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/confidential-computing-for-accelerators-patent-landscape/ · Patsnap · data cut-off 2026-07-31 · downloaded from the live page
Patent Landscape · Trusted Execution EnvironmentsConfidential Computing for Accelerators Patents: Who Leads and Where Filings Concentrate
- 76.3% of records sit with just five assignees out of 131 records in scope — this is a field defined by a small group of chipmakers, not a broad ecosystem.
- Filings peaked in 2020 at 22 and 2021-to-2024 activity fell 19 to 9, a 53% drop across a complete three-year window.
- G06F and H04L dominate at 71.0% and 48.9% of records respectively, leaving AI-model computing (G06N, 5.3%) and memory-class claims (G11C, 3.8%) comparatively open.
131
Published Records
76%
Top-5 Share of All Records
-53%
Filing Growth 2021→2024
US
Leading Jurisdiction
Filing growth compares 2021 (19 records) with 2024 (9) — 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 131 records in scope (CR5), not by the ranked leaders only.
Published byPatsnap Research··7 min readSourced from Patsnap Eureka
Overview
What this landscape covers
Confidential computing for accelerators concerns how a system protects data and code running on GPUs, IPUs and other offload devices from the host, the hypervisor and other tenants sharing the same hardware. The search underlying this page pairs attestation and link-encryption terms with the operational vocabulary of the field — bounce buffers, memory bandwidth, multi-tenant isolation, trusted drivers, performance overhead and address translation service — to isolate records that actually engineer trust boundaries around accelerator memory and interconnect, rather than general cryptography filings.
131 published records fall within scope, spanning receiving offices led by the United States, with meaningful activity at the EPO, WIPO and in Germany. Because publication lags filing by roughly 18 months, the most recent years in any trend understate real activity; 2024 is the newest year that can be read as a complete data point.
Filing activity, 2017–2026
Source: Patsnap Eureka. Assignee ranking and totals. Derived from a Patsnap search on Confidential Computing for Accelerators 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 131 records: how filing volume moved year over year, and which IPC subclasses the claims actually sit in.
A sharp rise, then a pullback from a 2020 peak
Filings climbed from nothing in 2017 to a peak of 22 in 2020, then eased. Using the exact figures in the record, 2021's 19 filings fell to 9 by 2024 — a 53% decline across a window that is fully published. Treat 2025 and 2026 as still filling in rather than as evidence of a continued slide.
Two classes carry the field
G06F (electric digital data processing) appears on 71.0% of the 131 records and H04L (digital information transmission) on 48.9% — expected given the subject sits at the boundary of compute and network security. Because records carry multiple classes, these shares add to more than 100%. The smaller classes — G06N, G11C, H04W, G07F, H04J and A01M — mark where accelerator-specific memory and AI-model claims are still thin.
Shares are the percentage of the 131 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 Confidential Computing for Accelerators covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.
Go deeper on Confidential Computing for Accelerators with Eureka
This page is one run against one query. Ask Eureka your own question about confidential computing for accelerators and every answer comes back with the patent numbers behind it.
Try EurekaKey Patents
A representative recent filing
Representative Record
Memory Bus Integrity and Data Encryption (IDE) — US12549331B2
INTEL CORPORATION
A memory subsystem includes link encryption for the system memory data bus. The memory controller can provide encryption for data at rest and link protection, and can optionally provide link encryption so the system protects data in transit. A link decryption engine on the memory module can decrypt link encryption where used and perform a link integrity check against a link integrity tag, with the memory devices storing the encrypted protected data and ECC data only after link protection verification.Filed by Intel Corporation, published 2026-02-10.
Most-cited records in this landscape
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US20050102529A1 | Mobility access gateway | 319 |
| 2 | US20210117249A1 | Infrastructure processing unit | 222 |
| 3 | US20100042824A1 | Hardware trust anchors in SP-enabled processors | 160 |
| 4 | US20070208864A1 | Mobility access gateway | 155 |
| 5 | US20210117242A1 | Infrastructure processing unit | 152 |
| 6 | US20040087304A1 | Integrated web cache | 142 |
| 7 | US7562393B2 | Mobility access gateway | 114 |
| 8 | US20110007705A1 | Mobility access gateway | 87 |
| 9 | US20190042463A1 | Apparatus and method for secure memory access using trust domains | 85 |
| 10 | US7499401B2 | Integrated web cache | 70 |
Citation counts favour older records simply because they have had longer to accumulate references — read them as a signal of influence within this corpus, not as a ranking of current 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 Confidential Computing for Accelerators 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, trend and citation figures above.
Concentration
76.3%
of 131 records held by top 5 assignees
This is a chipmaker's field, not a startup field
With the top five assignees combining for 76.3% of all 131 records and the top ten reaching 89.3%, freedom-to-operate work here means clearing a small number of well-resourced portfolios rather than a diffuse field of small filers.
Top 10 = 89.3% of records
Trend
-53%
filings, 2021 to 2024
Volume has cooled from its 2020 peak
Filing activity peaked at 22 in 2020 and had fallen to 9 by 2024, a 53% decline across a fully published window. That is a pullback from an unusually active period rather than proof the underlying technology is exhausted — claim space that was heavily contested in 2019-2021 is now easier to survey.
