Skip to main content

Water Electrolysis — Patent Landscape 2000–2024

Patent landscape of water electrolysis and electrolytic hydrogen production — 33,424 DOCDB patent families filed at 75 offices worldwide, mapping the shift from a small electrochemical engineering field to a China-centred research front driven by universities and state energy groups.

Data: EPO PATSTAT Global, Spring 2026 (patstat_2026a) Period: 2000–2024 Created: March 2026 Author: mtc.berlin
33,424
Patent Families
2000–2024
75
Filing Offices
62,869 applications
17.6%
CAGR 2000–2022
122 → 4,354 families/yr
61.8%
Grant Rate
2000–2020 cohort
17.4%
Co-Filing Rate
5,824 co-filed families

Executive Summary

Water electrolysis and electrolytic hydrogen production, 2000–2024, at DOCDB family level.

The search profile behind this report — the CPC/IPC group C25B 1/02 (electrolytic production of hydrogen or oxygen) with its subtree, the photoelectrolysis group C25B 1/55, and the four groups retired in the 2021 edition that fed into C25B 1/04 — yields 33,424 DOCDB patent families with an earliest filing year between 2000 and 2024, covering 62,869 individual applications at 75 patent offices.

Filing activity rose from 122 families in 2000 to 4,354 families in 2022 — a factor of 35.7 and a compound annual growth rate of 17.6% over those 22 years. The curve is not smooth: annual filings grew by 47.6% between 2015 (701) and 2016 (1,035), and the 2015–2022 stretch alone runs at a 29.8% CAGR. Two thirds of the whole 25-year corpus — 21,738 families, or 65.0% — carries an earliest filing year of 2020 or later. The 2023 and 2024 counts (5,927 and 6,065) are provisional and will rise as publication catches up, so no growth figure in this report ends in those two years.

Geographically the field has one centre. 21,134 families (63.2%) include a filing at CNIPA, ahead of international PCT filings at WIPO (6,887; 20.6%), the United States (6,204; 18.6%), Japan (5,405), Korea (4,349) and the EPO (4,336). The Chinese share is a recent phenomenon: 524 families with an earliest filing year in 2000–2009 against 15,768 in 2020–2024.

By applicant sector the corpus splits almost evenly between companies (13,181 families; 39.4%) and universities (12,174; 36.4%). Read year by year, universities have filed more families than companies in every filing year from 2018 onwards except 2023, where the two are level (2,151 company against 2,146 university families); in 2024 the gap is 2,537 to 1,891. Technologically the corpus is concentrated in component work: 14,790 families touch electrodes (C25B 11/00) and 13,603 cell construction (C25B 9/00), against 1,252 families in high-temperature steam electrolysis (C25B 1/042) and 2,060 in photoelectrolysis (C25B 1/55).

Robert Bosch GmbH holds the largest single harmonised portfolio with 376 families, followed by the Dalian Institute of Chemical Physics (260), Honda Motor (255) and the Huaneng Clean Energy Research Institute (247). Eleven of the twenty largest names are Chinese universities. Measured as families per year, the steepest climb between 2016–2019 and 2020–2024 belongs to Robert Bosch (9.5×), Tsinghua University (8.5×) and Siemens Energy Global (8.1×).

Filing Trend 2000–2024

DOCDB patent families by earliest filing year. Each family is counted once, in the year of its earliest application, so the yearly values sum exactly to the 33,424 families in scope.

Between 2000 and 2022 annual filings rose from 122 to 4,354 families, a compound annual growth rate of 17.6% over 22 intervals. The two largest year-on-year steps in the series are 2020 to 2021 (2,141 → 3,251, +51.8%) and 2015 to 2016 (701 → 1,035, +47.6%). Growth is continuous rather than event-driven: from 2016 onwards every year exceeds the one before it, which is the pattern the data supports; the report does not attribute this to any specific policy or market cause.

* 2023 and 2024 are provisional. Applications are published roughly 18 months after filing and reach PATSTAT later still, so the two most recent filing years in the Spring 2026 (patstat_2026a) edition are systematically under-counted and will rise. No growth rate in this report ends in a provisional year.

