Query Optimization Patents: Who Leads, Where the Gaps Are 2026
- Filing peaked in 2022 at 9 families and has not been matched since — this is a cooling filing cycle, not a rising one.
- Every tracked major assignee shows zero filings in the latest full year, including SAP with a -100% year-on-year drop — a pause worth watching, not necessarily an exit.
- Co-assignment is almost non-existent only 2 co-assignee pairs across 72 families, meaning most claim territory here was staked out solo.
What this landscape covers
This landscape tracks patent families where the title or claims name a query optimizer, query execution engine, or vectorized execution, combined with technical anchors like cardinality estimation, join ordering, adaptive execution, code generation, or a cost model, filed under classifications covering digital data processing, program execution, and software engineering. The corpus spans filings from 2015 through the 2026 cut-off, with 72 published families forming the ranking base.
Because publication typically lags filing by around 18 months, the most recent one to two years in the trend chart will always look thinner than they eventually turn out to be. Read the tail of the chart as a floor, not a ceiling.
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 72-family corpus: how filing activity has moved year over year, and which classification codes the claims actually sit under.
A flat-to-declining filing curve
Filings rose from 2 families in 2017 to a peak of 9 in 2022, then did not sustain that pace. With 2022 sitting at the dataset's midpoint and no year since exceeding it, the underlying trend reads flat to declining rather than growing — consistent with a technology area where the foundational claim positions were staked earlier in the window.
Concentrated in core data processing, with an AI overlay
All 72 records classify under G06F (electric digital data processing), confirming this is squarely a core software-execution corpus rather than an adjacent field. A meaningful secondary cluster of 14 records also touches G06N (AI-based computing), pointing to cost models and cardinality estimators increasingly framed as learned or AI-assisted components. Everything else — networking, wireless, even an internal-combustion-engine and lubrication-system stray — is incidental noise at 1-3 records each.
Shares are the percentage of the 72 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Go deeper on Query Optimization and Execution Engines with Eureka
This page is one run against one query. Ask Eureka your own question about query optimization and execution engines and every answer comes back with the patent numbers behind it.
Try EurekaThe most-cited records in this corpus
System and method for cardinality estimation feedback loops in query processing (US11334538B2)
A query host executes queries against data sources via an engine based on estimated cardinalities, while query monitors generate event signals during and at completion of execution. These signals — covering actual data cardinality, runtime statistics and query parameters — feed a feedback optimizer that analyzes them and generates change recommendations for later executions of the same or similar queries.Filed by Microsoft Technology Licensing, LLC; granted 2022-05-17.
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US20050004892A1 | Query optimizer system and method | 102 |
| 2 | US7146352B2 | Query optimizer system and method | 81 |
| 3 | US10592506B1 | Query hint specification | 29 |
| 4 | US20200379963A1 | System and method for cardinality estimation feedback loops in query processing | 21 |
| 5 | US20220004553A1 | Automated feedback and continuous learning for query optimization | 19 |
| 6 | US20160306847A1 | Apparatus and Method for Using Parameterized Intermediate Representation for Just-In-Time Compilation in Data… | 14 |
| 7 | US9489180B1 | Methods and apparatus for joint scheduling and layout optimization to enable multi-level vectorization | 14 |
| 8 | US20200142894A1 | Systems, methods, and media for improving the effectiveness and efficiency of database query optimizers | 12 |
| 9 | US20210263932A1 | System and method for machine learning for system deployments without performance regressions | 11 |
| 10 | US20230385261A1 | Scalable index tuning with index filtering and index cost models | 10 |
Citation counts reflect influence within the searched corpus and are skewed toward older filings; treat them as historical signal rather than 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 says about this field
Three patterns stand out once the filing curve, citation table and assignee activity are read together.
The build-out phase looks largely done
A rise from 2 families in 2017 to 9 in 2022, followed by no year exceeding that peak, suggests the core patentable ideas around cardinality estimation, join ordering and adaptive execution were substantially claimed by the early 2020s. New entrants now face denser prior art in the exact terms this search targets.
Influence sits with early, broad filings
The two most-cited records in the table — both titled "Query optimizer system and method" — carry citation counts far above the rest of the corpus (102 and 81). That gap indicates foundational optimizer architecture claims that later filings had to design around or build on, rather than a field with many equally influential contributions.
Filers work alone, not in consortia
With only two co-assignee pairs identified across the entire corpus, cross-organization filing is rare. Most families here represent a single company's or university's independent claim, which makes the ranking table a reasonably clean read on who actually built what.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to query optimization and execution engines, with the prior art for and against each one.
| Assignee | Co-assignee | Shared families |
|---|---|---|
| Huawei Technologies Co., Ltd. | East China Normal University | 1 |
| Huawei Technologies Co., Ltd. | KARTHIK VENKATESH SRINIVAS | 1 |
Both identified co-assignee pairs involve the same lead filer working with a university and an individual inventor respectively — a pattern of occasional academic or named-inventor collaboration rather than joint-venture filing.