Peak year: 2020 at 22 filings
Citations
319 cites
on the most-cited record in scope
Influence sits with early infrastructure patents
The most-cited records in this set date to the mid-2000s and describe mobility gateways and hardware trust anchors, well ahead of the accelerator-specific surge. Their high citation counts reflect age and foundational status within this searched corpus, not current commercial weight.
Oldest top-cited record: US20050102529A1
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 confidential computing for accelerators, with the prior art for and against each one.
Source: Patsnap Eureka. Co-assignee relationships and derived observations. Derived from a Patsnap search on Confidential Computing for Accelerators 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
Players
Who is filing, and where the gaps sit
39 companies make up the entire ranked list returned for this dataset — this is not a top-50 or top-100 cut, it is the full set of ranked assignees for these 131 records.
Leader
71 records
held by the top-ranked assignee
One filer sets the pace by a wide margin
The leading assignee's 71 records dwarf the fifth-place count of 4, meaning most of the concentration in this field comes from a single dominant portfolio rather than an even spread among a handful of firms.
Fifth place: 4 records
Mid-field
3 records
held by the tenth-ranked assignee
The drop-off past the leader is steep
Once past the top handful, counts fall quickly to single digits — tenth place holds just 3 records — before the long tail of single- and double-filing entrants that makes up most of the 39-company ranking.
39 companies ranked in total
Momentum
-100% YoY
for several previously active filers
Recent-year activity has gone quiet across the board
Several of the historically active assignees show zero filings in the latest year, with year-over-year drops as steep as -100% for some and -67% for others. Given the 18-month publication lag, this likely reflects filings still working through the pipeline rather than a genuine stop.
Latest-year activity is still filling in
Recent-year filing momentum by assignee
| Assignee | Recent year | YoY |
|---|---|---|
| Micron Technology, Inc. | 1 | -67% |
| Intel Corporation | 0 | -100% |
| Advanced Micro Devices, Inc. (AMD) | 0 | -100% |
| SMITH NED M | 0 | — |
| DOSHI KSHITIJ ARUN | 0 | — |
| Alcatel | 0 | — |
| Nokia of America Corporation | 0 | — |
| Dell Products L.P. | 0 | — |
Source: Patsnap Eureka. Assignee-level momentum. Derived from a Patsnap search on Confidential Computing for Accelerators 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 few concrete next steps for teams doing freedom-to-operate or whitespace work in this area.
Map the leader's claim boundaries
With one assignee holding 71 of 131 records, understanding exactly what that portfolio covers — and where its claims stop short — is the single highest-value piece of analysis for anyone filing in this space.
Explore assignee portfolios in EurekaWatch the smaller IPC classes
G06N, G11C, H04W, G07F, H04J and A01M together account for a small share of records but mark where accelerator-specific AI and memory claims are least crowded.
Track technology classes in EurekaReassess once 2025-2026 fully publishes
Because publication lags filing by roughly 18 months, the apparent cooling after 2020 needs to be revisited once the most recent two years finish populating.
Set a filing-trend alert in EurekaSource: Patsnap Eureka. Forward-looking reading of the same dataset. Derived from a Patsnap search on Confidential Computing for Accelerators 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 this landscape with 71 of the 131 records in scope, far ahead of the fifth-ranked filer at 4 records. The top five assignees combined hold 76.3% of all records, and the top ten reach 89.3%, so this is a field where a small number of chipmakers control most of the claim space. The full ranking covers 39 companies, including memory makers, processor vendors and a few infrastructure and telecom filers with narrower but relevant portfolios.
Filing activity rose sharply to a peak of 22 records in 2020, then declined to 9 by 2024 — a 53% drop across that fully published three-year window. Because patent publication typically lags filing by around 18 months, the lower counts shown for 2025 and 2026 are still filling in and should not be read as continued decline. The honest read is that the field cooled from an unusually active 2019-2021 period rather than that innovation has stopped.
The underlying search combines attestation and encryption terms — device attestation, encrypted device memory, link encryption — with accelerator-specific operational terms such as bounce buffer, memory bandwidth, multi-tenant isolation, trusted driver, performance overhead and address translation service. This targets patents that engineer trust boundaries specifically around accelerator memory and interconnects, rather than general-purpose cryptography or generic trusted-execution-environment filings. G06F and H04L are the dominant IPC classes, appearing on 71.0% and 48.9% of the 131 records respectively.
The smaller IPC classes in this dataset point to comparatively under-claimed branches: G06N (AI-model computing) at 5.3% of records, G11C (static and digital memories) and H04W (wireless communication) each at 3.8%, and H04J (multiplex communication) at 1.5%. These sub-areas sit adjacent to the dense G06F/H04L core but carry far fewer filings, suggesting narrower prior art to design around for claims that combine accelerator isolation with AI workloads, memory-class hardware, or wireless-linked deployments.
US12549331B2, filed by Intel Corporation and published 2026-02-10, covers a memory subsystem with link encryption for the system memory data bus, including a memory controller providing data-at-rest encryption and link protection, and a link decryption engine on the memory module that verifies a link integrity tag before storing encrypted data and ECC data. It is a recent, concrete example of the memory-bus and link-encryption claim style that runs through much of this landscape's G06F/H04L core. Teams working on memory-bus security for accelerators should review its specific claim scope before designing similar link-encryption or integrity-check mechanisms.
Answers are grounded in the same dataset. Derived from a Patsnap search on Confidential Computing for Accelerators 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 Confidential Computing for Accelerators in depth with Eureka
Go past this page: query the whole confidential computing for accelerators 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.