Data table: Patent families by earliest filing year, 2000–2024
Filing Year Patent Families Year-on-Year Change
2000122
2001148+21.3%
2002171+15.5%
2003217+26.9%
2004242+11.5%
2005267+10.3%
2006248−7.1%
2007320+29.0%
2008368+15.0%
2009365−0.8%
2010479+31.2%
2011471−1.7%
2012534+13.4%
2013543+1.7%
2014617+13.6%
2015701+13.6%
20161,035+47.6%
20171,238+19.6%
20181,646+33.0%
20191,954+18.7%
20202,141+9.6%
20213,251+51.8%
20224,354+33.9%
2023*5,927+36.1%*
2024*6,065+2.3%*
Total33,424

Top Applicants

The twenty largest harmonised applicant names (han_name) by DOCDB families in scope, 2000–2024. Applicants only (applt_seq_nr > 0); inventors are excluded.

# Harmonised Applicant Name Sector Families Filing Years
1ROBERT BOSCH GMBHCompany3762000–2024
2DALIAN INSTITUTE OF CHEMICAL PHYSICS, CHINESE ACADEMY OF SCIENCESUniversity2602014–2024
3HONDA MOTOR CO LTDCompany2552000–2024
4HUANENG CLEAN ENERGY RESEARCH INSTITUTECompany2472015–2024
5TOSHIBA CORPCompany2222000–2024
6SHAANXI UNIVERSITY OF SCIENCE AND TECHNOLOGYUniversity2112017–2024
7BEIJING UNIVERSITY OF CHEMICAL TECHNOLOGYUniversity1932010–2024
8MATSUSHITA ELECT IND CO LTDCompany1922004–2024
9ZHEJIANG UNIVERSITYUniversity1772016–2024
10CEA & AUX ENERGIES ALTERNATIVESGovernment / Non-Profit1722005–2024
11CO LTD TOSHIBACompany1702007–2024
12QINGDAO UNIVERSITY OF SCIENCE & TECHNOLOGYUniversity1692016–2024
12KR ADVANCED INSTITUTE OF SCIENCE & TECHGovernment / Non-Profit1692004–2024
14SIEMENS ENERGY GLOBAL GMBH & CO KGCompany1672015–2024
15JILIN UNIVERSITYUniversity1662013–2024
16TSINGHUA UNIVERSITYUniversity1622017–2024
17TIANJIN UNIVERSITYUniversity1592002–2024
18SOUTH CHINA UNIVERSITY OF TECHNOLOGYUniversity1542010–2024
19JIANGSU UNIVERSITYUniversity1412010–2024
19DALIAN UNIVERSITY OF TECHNOLOGYUniversity1412011–2024

Eleven of the twenty largest names carry PATSTAT's University sector, and all eleven are Chinese institutions; two more — CEA in France and KAIST in Korea — are public research organisations. The industrial names at the top divide into two groups by filing history: long-standing filers active across the whole window (Robert Bosch 2000–2024, Honda Motor 2000–2024, Toshiba 2000–2024) and entrants whose first family in scope falls in 2015 or later (Huaneng Clean Energy Research Institute 2015, Siemens Energy Global 2015). Among the academic names, Tsinghua University, Shaanxi University of Science and Technology and Qingdao University of Science & Technology all enter the field from 2016–2017 onwards.

Family counts run through filing year 2024. The two most recent filing years, 2023 and 2024, are provisional because of the roughly 18-month publication lag, so every portfolio figure in this section is a lower bound — most of all for the applicants whose activity starts after 2020.

Data Quality Note — Applicant Name Harmonisation

This ranking uses PATSTAT's harmonised applicant name (han_name) exactly as stored. That harmonisation does not merge corporate groups, legal-form variants, transliterations or renamings, and the query output shows several portfolios split across multiple entries:

  • Toshiba appears as TOSHIBA CORP (222), CO LTD TOSHIBA (170), TOSHIBA ENERGY SYSTEM&SOLUTION CORP (84), TOSHIBA ENERGY SYSTEMS & SOLUTIONS CORP (54) and KABUSHIKI KAISHA TOSHIBA (47).
  • Panasonic appears as MATSUSHITA ELECT IND CO LTD (192), パナソニックIPマネジメント株式会社 (128), PANASONIC IP MANAGEMENT CORP (124) and PANASONIC INTELLECTUAL PROPERTY MANAGEMENT CO., LTD. (81).
  • Honda appears as HONDA MOTOR CO LTD (255) and Honda Motor Co., Ltd. (90); Hyundai as HYUNDAI MOTOR CO (94) and HYUNDAI MOTOR COMPANY (47); Topsoe as TOPSOE AS (73), HALDOR TOPSØE AS (45) and HALDOR TOPSOE AS (42).

