Photonic Computing Patents: Who Leads, Where the Gaps Are 2026
- Filings have already peaked. The 2025 high of 88 families sits well above the 2022 midpoint of 66, but growth since then is flat rather than accelerating.
- China's receiving office leads by a wide margin. 206 filings there versus 126 in the United States, with WIPO PCT and EPO filings well behind — a strong signal of where enforcement and freedom-to-operate checks matter most.
- The foundational claims are three decades old. The most-cited record, an optical neural network apparatus, dates to a 1990s priority chain, and today's filers are still building around that architecture rather than replacing it.
What the photonic computing patent record actually shows
Photonic computing covers hardware that performs computation — usually matrix multiplication for neural network inference — using light rather than electrons. The dataset behind this page pulls 548 patent families published between 2015 and mid-2026 that combine optical-computing terminology in the title or abstract with classification codes for optical computing, AI-model computation, and optical control hardware. That triangulation is deliberate: optical neural network as a phrase spans both decades-old opto-electronic prototypes and current silicon-photonic accelerator chips, so the classification filter is what separates a genuine hardware claim from a software patent that merely mentions light.
Because publication lags filing by roughly 18 months, the 2026 count in any trend chart understates real filing activity for that year — treat the last one or two bars 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 548 families: how filing activity has moved year over year, and which IPC subclasses carry the underlying claims.
A decade of growth that has already plateaued
Annual filings rose from 9 in 2017 to a peak of 88 in 2025. The climb from the 2022 midpoint of 66 to that peak is real, but it is not steep — this is a technology moving from early demonstration into a settled filing rate, not one accelerating into a boom.
AI-model computation dominates the classification mix
G06N (computing based on AI models) appears in 497 of 548 families, far ahead of G06E (optical computing proper, 148) and G02F (optical control and modulation, 139). That gap says most filers are claiming the neural network architecture and training method first, with the optical implementation as a supporting element rather than the primary inventive claim — a pattern worth checking before assuming a hardware-first competitor exists.
Shares are the percentage of the 548 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Go deeper on Photonic Computing Technology Landscape with Eureka
This page is one run against one query. Ask Eureka your own question about photonic computing technology landscape and every answer comes back with the patent numbers behind it.
Try EurekaThe patents that anchor this field
Optical neural network apparatus using primary processing (US5080464A)
For inputting a two-dimensional image into an optical neural network apparatus, a primary processing device extracts the characteristic feature of an object pattern. Compressed information from that processing is then fed into an all-optical neural network apparatus that performs parallel processing adaptively through optical computing at individual points on its input. The primary processing device, previously limited to logical input information, can therefore handle even vague input by pairing it with the downstream optical neural network stage.Filed by Hamamatsu Photonics, granted 1992 — one of the earliest hardware-level optical neural network filings still active in citation chains today.
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US20170351293A1 | Apparatus and Methods for Optical Neural Network | 260 |
| 2 | US10268232B2 | Apparatus and methods for optical neural network | 112 |
| 3 | US5095459A | Optical neural network | 98 |
| 4 | US7847225B2 | Optical neural network | 84 |
| 5 | WO2020102204A1 | Large-scale artificial neural-network accelerators based on coherent detection and optical data fan-out | 63 |
| 6 | US20190294199A1 | Apparatus and methods for optical neural network | 63 |
| 7 | US20210064958A1 | Heterogeneously integrated optical neural network accelerator | 61 |
| 8 | US20210357737A1 | Large-Scale Artificial Neural-Network Accelerators Based on Coherent Detection and Optical Data Fan-Out | 54 |
| 9 | US20210133547A1 | High efficiency optical neural network | 53 |
| 10 | CN110197277A | 实现数字识别的光学神经网络方法 | 52 |
Citation counts favour older records simply because they have had longer to accumulate them — read this table as a map of influence, not of current commercial relevance.
Patent titles are shown in the language they were filed in, not translated, so that each record stays verifiable against the original filing — a translated title will not match in Eureka or in any national register. 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 patterns stand out once family counts, classification codes and receiving offices are read together.
Growth has flattened, not stopped
The rise from 66 families at the 2022 midpoint to 88 in 2025 is a real increase, but the shape is closer to a maturing filing rate than a technology in early exponential growth. New entrants should expect a crowded, established claim landscape rather than open ground.
China is the primary filing venue
China's receiving office accounts for 206 of the tracked families against 126 for the United States, with Europe, WIPO and Japan trailing further behind. Any freedom-to-operate check that stops at the USPTO and EPO is missing the largest single filing venue in this field.
Software-adjacent claims outnumber pure optics claims
G06N (AI-model computation) appears in nearly all families, while G06E (optical computing itself) appears in only 148. Most applicants are protecting the neural network method and training approach, with the optical substrate as an enabling detail — a gap that leaves room for claims written the other way round.
