverless conditions.Filed by Meta Platforms, published 2023-10-05 - illustrates the placement-and-feedback pattern common across this dataset's serverless-serving claims.
Most-cited records in this dataset
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US20250259085A1 | Convergent Intelligence Fabric for Multi-Domain Orchestration of Distributed Agents with Hierarchical Memory … | 208 |
| 2 | US5146460A | Logic simulation using a hardware accelerator together with an automated error event isolation and trace faci… | 121 |
| 3 | US4970658A | Knowledge engineering tool | 67 |
| 4 | US6059835A | Performance evaluation of processor operation using trace pre-processing | 64 |
| 5 | US20250390352A1 | AI Serving Hardware and Software Frontier Enhancements | 39 |
| 6 | US20230144662A1 | Techniques for partitioning neural networks | 38 |
| 7 | US20220239758A1 | Distributed Machine-Learning Resource Sharing and Request Routing | 34 |
| 8 | US20250258708A1 | Federated distributed graph-based computing platform with hardware management | 33 |
| 9 | US11876858B1 | Cloud-based fleet and asset management for edge computing of machine learning and artificial intelligence wor… | 29 |
| 10 | US20210390460A1 | Compute and memory based artificial intelligence model partitioning using intermediate representation | 20 |
Citation counts accumulate over time, so older records such as the 1990s-era filings naturally lead this table; treat citation rank as a signal of influence within this corpus, not of current commercial relevance.
Each row carries its publication number; clicking a row searches Eureka by that number.
Source: Patsnap Eureka. Citation counts and representative records. Derived from a Patsnap search on Distributed Inference Partitioning covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
Run it yourself
Put your own technology through the same analysis
Where to run it
For builders
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 →Insights
What the data means for filing strategy
Three read-throughs from the concentration, growth and classification figures above.
Concentration
48.2% of 193 records
held by the top 5 assignees
The top of the field is dense, the tail is long
With one assignee alone holding 41 records and the top 5 combined covering 48.2% of all 193 records in scope, core partitioning and placement mechanics are well-claimed by a small group. The remaining 75 ranked companies each hold comparatively few filings, which typically means narrower, more defensible niches rather than open ground on the core mechanics.
Source: assignee ranking, 80 companies
Growth
+525%
filings, 2021 to 2024
Growth is real and recent, not a historical bump
Annual filings rose from 8 in 2021 to 50 in 2024, a +525% increase over three years and the clearest sign that distributed inference partitioning has moved from a niche systems topic to an actively contested filing area. Because 2025-2026 records are still publishing, the true current rate is likely higher than the chart shows.
Source: filing trend, 2021-2024
Classification
27.5% vs 52.3%
H04L share vs G06F share
Transmission-layer claims are comparatively open
G06F and G06N each sit above 50% of the 193 records, showing that most applicants claim partitioning through compute or model-architecture language. H04L, covering digital information transmission, appears in only 27.5% of records - a gap that points toward interconnect and network-scheduling claims as a less saturated angle on the same problem.
Source: IPC composition, 193 records
Eureka AI Agent
Looking for what nobody has claimed yet?
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to distributed inference partitioning, with the prior art for and against each one.
Source: Patsnap Eureka. Co-assignee relationships and derived observations. Derived from a Patsnap search on Distributed Inference Partitioning covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
Who's Filing
Players and where new activity is concentrated
The assignee ranking spans 80 companies across the full 193-record dataset, but recent momentum is uneven - most of the ranked leaders show flat or falling activity in the latest year, while a handful of newer entrants are still filing.
Momentum
+200% YoY
latest-year filings
New entrants are moving faster than incumbents
One newer entrant added 6 filings in the latest year, a +200% year-over-year increase, while several of the longest-established leaders in the ranking show zero or negative latest-year activity. That divergence suggests the earliest, broadest claims are largely staked out, and new activity is shifting toward more specific implementations.
Source: recent-year momentum by assignee
Receiving offices
92 of 193
filed at the USPTO
Filing is US-centred with a real international spread
The United States receives the largest share of filings at 92 of 193 records, followed by WIPO PCT filings, India, and the European Patent Office. Canada and Australia each account for a smaller number, indicating that most applicants treat the US as the primary market and use PCT or regional routes for secondary coverage.