These counts must not be added together: a family co-filed under two variants would be counted twice. Consolidated portfolios for these groups are therefore larger than any single row above, but the correct consolidated figure requires a variant-merging step that is not part of this report. Sector labels come from PATSTAT's psn_sector; note that it classifies Chinese Academy of Sciences institutes such as the Dalian Institute of Chemical Physics as UNIVERSITY rather than as a public research organisation.

Applicant country (person_ctry_code) is blank for a large share of the Chinese applicant records in this scope, so an applicant-origin breakdown is not reported here. The geography section works from filing authority (appln_auth) instead, which is a market question rather than an origin question.

Technology Sub-Areas

Where in subclass C25B the families in scope are classified. Counts are IPC-based and aggregated across all members of each family, at main-group level.

Beyond the scope group itself, the corpus is concentrated in hardware rather than process chemistry: 14,790 families (44.2%) carry an electrode group (C25B 11/00) and 13,603 (40.7%) a cell-construction group (C25B 9/00), against 8,231 (24.6%) for operating and servicing of cells and 2,538 (7.6%) for diaphragms and spacing elements. A family typically carries several of these groups at once, so the figures are shares of the 33,424 families in scope and not parts of a whole.

Data table: C25B main groups within the scope
IPC Main Group Title Families Share of 33,424
C25B 1/00Electrolytic production of inorganic compounds or non-metals31,00792.8%
C25B 11/00Electrodes; manufacture thereof not otherwise provided for14,79044.2%
C25B 9/00Cells or assemblies of cells; constructional parts of cells13,60340.7%
C25B 15/00Operating or servicing cells8,23124.6%
C25B 13/00Diaphragms; spacing elements2,5387.6%
C25B 3/00Electrolytic production of organic compounds1,2573.8%
C25B 5/00Electrogenerative processes2730.8%
C25B 7/00Electrophoretic production of compounds or non-metals100.0%

Inside the Scope Definition

The nine classification symbols that define the scope are not equally weighted. Families are counted once per symbol they carry in either scheme, so the column below sums above 33,424.

Symbol Title in the current scheme Families Share of 33,424
C25B 1/04by electrolysis of water28,60285.6%
C25B 1/02Hydrogen or oxygen (parent group)4,05212.1%
C25B 1/55Photoelectrolysis2,0606.2%
C25B 1/042by electrolysis of steam1,2523.7%
C25B 1/10retired in the 2021 edition — no current title8952.7%
C25B 1/044producing mixed hydrogen and oxygen gas, e.g. Brown's gas4691.4%
C25B 1/06retired in the 2021 edition — no current title3421.0%
C25B 1/12retired in the 2021 edition — no current title2110.6%
C25B 1/08retired in the 2021 edition — no current title1020.3%

High-temperature steam electrolysis (C25B 1/042) accounts for 1,252 families and photoelectrolysis (C25B 1/55) for 2,060 — 3.7% and 6.2% of the corpus respectively. Both symbols were created in the 2021 edition, so their counts include only what has been classified or reclassified into them so far and are a lower bound on the underlying research activity. The four retired symbols still carry 1,464 families between them; 567 of those carry no current symbol at all and would be absent from a code list built only from today's scheme.

A family can be assigned to several classification symbols and to several main groups at once. Column totals in this section therefore exceed the distinct family count of 33,424 and must not be read as parts of a whole. All counts run through filing year 2024 and include the two provisional years 2023 and 2024, so each is a lower bound. C25B 7/00 (10 families) is listed in the data table but omitted from the chart, where it would not be legible.

Geographic Analysis

Where the families in scope are filed. Both cuts use the filing authority (appln_auth), which answers a market question — not where the applicant is based.

21,134 families (63.2%) include a filing at CNIPA, more than three times the next entry. International PCT filings account for 6,887 families (20.6%), the United States for 6,204 (18.6%), Japan for 5,405 (16.2%), Korea for 4,349 and the EPO for 4,336 (13.0% each). Filings are spread across 75 offices in total.

The period split shows how recent the Chinese concentration is: 524 families with an earliest filing year in 2000–2009 against 15,768 in 2020–2024, a factor of 30. Over the same two periods the United States moves from 908 to 2,787 (3.1×), Japan from 960 to 2,393 (2.5×) and the EPO from 447 to 2,325 (5.2×). Every office in the table grows; the difference is one of magnitude, not direction.

