Continuous Batching Patents: Who Leads, Where the Gaps Are 2026
- 39.3% concentration at the top. The five leading assignees hold 46 of 117 records in scope, while a long tail of single- and few-filing entrants fills out the rest of the 62 ranked companies.
- Filings grew 75% from 2021 to 2024. Annual filings rose from 4 in 2021 to 7 in 2024 — the last year the dataset treats as complete, since publication lags filing by roughly 18 months.
- G06F and H04L dominate the class mix. 53.0% of records carry a G06F digital-data-processing class and 36.8% carry H04L, while AI-model classing under G06N sits at just 6.8% of the 117 records.
Filing growth compares 2021 (4 records) with 2024 (7) — 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 117 records in scope (CR5), not by the ranked leaders only.
What this landscape covers
Continuous batching and iteration-level scheduling sit at the intersection of systems software and network transmission: the claims in scope describe how a server admits, queues, preempts and completes competing requests under throughput, tail-latency, fairness or memory-fragmentation constraints. This search pulls 117 published records filed or published between 2015 and mid-2026, spanning both general-purpose compute scheduling and wireless/network request scheduling that shares the same claim language.
The dataset mixes classic OS- and storage-scheduling priority art with newer inference-serving and wireless-network scheduling filings, so a search on this exact string surfaces both decades-old queueing patents and recent iteration-level batching claims side by side.
Filing trend and technology composition
Two views of the same 117 records: how filing activity has moved year over year, and which IPC subclasses the claims fall under.
Filing trend, 2017–2026
Annual filings moved from 6 in 2017 to a peak of 15 in 2025, with 2026 showing only 2 so far — expected, since publication trails filing by about 18 months and the most recent years are still filling in. The reliable growth signal is the 2021-to-2024 stretch, where filings rose from 4 to 7, a 75% increase.
IPC subclass composition
G06F (electric digital data processing) covers 53.0% of the 117 records and H04L (digital information transmission) covers 36.8%, confirming that most claims are framed as compute or network scheduling mechanics rather than as AI-model-specific methods — G06N appears on only 6.8% of records. Because a record can carry multiple classes, these shares sum to well over 100%.
Shares are the percentage of the 117 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Go deeper on Continuous Batching and Request Scheduling with Eureka
This page is one run against one query. Ask Eureka your own question about continuous batching and request scheduling and every answer comes back with the patent numbers behind it.
Try EurekaRepresentative and most-cited records
US20090031314A1 — Fairness in memory systems
Architecture for a multi-threaded system that applies fairness to thread memory request scheduling so that access to shared memory is fair among different threads and applications. A fairness scheduling algorithm provides fair memory access to different threads in multi-core systems, avoiding thread starvation and performance loss caused by a memory performance hog application, by computing thread slowdown as the ratio of real latency experienced to the ideal latency the thread would see running alone.Filed by Microsoft Technology Licensing, LLC, published 2009-01-29.
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US5619695A | Method and apparatus for scheduling resources | 100 |
| 2 | US20090303937A1 | Mobile station, base station and communications method | 67 |
| 3 | US7035971B1 | Request scheduling to mirrored heterogeneous storage arrays | 60 |
| 4 | WO2002039696A2 | Assured quality-of-service request scheduling | 55 |
| 5 | US6212623B1 | Universal dependency vector/queue entry | 50 |
| 6 | US20090031314A1 | Fairness in memory systems | 45 |
| 7 | US20090217273A1 | Controlling interference in shared memory systems using parallelism-aware batch scheduling | 44 |
| 8 | US20090044189A1 | Parallelism-aware memory request scheduling in shared memory controllers | 41 |
| 9 | US20090216962A1 | Prioritization of multiple concurrent threads for scheduling requests to shared memory | 39 |
| 10 | US8180975B2 | Controlling interference in shared memory systems using parallelism-aware batch scheduling | 38 |
Citation counts favour older records simply because they have had longer to accumulate citations inside the searched corpus — read them as a signal of influence on the field, not as a measure of current 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 tells a filer
Three read-throughs of the concentration, class and geography figures that matter for a freedom-to-operate or whitespace call.
The top of the field is dense, the rest is thin
The five leading assignees account for 46 of the 117 records in scope — 39.3% — while the top 10 combined reach 62.4% (73 records). Beyond that the ranking of 62 companies thins quickly into single- and few-filing entrants, which is where most freedom-to-operate risk in adjacent claim language actually sits.
Growth is real but the newest years understate it
Filings rose from 4 in 2021 to 7 in 2024, a 75% increase over three years, with 2025 posting the dataset's peak at 15. Treat 2025 and 2026 counts as provisional floors rather than a plateau, since publication typically lags filing by about 18 months.