Historic leaders have gone quiet
Several of the assignees with the deepest filing history, including major research universities, show zero filings in the latest year and a -100% year-on-year change. Recent activity is concentrated among a narrower, more current set of filers than a cumulative ranking alone would suggest.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to photonic computing technology landscape, with the prior art for and against each one.
Who is filing, and who has slowed down
Cumulative rankings and recent-year momentum tell different stories in this dataset — several long-standing filers show no activity in the latest tracked year.
Tsinghua University and MIT have paused
Both Tsinghua University and MIT — historically among the more active filers in optical neural network research — show zero filings in the latest year and a -100% year-on-year change, alongside NTT and IBM. That does not mean the research has stopped; it more likely reflects the 18-month publication lag catching up with recent work.
University of California remains active
The Regents of the University of California recorded a filing in the latest tracked year, making it one of the few historically significant assignees still visibly active as the data cut-off approaches.
Co-filing is limited and university-anchored
Only 10 co-assignee pairs appear across the corpus, the strongest linking Tsinghua University with two separate Beijing-based optics firms, and MIT with an individual named co-inventor. Collaboration is the exception here, not the norm — most families are filed by a single assignee.
| Assignee | Recent year | YoY |
|---|---|---|
| The Regents of the University of California | 1 | — |
| Fujikura Ltd. | 0 | — |
| Tsinghua University | 0 | -100% |
| Massachusetts Institute of Technology (MIT) | 0 | -100% |
| Nippon Telegraph and Telephone Corporation (NTT) | 0 | — |
| International Business Machines Corporation (IBM) | 0 | — |
| Huawei Technologies Co., Ltd. | 0 | — |
| Intel Corporation | 0 | — |
Where to take this analysis
The dataset points to specific follow-up checks rather than a single conclusion.
Check freedom-to-operate against China filings first
With 206 of 548 families filed through China's receiving office against 126 in the US, a freedom-to-operate review anchored only in US and EPO records will miss the largest filing venue for this technology.
Run a China-anchored FTO searchRe-read the 1990s-era anchor claims before drafting new ones
The most-cited records in this corpus trace back to 1990s optical neural network architectures. Any new filing in this space should be checked against those foundational claims, not just against recent competitors.
Review the featured patent in detailWatch for a lag-driven rebound in 2026 filings
Several historically active assignees show zero filings in the latest year, which is consistent with the normal 18-month publication lag rather than a genuine exit from the field. Re-check momentum once later data lands.
Track assignee momentum over timeCommon questions about photonic computing patents
This dataset identifies 548 patent families published between 2015 and mid-2026 that combine optical-computing language in the title or abstract with classification codes covering optical computing, AI-model computation and optical control hardware. That figure counts families, not raw document filings, so continuations and multiple-jurisdiction copies of the same invention are not double-counted. Because publication lags filing by roughly 18 months, the true total for the most recent one to two years will rise as later data arrives.
Cumulative filing history in this corpus includes major research universities and large electronics manufacturers alongside specialist optics firms, but cumulative rank and recent activity diverge sharply here. Several long-standing filers, including Tsinghua University and MIT, show zero filings in the latest tracked year, while the University of California recorded activity in that same period. Any 'who leads' answer needs to specify whether it means all-time filings or current momentum, because this dataset gives different answers to each.
China's receiving office accounts for the largest share of filings tracked here, at 206 families, ahead of the United States at 126, the European Patent Office at 74, WIPO's PCT route at 58, Japan at 49 and South Korea at 10. A competitive or freedom-to-operate review that only checks US and European filings will miss the single largest filing venue for this technology. This pattern is worth confirming against your own product's target markets before drawing conclusions.
Filings rose from 9 in 2017 to a peak of 88 in 2025, with a 2022 midpoint of 66 — real growth, but the pace from 2022 to the 2025 peak is closer to steady expansion than acceleration. Because the 2026 figure is still partial due to publication lag, it is too early to call a downturn, but the trend line does not show the technology in an early exponential phase anymore. Treat this as a maturing filing rate rather than an emerging one.
The most heavily cited record in this corpus is a 1990s-era optical neural network apparatus filing, and a related Hamamatsu Photonics filing, US5080464A, describes an optical neural network apparatus using primary processing, granted in 1992. These older filings still anchor citation chains for current filings, meaning new applicants should check their claims against this foundational architecture rather than assuming it has aged out of relevance. Citation weight in an older record reflects accumulated influence over time, not necessarily current commercial dominance.
Research Photonic Computing Technology Landscape in depth with Eureka
Go past this page: query the whole photonic computing technology 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.