Data table: Top filing offices
Office Patent Families Share of 33,424
CN — China (CNIPA)21,13463.2%
WO — International (PCT)6,88720.6%
US — United States6,20418.6%
JP — Japan5,40516.2%
KR — Korea4,34913.0%
EP — European Patent Office4,33613.0%
AU — Australia1,5784.7%
DE — Germany1,5594.7%
CA — Canada1,5394.6%
TW — Taiwan7272.2%
Data table: Geographic shift by period
Office 2000–2009 2010–2015 2016–2019 2020–2024* Total
CN5241,1283,71415,76821,134
WO6561,2271,1843,8206,887
US9081,2491,2602,7876,204
JP9609631,0892,3935,405
KR5106528562,3314,349
EP4477328322,3254,336
DE1802432408961,559

* The 2020–2024 band contains the two provisional filing years 2023 and 2024 and is therefore a lower bound for every office. Families are assigned to a period by their earliest filing year; a family filed at several offices is counted at each of them, so the column values are non-exclusive across offices and sum above 33,424.

Sector Analysis

Applicant sectors as classified by PATSTAT's psn_sector, at DOCDB family level.

Sector Families Share of 33,424
Company13,18139.4%
University12,17436.4%
Unknown / Unclassified7,88623.6%
Individual3,80411.4%
Government / Non-Profit1,5544.6%
Hospital40.0%

A family with several applicants can belong to several sectors. The six values sum to 38,603 against 33,424 distinct families, so these are shares of the corpus rather than parts of a whole. This is why the breakdown is shown as a table and a line series, not as a pie chart.

Sector Development 2000–2024

University filings exceed company filings in every filing year from 2018 onwards except 2023, where the two are effectively level (2,151 company against 2,146 university families). The crossover year is 2018 (621 company, 820 university); in 2024 the two series stand at 1,891 and 2,537. Individual applicants follow a different curve entirely: they peak at 320 families in 2010 and stand at 53 in 2024 — the only one of the six sectors whose annual count is lower in 2024 than it was in 2010.

* 2023 and 2024 are provisional filing years and will rise. The 2024 company value is below the 2023 value in the current data; that gap is consistent with publication lag and is not read here as a decline.

Data table: Families per sector by earliest filing year
Filing Year Company University Individual
200083445
200194251
2002112470
20031431080
200415812107
200516822131
200612725158
200718135192
200819737261
200919744259
201028663320
201124984286
201229989233
201332790133
2014341136143
2015381182151
2016462333212
2017566507149
2018621820156
20196821,072132
20208081,043125
20211,1591,273142
20221,4981,604103
2023*2,1512,146112
2024*1,8912,53753

Data Quality Note — Unclassified Applicants

7,886 families (23.6%) have at least one applicant whose psn_sector is empty or unknown. That bucket is not evenly spread over time: it holds 11 families in filing year 2000 and 2,034 in 2024, and it therefore grows faster than any named sector. Chinese state energy entities and newly registered hydrogen companies appear repeatedly among the unclassified names in the applicant ranking. Sector shares in this section should be read as a lower bound for the named sectors.

Growth Dynamics

Change in filing rate between 2016–2019 (four filing years) and 2020–2024 (five filing years). Because the two windows differ in length, the comparison is made on families per year, not on raw totals.

Robert Bosch multiplied its annual filing rate by 9.5, from 6.75 families per year in 2016–2019 to 64.4 in 2020–2024 — the steepest climb among applicants with a comparable base, ahead of Tsinghua University (8.5×) and Siemens Energy Global (8.1×). Below those three the field is uniformly academic and Chinese, with rate multiples between 3.4 and 4.4. Because 2023 and 2024 are provisional, every multiple in this table is a lower bound.

Data table: Filing rate change 2016–2019 to 2020–2024
Applicant (han_name) Families 2016–2019 Families 2020–2024* Per year, earlier Per year, later Rate multiple
ROBERT BOSCH GMBH273226.7564.49.5×
TSINGHUA UNIVERSITY141483.5029.68.5×
SIEMENS ENERGY GLOBAL GMBH & CO KG151513.7530.28.1×
HUAZHONG UNIVERSITY OF SCIENCE AND TECHNOLOGY13723.2514.44.4×
CHINA UNIVERSITY OF PETROLEUM (EAST CHINA)12663.0013.24.4×
XI'AN JIAOTONG UNIVERSITY17894.2517.84.2×
JILIN UNIVERSITY261346.5026.84.1×
CHANGZHOU UNIVERSITY18894.5017.84.0×
BEIJING UNIVERSITY OF CHEMICAL TECHNOLOGY341538.5030.63.6×
UNIVERSITY OF SCIENCE AND TECHNOLOGY OF CHINA21915.2518.23.5×
TONGJI UNIVERSITY22955.5019.03.5×
QINGDAO UNIVERSITY OF SCIENCE & TECHNOLOGY321378.0027.43.4×