Compute scheduling outweighs network scheduling, but overlaps heavily
G06F covers 53.0% of the 117 records and H04L covers 36.8%; because records can carry both, a large share of filings frame the same scheduling mechanism as both a compute and a transmission problem. G06N, the AI-model class, appears on only 6.8% of records, suggesting claim language has not yet caught up to inference-serving-specific framing.
Filing office split is close, not lopsided
United States receiving offices account for 42 records and China for 34, with EPO and WIPO PCT each at 14 and Taiwan and India trailing well behind. No single office dominates the way the assignee ranking does, which points to parallel rather than concentrated jurisdictional strategy.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to continuous batching and request scheduling, with the prior art for and against each one.
Who is filing, and where the gate sits
The ranked leaders span systems vendors, telecom equipment makers and one research institute; momentum among the top names has cooled in the most recent complete year, consistent with the publication lag across the whole dataset.
A single filer sits well ahead of fifth place
The leading assignee holds 13 records against 6 for the fifth-ranked filer and 5 for tenth place, a steep early drop-off typical of a field where one systems vendor treated scheduling and fairness claims as a defensible portfolio early.
Recent-year counts read low across the board
Several of the most active historical filers show zero filings in the latest tracked year, including a -100% year-on-year move for one. This is consistent with the 18-month publication lag rather than a genuine pullback, since 2025 was the dataset's peak filing year.
Co-filing is limited and clustered around named inventors
Only 9 co-assignee pairs appear across the dataset, and the strongest repeated pairings link a single corporate assignee with individual named inventors rather than joint corporate filings — collaboration in this field is the exception, not the norm.
| Assignee | Recent year | YoY |
|---|---|---|
| Microsoft Corp | 0 | — |
| Huawei Technologies Co., Ltd. | 0 | -100% |
| Advanced Micro Devices, Inc. | 0 | — |
| Industrial Technology Research Institute | 0 | — |
| Hewlett Packard Development Company, L.P. | 0 | — |
| Microsoft Technology Licensing, LLC | 0 | — |
| ZTE Corporation | 0 | — |
| Board of Regents of the University of Nebraska | 0 | — |
Where to take this next
The landscape numbers point to specific follow-up work rather than a single conclusion.
Check freedom-to-operate against the top 10
62.4% of all 117 records sit with the ten leading assignees, so any new filing in core admission-control or preemption claim language should be checked against that group first before the long tail.
Explore the assignee rankingWatch the 2025 peak resolve
2025's count of 15 is the dataset's high point but is still an incomplete publication year; re-running this search in a year will show whether the 2021–2024 growth trend continued or was a one-year spike.
Track filing trends in EurekaProbe the under-claimed branches
G06N classing sits at just 6.8% of records despite the underlying technology's inference-serving relevance, suggesting AI-specific scheduling claims are still open relative to general compute and network scheduling art.
Run a whitespace search in EurekaCommon questions on this landscape
It is moderately concentrated at the top and thin below it. The five leading assignees hold 46 of the 117 records in scope, or 39.3%, and the top ten combined reach 62.4%. Beyond the ranked leaders, the field spreads across 62 companies total, many with just one or two filings, so most of the crowding sits in a small set of core claim families rather than across the whole dataset.
Yes, through the last complete data year. Filings rose from 4 in 2021 to 7 in 2024, a 75% increase, and 2025 recorded the dataset's peak at 15. Counts for 2025 and 2026 should be read as provisional floors rather than a slowdown, because publication typically lags actual filing by around 18 months.
Most records fall under G06F (electric digital data processing), which covers 53.0% of the 117 records, and H04L (digital information transmission), covering 36.8%. Wireless-specific claims under H04W appear on 14.5% of records, while AI-model-specific classing under G06N is comparatively rare at 6.8%, indicating claim language in this space is still largely framed as general compute or network scheduling rather than model-serving specific.
The class composition data points to AI-model-aware scheduling as relatively open: G06N classing sits at only 6.8% of the 117 records despite continuous batching's strong ties to inference serving. Specific sub-areas worth checking include iteration-level batch admission control, SLO-aware preemption policies and cross-request cache eviction ordering, all of which sit adjacent to the dense core scheduling art without being fully occupied by it.
The leading assignee holds 13 of the 117 records, well ahead of the fifth-ranked filer at 6 and the tenth-ranked filer at 5. The ranked group spans systems and cloud vendors, telecom equipment makers and at least one research institute, but recent-year momentum has cooled across several of the most active historical filers, consistent with the publication lag affecting the newest years of data.
Research Continuous Batching and Request Scheduling in depth with Eureka
Go past this page: query the whole continuous batching and request scheduling 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.