Source: receiving office counts
Co-filing
2 records
strongest co-assignee pair
Joint filings are rare in this dataset
Co-assignment between organizations is uncommon here; the strongest pair in the dataset links two IBM-affiliated entities across two records. Most filings in this space are attributed to a single corporate applicant rather than a joint venture or research consortium.
Source: co-assignee pairs
Recent-year filing momentum by assignee
| Assignee | Recent year | YoY |
|---|---|---|
| ARMADA SYST INC | 6 | +200% |
| Google LLC | 1 | — |
| Avago Technologies | 0 | — |
| Microsoft Technology Licensing, LLC | 0 | -100% |
| International Business Machines Corporation (IBM) | 0 | — |
| NEOTERICC LLC | 0 | -100% |
| QOMPLX Inc. | 0 | -100% |
| Intel Corporation | 0 | -100% |
Source: Patsnap Eureka. Assignee-level momentum. Derived from a Patsnap search on Distributed Inference Partitioning covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
What's Next
Where to take this analysis
The dataset points to a field with dense compute-side claims and thinner interconnect-side coverage. The next steps depend on whether the goal is freedom-to-operate, licensing, or identifying open filing space.
Map claims against your own architecture
Run your specific partitioning or placement mechanism against the most-cited records and the top assignee's portfolio to see which claim elements actually overlap with your implementation.
Explore with EurekaTrack the under-claimed branches
Interconnect-aware placement and cold-start mitigation show thinner IPC coverage than the compute-heavy classes - worth monitoring as filing continues to climb through 2025-2026.
Set up monitoring in EurekaWatch momentum, not just rank
The largest historical filer shows flat recent activity while newer entrants are accelerating - portfolio rank today may not predict who is active in this space in two years.
Compare assignees in EurekaSource: Patsnap Eureka. Forward-looking reading of the same dataset. Derived from a Patsnap search on Distributed Inference Partitioning covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
FAQ
Common questions about this landscape
One assignee leads the ranked field with 41 of the 193 records in scope, and the top 5 assignees combined account for 48.2% of all records. That leaves a long tail of 75 further ranked companies each holding comparatively few filings. This concentration pattern means the core placement and partitioning mechanics are well-claimed by a small group, while more specific implementations remain open to a wider set of filers.
Filings rose from 8 in 2021 to 50 in 2024, a documented +525% increase over that three-year span, with a peak so far of 55 records in 2025. Because publication lags filing by roughly 18 months, the 2025 and 2026 figures will likely rise further as more filings from those years are published. The trend so far shows sustained acceleration rather than a plateau.
Most records sit in G06F (electric digital data processing, 52.3% of the 193 records) and G06N (AI-model computing, 51.3%), reflecting a focus on compute and model-architecture aspects of partitioning. H04L (digital transmission) covers a smaller 27.5% share, and smaller classes like G06Q, H04W, H04B, G06T and G08G each cover single-digit percentages, pointing to niche applications such as traffic control or wireless-specific implementations. Because records can carry multiple IPC classes, these shares add up to more than 100%.
The gap between the compute-heavy classes (G06F and G06N, each above 50% of records) and the transmission class H04L (27.5%) suggests that interconnect-aware placement and network-scheduling claims are comparatively less crowded than compute-side claims. Sub-areas like cold-start latency mitigation for replica scaling and cross-region interconnect-aware placement show thinner coverage relative to the core classes. That does not guarantee those areas are unclaimed, but it does indicate less dense prior art to search against.
The United States receives the largest share of filings with 92 of the 193 records, followed by WIPO PCT filings at 27, India at 26, and the European Patent Office at 23. Canada and Australia trail with smaller counts. This pattern indicates the US is the primary filing jurisdiction for this technology, with PCT and regional offices used for secondary or international coverage.
Answers are grounded in the same dataset. Derived from a Patsnap search on Distributed Inference Partitioning covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
Research Distributed Inference Partitioning in depth with Eureka
Go past this page: query the whole distributed inference partitioning 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.