Entrants Without a Comparable Base

A second group of applicants had fewer than ten families in 2016–2019, so a rate multiple would be arithmetically unstable. Their absolute counts are reported instead:

  • Huaneng Clean Energy Research Institute — 9 families in 2016–2019, 235 in 2020–2024
  • China Petroleum & Chemical Corporation — 8, then 118
  • PetroChina Company Limited — 2, then 107
  • Huaneng Group Technology Innovation Center — 1, then 114
  • North China Electric Power University — 1, then 70
  • Linde GmbH — 8, then 57
  • Schaeffler Technologies AG & Co KG — 7, then 59

* The 2020–2024 window contains the two provisional filing years 2023 and 2024, so both the later counts and the resulting multiples understate the true change. Applicants are counted by harmonised name; the fragmentation described in the Top Applicants section applies here too, which is why several large groups do not appear in this table under a single entry.

Collaboration Network

Co-filing between applicants. A family counts as co-filed when at least one of its applications carries more than one applicant.

5,824 of 33,424 families (17.4%) are co-filed. The co-applicant pairs behind that figure are not, for the most part, collaborations between independent organisations.

The largest co-filing cluster in the corpus is a set of nine applicant names that all contain "Huaneng": all 36 possible pairs among those nine names share 93 or more families, and the strongest pair — Huaneng Clean Energy Research Institute with Huaneng Group Technology Innovation Center — shares 112. A uniform pairwise count across a complete set of names is the signature of a block of applications filed jointly by the whole group rather than of a network of separate partnerships.

Pairs Across Different Applicant Names

Removing every pair whose two names share their leading word — which strips out the Huaneng, Toshiba, Mitsubishi and Sinopec self-pairings — leaves the following. Several of the remaining pairs still connect entities within one corporate group; that reading is an interpretation of the names, not a value returned by the query.

Applicant A Applicant B Shared Families
HYUNDAI MOTOR COKIA CORP78
CNPC SHENZHEN NEW ENERGY RESEARCH INSTITUTE CO., LTD.PETROCHINA COMPANY LIMITED76
HYUNDAI MOTOR COMPANYKIA MOTORS CORP.43
CHN ENERGY INVESTMENT GROUP CO., LTD.NATIONAL INSTITUTE OF CLEAN-AND-LOW-CARBON ENERGY37
ARAMCO SERVICES COSAUDI ARABIAN OIL CO33
GENVIASERVICES PETROLIERS SCHLUMBERGER27
CHINA PETROLEUM & CHEMICAL CORPORATIONSINOPEC RESEARCH INSTITUTE OF PETROLEUM PROCESSING25
FUJIFILM CORPJP TECHNOLOGICAL RESEARCH ASSOCIATION OF ARTIFICIAL PHOTOSYNTHESIS24
GENVIASCHLUMBERGER TECH CORP23
HONDA MOTOR CO LTDOkabe, Masanori20
DE NORA PERMELEC LTDNATIONAL UNIVERSITY CORP YOKOHAMA NATIONAL UNIVERSITY20
CHINA PETROLEUM & CHEMICAL CORPORATIONSINOPEC (BEIJING) CHEMICAL RESEARCH INSTITUTE CO., LTD.20

Two of the twelve entries pair a company with a university or a research association: Fujifilm with the Japanese Technological Research Association of Artificial Photosynthesis (24 families) and De Nora Permelec with Yokohama National University (20 families). A third, CHN Energy Investment Group with the National Institute of Clean-and-Low-Carbon Energy (37 families), pairs a company with an applicant PATSTAT classifies as a public research organisation. Every other entry links commercial entities. Against a corpus in which universities account for 12,174 families and companies for 13,181, cross-sector joint filings are rare at this scale. The data supports the observation; it does not explain it, and a low count of joint filings is not by itself evidence about the level of underlying research cooperation, which often runs through contract research rather than shared ownership of a patent.

Data Quality Note — Co-Applicant Pairs and Name Variants

  • Pairs are built inside a single application and deduplicated by ordering the two names alphabetically, so no applicant can appear as its own partner.
  • Because harmonised names are not merged across variants, a machine-translated duplicate of the same entity can surface as a partner. The query returns, for example, "HUA ENERGY GROUP TECHNOLOGY INNOVATION CENTER LIMITED COMPANY" and "CHINESE CHINA ENERGY GROUP CLEAN ENERGY TECHNOLOGY RESEARCH INSTITUTE" alongside the Huaneng and CHN Energy names — these are translation variants, not additional partners.
  • "Okabe, Masanori" is an individual named as co-applicant on Honda filings, not an organisation.

