HPC Patents: Who Leads, Where the Gaps Are 2026
- 33.2% concentration. The five most active filers account for 14,409 of the 43,368 records in scope, a level of concentration that makes clearance searches around the leaders' filings a first, not optional, step.
- Growth cooled, not stalled. Filings ran 2021 = 1,793 down to 2024 = 1,405, a 22% pull-back over that three-year span, after a peak of 1,814 in 2022 — the most recent complete years, since publication lags filing by about 18 months.
- Core compute still dominates. G06F (electric digital data processing) touches 28.8% of all records, more than triple the next largest class, H04L at 7.4% — most of the field's activity is still in the data-processing core, not the surrounding transport or AI layers.
Filing growth compares 2021 (1,793 records) with 2024 (1,405) — 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 43,368 records in scope (CR5), not by the ranked leaders only.
What the high-performance computing patent record shows
High-performance computing spans processor architecture, cache coherence, memory access and parallel task scheduling — the plumbing that lets many cores or nodes work on one problem. The dataset behind this page pulls 43,368 published records filed against search terms covering HPC systems, parallel computing and the low-level mechanisms (cache coherence, compute kernels, memory access, task and parallel processing) that make them work. Coverage runs from 2015 through the current data cut-off in 2026, with the most recent year necessarily partial.
Filing activity is concentrated among a small number of large compute and storage vendors, with a long tail of single- or few-filing entrants behind them. The technology composition skews heavily toward core digital data processing, with transport, imaging and AI-adjacent classes present but smaller by comparison.
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 43,368-record corpus: how filing volume has moved year over year, and which IPC subclasses the records fall into.
Filing trend, 2017–2026
Annual filings rose from 893 in 2017 to a peak of 1,814 in 2022, then eased to 175 in the partial 2026 year. Treat the last one to two years as undercounted rather than as evidence of a slowdown — publication lags filing by roughly 18 months.
Technology composition by IPC subclass
G06F (electric digital data processing) covers 28.8% of the 43,368 records, well ahead of H04L (digital information transmission, 7.4%), G06T (image data processing, 6.5%) and G06N (AI-based computing, 6.0%). Because a record can carry several IPC classes, these shares add up to more than 100% of the record total — they are not mutually exclusive slices of one pie.
Shares are the percentage of the 43,368 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Go deeper on High-Performance Computing Patent Landscape with Eureka
This page is one run against one query. Ask Eureka your own question about high-performance computing patent landscape and every answer comes back with the patent numbers behind it.
Try EurekaRepresentative filing and most-cited records
Caching Network Fabric for High Performance Computing
An apparatus and method exchange data between two nodes of a high performance computing (HPC) system using a data communication link. The apparatus has one or more processing cores, RDMA engines, cache coherence engines, and multiplexers, programmable through an API to selectively couple RDMA engines, cache coherence engines, or a mix of these to the link. Bulk data transfer uses paged RDMA during initialisation; during computation, remote data access uses a coherence protocol such as MESI operating on much smaller transfer units.Filed by Hewlett Packard Enterprise Development LP, published 2018-01-18 as US20180020054A1.
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US20170006135A1 | Systems, methods, and devices for an enterprise internet-of-things application development platform | 1,905 |
| 2 | US5485474A | Scheme for information dispersal and reconstruction | 1,798 |
| 3 | US20190201140A1 | Surgical HUB situational awareness | 1,784 |
| 4 | US20190200844A1 | Method of HUB communication, processing, storage and display | 1,718 |
| 5 | US20190201136A1 | Method of HUB communication | 1,635 |
| 6 | US20190201115A1 | Aggregation and reporting of surgical HUB data | 1,549 |
| 7 | US20190201046A1 | Method for controlling smart energy devices | 1,448 |
| 8 | US20190206562A1 | Method of HUB communication, processing, display, and cloud analytics | 1,350 |
| 9 | US20190201033A1 | Surgical system distributed processing | 1,337 |
| 10 | US20190125361A1 | Method for operating a powered articulating multi-clip applier | 1,281 |
Citation counts inside a searched corpus favour older filings that have had more years to accumulate citations — read them as a signal of influence, not of current technical importance.
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 mean for a filing decision
Three read-outs from the same corpus, each pointing at a different practical question: where the field is crowded, where activity is cooling, and where the claim space still has room.
The core is crowded at the top
Fourteen thousand four hundred and nine records sit with just five assignees. A new filing that touches core cache-coherence or memory-access mechanisms is filing into territory these five have already claimed heavily; freedom-to-operate work should start there, not with a broad novelty search.
Growth cooled from a 2022 peak
Filings peaked at 1,814 in 2022 and eased to 1,405 by 2024 — the most recent year that can be read as complete. That is a pull-back from a peak, not a collapse, and the 2025–2026 figures are still filling in given the usual 18-month publication lag.