Who is active, and who has gone quiet
Every major assignee tracked in the recent-momentum data shows zero filings in the latest full year — including a documented -100% year-on-year drop for SAP. That is unusual for a field with 72 total families, and it reads less like exit and more like a pause after the 2022 peak, with publication lag likely masking filings still working through the pipeline.
The largest filers have all stopped short-term
Microsoft (both its Technology Licensing and corporate entities), Huawei, SAP, and IBM all show zero filings in the most recent full year in this dataset. Given the 18-month publication lag, some of this is filings not yet public rather than a genuine halt.
SAP's year-on-year decline is the sharpest documented
SAP is the only assignee in the momentum data with an explicit year-on-year percentage change, and it is a full drop to zero. Whether that reflects a strategic shift away from patenting query-engine internals or simply a gap before the next filing cohort publishes is not answerable from filing counts alone.
Universities and individuals appear as co-filers, not leaders
East China Normal University and a named inventor both appear paired with Huawei in the co-assignee data, suggesting Huawei sources some of its query-engine IP through academic or individual collaboration rather than filing purely in-house.
| Assignee | Recent year | YoY |
|---|---|---|
| Microsoft Technology Licensing, LLC | 0 | — |
| Huawei Technologies Co., Ltd. | 0 | — |
| SAP SE | 0 | -100% |
| Microsoft Corporation | 0 | — |
| International Business Machines Corporation (IBM) | 0 | — |
| Stream Analyze Sweden AB | 0 | — |
| Red Hat, Inc. | 0 | — |
| Oracle International Corporation | 0 | — |
Where to take this
The filing curve and assignee momentum both point to a field past its initial claim-staking phase, which changes what a freedom-to-operate or whitespace exercise should focus on.
Map the top-cited families claim by claim
The two highest-cited optimizer patents anchor much of the downstream prior art in this corpus; understanding their claim scope is a faster route to freedom-to-operate clarity than reading the full 72-family set.
Explore claim structures in EurekaWatch for the 2024-2026 filings still in the pipeline
Given the roughly 18-month publication lag, the apparent zero-filing years for major assignees may fill in as pending applications publish. A pipeline check now avoids acting on a stale picture later.
Track new publications in EurekaProbe the under-claimed branches directly
Learned cost models and adaptive mid-query re-optimization show thin coverage relative to the corpus size — a targeted search in those exact terms will clarify whether that is true white space or just a labelling gap.
Run a targeted search in EurekaCommon questions about this landscape
The two highest-cited records in this corpus, both titled "Query optimizer system and method," carry citation counts of 102 and 81 respectively and predate most of the rest of the dataset. High citation counts inside a searched corpus tend to favour older filings simply because they have had more time to be cited, so treat them as markers of foundational influence rather than proof that they are the most commercially active patents today. For current commercial activity, cross-reference against the recent-year momentum data rather than the citation table alone.
No — filings rose from 2 families in 2017 to a peak of 9 in 2022, and no subsequent year has matched that peak, which reads as a flat-to-declining trend rather than growth. Because publication lags filing by around 18 months, the most recent one to two years will always understate true filing activity, so some of the apparent decline may still fill in. Even accounting for that lag, the shape of the curve suggests the field's core claim positions were largely established by the early 2020s.
Cardinality estimation is the process a query optimizer uses to predict how many rows a given operation will return, and that prediction drives decisions like join order and access-path selection. It appears throughout this corpus because getting it wrong is one of the most common causes of poor query plans, making it a natural target for patentable improvements — including feedback-loop approaches, as in the representative Microsoft filing in this dataset, that adjust future estimates based on actual runtime statistics.
Every assignee tracked in the recent-year momentum data, including Microsoft, Huawei, SAP, and IBM, shows zero filings in the most recent full year, and SAP shows an explicit -100% year-on-year drop. Part of this is a genuine slowdown following the 2022 peak in the broader dataset, and part is almost certainly the roughly 18-month gap between filing and publication, which means some 2025-2026 filings from these same companies have not yet appeared in public records. It is not safe to conclude these companies have abandoned the space based on this figure alone.
Based on the technology composition data, learned or AI-assisted cost models sit at the intersection of the core G06F corpus and a smaller 14-record G06N cluster, suggesting that combination is still comparatively thin relative to traditional cost-model claims. Adaptive re-optimization for streaming or federated join scenarios, and vectorized code generation targeted at heterogeneous hardware, also show limited direct representation in this corpus. These are starting points for a deeper freedom-to-operate search, not confirmed gaps, since a narrow search string will always miss some adjacent filings.
Research Query Optimization and Execution Engines in depth with Eureka
Go past this page: query the whole query optimization and execution engines 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.