Grant Rate Analysis

Share of families with at least one granted application per office, restricted to the 2000–2020 filing cohort.

Across the whole corpus, 8,542 of the 13,827 families with an earliest filing year up to and including 2020 have a granted member — a grant rate of 61.8%. Families filed from 2021 onwards are excluded from every figure in this section: their examination is largely still running, and including them would measure examination speed rather than outcome.

China's grant rate of 57.5% sits below the United States (62.5%), Japan (61.4%) and Korea (67.2%) while carrying by far the largest cohort — 6,818 families against 3,867 at the USPTO. The EPO reaches 50.2% on 2,337 cohort families. Germany's 19.6% is the outlier in the table and reflects how the DPMA route is used rather than examination outcome: a German application can remain unexamined unless examination is requested separately, so a low ratio of granted members is expected there and should not be compared directly with the other offices.

Data table: Grant rate by office, 2000–2020 cohort
Office Cohort Families With Granted Member Grant Rate
TW — Taiwan42731473.5%
KR — Korea2,3581,58467.2%
AU — Australia74847163.0%
US — United States3,8672,41562.5%
JP — Japan3,3482,05561.4%
CN — China6,8183,92057.5%
EP — European Patent Office2,3371,17350.2%
CA — Canada98544745.4%
DE — Germany75014719.6%
All offices, whole corpus13,8278,54261.8%

How to Read These Rates

  • Pendency. granted = 'Y' means granted as of the Spring 2026 snapshot. Even inside the 2000–2020 cohort the most recent filing years are not fully decided, so the rates are a lower bound and lowest for the youngest part of the cohort.
  • WIPO is excluded. The 3,491 cohort families with a WO member show a 6.3% "grant rate", which is an artefact: a PCT application is not granted, and any grant follows in the national phase. Including WO in an office ranking would be misleading.
  • Unit of counting. A family counts as granted at an office if any of its applications at that office is granted. Office columns are therefore non-exclusive and do not sum to the corpus total.

Application Domains

Which other CPC subclasses the families in scope also carry. Labels come from the classification views, so the CPC-only Y sections appear alongside the ordinary technical subclasses.

After the electrolysis subclass itself, the largest co-occurring subclass is a tagging scheme rather than a technical field: 23,407 families (70.0% of the corpus) carry a code in Y02E — greenhouse-gas reduction related to energy generation, transmission or distribution — and a further 4,279 (12.8%) carry Y02P, the equivalent tag for production processes. The technical neighbours follow at a distance: electrochemical energy conversion (H01M, 4,262), catalysis and process apparatus (B01J, 3,643), non-metallic elements and their compounds (C01B, 2,887), water and waste-water treatment (C02F, 1,921) and nanostructures (B82Y, 1,711). Electric power networks (H02J, 1,216) is the smallest entry in the table and the one that marks the grid-coupling end of the field.

Data table: Co-occurring CPC subclasses
Subclass Title Families Share of 33,424
C25BElectrolytic or electrophoretic processes for the production of compounds or non-metals26,94880.6%
Y02EReduction of greenhouse gas emissions, related to energy generation, transmission or distribution23,40770.0%
Y02PClimate change mitigation technologies in the production or processing of goods4,27912.8%
H01MDirect conversion of chemical energy into electrical energy, e.g. batteries and fuel cells4,26212.8%
B01JChemical or physical processes, e.g. catalysis or colloid chemistry; relevant apparatus3,64310.9%
C01BNon-metallic elements; compounds thereof2,8878.6%
C02FTreatment of water, waste water, sewage or sludge1,9215.7%
B82YSpecific uses or applications of nanostructures1,7115.1%
B01DSeparation1,5814.7%
H02JElectric power networks; supply or distribution of electric power1,2163.6%

The Y Sections as a Search Layer

Y02 and Y04 are CPC-only tagging schemes with no IPC counterpart, applied across the whole classification rather than within one field. For a climate-technology subject they are the obvious second search layer, and this report measured what they contribute before deciding whether to use them for scope.