Activity still centres on core data processing
G06F outweighs every other subclass by a wide margin, with H04L, G06T and G06N each in single digits. That imbalance suggests the transport, imaging and AI-adjacent layers around HPC are comparatively open relative to the processing core itself.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to high-performance computing patent landscape, with the prior art for and against each one.
| Assignee | Co-assignee | Shared families |
|---|---|---|
| OnLive, Inc. | PERLMAN STEPHEN G | 36 |
| OnLive, Inc. | VAN DER LAAN ROGER | 33 |
| International Business Machines Corp (IBM) | IBM UNITED KINGDOM LTD INTELLECTUAL PROPERTY DEPARTMENT | 27 |
| OnLive, Inc. | Altour, Inc. | 13 |
| Citrix Systems Inc | CHOUDHARY AKSHAT | 12 |
| Citrix Systems Inc | SUGANTHI JOSEPHINE | 11 |
| Citrix Systems Inc | RAJA MURALI | 11 |
| OnLive, Inc. | FURMAN SCOTT | 10 |
Co-assignee pairings in this corpus are few — ten pairs recorded overall — with the strongest pairings linking individual inventors to their assignee corporation rather than joint filings between competing companies.
Who is filing, and where the activity is slowing
The ranked leaders are large compute, storage and enterprise software vendors. Recent-year momentum figures show most of them pulling back from prior-year filing volumes, consistent with the corpus-wide 2021–2024 pull-back and the 18-month publication lag understating the very latest year.
A single dominant filer
The top-ranked assignee holds 6,370 records, well ahead of fifth place at 1,214 and tenth place at 674 — a steep drop-off rather than a gradual one.
Even the leader is filing less this year
The top filer recorded 114 filings in the latest year, down 76% year over year — consistent with the broader corpus pattern of fewer newly published records in the most recent, still-incomplete year rather than a genuine drop in R&D activity.
A wide field beyond the leaders
Just over half of all records sit outside the ten most active filers, spread across the remaining ranked companies and unranked entrants — a long tail of narrower, more opportunistic filings rather than broad platform claims.
| Assignee | Recent year | YoY |
|---|---|---|
| NVIDIA Corp | 114 | -76% |
| Pure Storage Inc | 4 | -64% |
| Huawei Technologies Co., Ltd. | 3 | -67% |
| Intel Corp | 2 | -93% |
| Oracle International Corp | 2 | -50% |
| Micron Technology Inc | 1 | -91% |
| International Business Machines Corp (IBM) | 0 | -100% |
| Citrix Systems Inc | 0 | — |
Where to take this
The dataset points at where claim space is dense and where it thins out. The next step is turning that into a specific freedom-to-operate or whitespace question.
Map the leader's claim boundaries
Start clearance work against the top-ranked assignee's core cache-coherence and memory-access filings before drafting new claims in that territory.
Explore assignee filings in Eureka →Test a white-space claim
Draft a first claim around one of the under-claimed sub-areas above and check it against the existing IPC-classified prior art before committing to a full application.
Run a claim check in Eureka →Track momentum, not just totals
Recent-year YoY figures for individual assignees move faster than the corpus-wide trend — worth monitoring quarterly rather than relying on the annual view alone.
Set up assignee tracking in Eureka →Questions about the HPC patent landscape
The assignee ranking in this dataset is led by a single company holding 6,370 of the 43,368 records in scope, with a steep drop to fifth place at 1,214 and tenth place at 674. The top five combined account for 33.2% of all records, so the field is concentrated at the top with a long tail of smaller filers behind it. Any competitive analysis should treat the leader's portfolio as the first thing to map, since it covers a disproportionate share of the core claim space.
Filings peaked at 1,814 in 2022 and eased to 1,405 by 2024, a 22% pull-back over that three-year span using the most recent years that can be treated as complete. That reads as a cooling from a peak rather than a collapse. The 2025 and 2026 figures in the raw trend look lower still, but publication typically lags filing by about 18 months, so those years are undercounted rather than genuinely slower.
Electric digital data processing (IPC class G06F) is by far the largest category, touching 28.8% of the 43,368 records in scope. Digital information transmission (H04L, 7.4%), image data processing (G06T, 6.5%) and AI-based computing (G06N, 6.0%) follow at much lower density. Because records can carry multiple IPC classes, these are overlapping shares of the same record set, not slices that sum to 100%, but the gap between G06F and everything else is wide enough to show where the field's filing weight actually sits.
Based on the technology composition, classes adjacent to the dominant G06F core — such as image/video recognition (G06V), pictorial communication (H04N) and static memory architectures (G11C) — carry noticeably lower filing density. That does not guarantee an easy grant, but it does suggest less-crowded claim space around cross-node coherence protocols beyond MESI, heterogeneous memory scheduling, and compute-kernel offload for imaging pipelines. A freedom-to-operate search focused on those specific mechanisms, rather than HPC broadly, is the more productive next step.
The most-cited record in this corpus is US20170006135A1, an enterprise IoT application development platform filing, cited 1,905 times, followed closely by other frequently-cited records including a 1990s information dispersal patent and several hub-communication filings. High citation counts favour older filings that have simply had more years to accumulate citations within the searched corpus, so treat them as a signal of historical influence rather than of current technical importance or of what is being actively litigated today.
Research High-Performance Computing Patent Landscape in depth with Eureka
Go past this page: query the whole high-performance computing patent landscape 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.