  • Volume. The group Y02E 60/36 — hydrogen production from non-carbon containing sources — together with its four retired subgroups matches 34,415 families with an earliest filing year in 2000–2024, more than the whole classification scope of this report.
  • Reach beyond the core. 17,449 of those families carry none of the nine C25B symbols in CPC. That is an upper bound on what the Y layer would add, not the addition itself: the Y codes exist only in CPC, so the comparison can only be made on the CPC side, and part of those 17,449 families carry a C25B symbol in IPC and are already inside this report's scope. Adding the Y layer would therefore have grown the corpus by at most two thirds.
  • Precision. A sample of 25 of those 17,449 families was read by title. 13 of 25 (52%) are not water electrolysis — photocatalytic water splitting without electrodes, covalent organic framework and carbon nitride photocatalysts, borane-based chemical hydrides, a biogas reactor and a hydrogen-enriched drinking-water device. That is what the group's own title says it covers: hydrogen from any non-carbon source, of which electrolysis is one example.
  • No finer cut is available. The four retired subgroups Y02E 60/362 to Y02E 60/368, which subdivided that group before 2020, carry zero assignments in the family-level CPC table on this edition, so there is no electrolysis-specific Y symbol left to search.

The Y sections are excluded from the search scope and reported as a measured layer instead. A profile whose additional half is half off-topic fails the one-in-five precision rule this report works to. What the Y codes do contribute is confirmation from the other direction: 70.0% of the families that the C25B profile selects already carry a Y02E tag, which is independent evidence that the corpus sits inside climate-relevant technology.

Subclass counts are non-exclusive — a family carrying Y02E also sits in its ordinary technical subclass — so the column sums far above 33,424. All counts run through filing year 2024 and include the two provisional years 2023 and 2024, so each is a lower bound. CPC is assigned mainly by the EPO, USPTO and KIPO; 26,948 of the 33,424 families in scope (80.6%) carry a CPC code in subclass C25B, so this table under-represents the purely national Chinese and Japanese filings in the corpus.

Methodology

Search strategy, data basis, and known limitations.

Data basis as of 15 August 2026: EPO PATSTAT Global, Spring 2026 (patstat_2026a) on Google BigQuery, with classification labels and hierarchy from IPC edition 20260101 (in force 1 January 2026) and CPC edition 202608 (in force 1 August 2026). The classification set was completed in all three respects the classification methodology requires: every seed symbol was expanded to its subtree, the predecessor symbols from the 2021 C25B reclassification were unioned in, and retired symbols were kept rather than filtered out with is_active.

Search strategy, counting rules and data source

Search Strategy

Codes were found through the three classification entry points, then completed over time:

  • Catchword index (dpt_catchword) — the WIPO natural-language index returns "ELECTROLYSIS of water" and "electrolysis of WATER", both pointing at C25B 1/04, and "preparation of oxygen by ELECTROLYSIS" pointing at C25B 1/02.
  • Title breadcrumbSEARCH(title_full, …) on dpt_ipc and dpt_cpc for "water electrolysis", "electrolysis of water", "electrolyser", "electrolysers", "electrolyzer" and "electrolyzers", in singular and plural.
  • Definition textsSEARCH(definition_scope, …) on dpt_cpc for the same terms. This is what surfaced Y02E 60/36 and the neighbouring water-treatment and fuel-cell groups.
  • Final scope — the C25B 1/02 branch (hydrogen or oxygen) with its full subtree C25B 1/04, C25B 1/042 and C25B 1/044; the photoelectrolysis group C25B 1/55; and the four groups retired at the end of 2020 whose content the reclassification chain records as moving into C25B 1/04 and the cell and electrode groups — C25B 1/06, C25B 1/08, C25B 1/10 and C25B 1/12. Nine symbols in total.
  • Note on breadth. C25B 1/02 reads "hydrogen or oxygen", not "hydrogen or oxygen from water". Families sitting on that parent group without a C25B 1/04 descendant are therefore slightly broader than literal water electrolysis; the group accounts for 4,052 families, 12.1% of the corpus.

Which Classification Scheme

The same nine symbols were run through both schemes and counted separately before the scope was fixed:

  • CPC at family level (tls225_docdb_fam_cpc): 26,425 families.
  • IPC at application level, aggregated to family (tls209_appln_ipc): 30,624 families.
  • Present in both: 23,625. CPC only: 2,800. IPC only: 6,999. Union: 33,424.

CPC is generally preferred for cross-office landscapes because it is applied consistently across family members, but the measurement above shows that a CPC-primary scope would have dropped 6,999 families — 21% of the union — in a field whose centre of gravity is at offices that assign IPC rather than CPC. Both schemes are therefore used, unioned on the identical symbol list.

Completing the Code Set Over Time

  • Subtree expanded. C25B 1/02 was expanded through C25B 1/04 to C25B 1/042 and C25B 1/044; C25B 1/55 has no descendants in the current scheme.
  • Predecessors unioned. 1,464 families in scope carry one of the four retired symbols, and 567 of them carry no current symbol at all — 1.7% of the corpus that a code list built from today's scheme alone would not have found. The retired symbols carry zero assignments in the family-level CPC table; all 1,464 come from the IPC side.
  • Retired codes kept. The code list is not filtered on is_active, precisely because the four retired symbols still sit on documents.
  • One predecessor hop was deliberately not unioned. The chain also records C25B 1/00 as a predecessor of the new photoelectrolysis group C25B 1/55. C25B 1/00 is the residual main group for the whole subclass and covers chlor-alkali, ozone, manganese oxides and ammonia, so unioning it would have pulled the entire inorganic-electrolysis field into a water-electrolysis scope. It is left out, and the consequence is stated where it matters: the C25B 1/55 count is a lower bound.
  • The trade-off. Unioning predecessors buys recall at the cost of precision: some documents in a retired group were redistributed to cell and electrode groups rather than to water electrolysis. The true figure lies between a reading with and a reading without the chain; this report states the reading with it, and states the size of the difference above.

Family Counting

  • Every count is COUNT(DISTINCT docdb_family_id). One invention filed in DE, US, CN and EP is one family, not four.
  • Each family is assigned to its earliest filing year across all its applications, so the annual series sums exactly to 33,424 rather than counting a family in every year it has a filing.
  • Only patents of invention are counted (ipr_type = 'PI'); utility models and designs are excluded. Pseudo-families (docdb_family_id = 0) are excluded.
  • Applicants and inventors are separated: applicant analyses use applt_seq_nr > 0.

Multi-Assignment Note

  • A family carries several classification symbols, is filed at several offices, and can have applicants in several sectors. Every breakdown in this report is therefore non-exclusive and its column sums above 33,424.
  • Concretely: the C25B main groups sum to 71,709; the applicant sectors to 38,603; the office table to more than the corpus. This is expected and is stated next to each table.
  • No pie or doughnut chart is used anywhere in this report, because none of these breakdowns are parts of a whole.

Data Source and Reproducibility

  • EPO PATSTAT Global, Spring 2026 edition, dataset patstat-mtc.patstat_2026a, queried on Google BigQuery through the patstat-mcp server.
  • Classification hierarchy, titles, definitions and the reclassification chain from patstat-mtc.classification (dpt_ipc, dpt_cpc, dpt_catchword, dpt_ipc_reclassification, dpt_edition).
  • All SQL is shipped with the report in queries.sql, including the frozen scope definition that every query in the set inherits.

Acknowledgement

The research question behind this report was suggested by the work of Riccardo Priore at Area Science Park (Centro Patlib, Trieste, Italy), who built an interactive applicant-exploration tool for the electrolyser domain on the EPO Technology Intelligence Platform. The analysis here is independent and uses PATSTAT Global on BigQuery.

Scope Limitations

  • Precision of the profile. Titles were read for two random samples. In the CPC-classified core, 2 of 25 families (8%) were off-topic. In the IPC-only stratum — the 6,999 families the CPC view does not reach — 10 of 24 readable titles (42%) were off-topic, mostly chlorine electrolysis, CO₂ electroreduction and hydrogen-water devices. Weighted by the size of the two strata, the profile carries an estimated 15% off-topic rate, inside the one-in-five limit but not far inside it.
  • What the search misses. 7,265 families filed 2000–2024 carry water electrolysis keywords in an application title; 2,791 of them are not in scope — 8.4% additional families that classification alone does not catch. They are recorded here rather than absorbed, because widening the profile to reach them would push the off-topic rate past the limit above.
  • The Y sections are not in scope. Y02E 60/36 reaches at most 17,449 families that carry no C25B symbol in CPC — an upper bound, because that comparison is possible only on the CPC side and some of those families are already in scope through IPC. A title sample of them found 13 of 25 (52%) to be outside water electrolysis. Measured and excluded — see the Application Domains section.
  • Two provisional filing years. 2023 and 2024 are incomplete because of the roughly 18-month publication lag. No growth rate in this report ends in either year, and every figure covering 2020–2024 is a lower bound.
  • Applicant names are not consolidated. PATSTAT's han_name does not merge corporate groups, legal-form variants or transliterations, so several large portfolios are split across entries. Named portfolio figures are a lower bound.
  • Applicant origin is not reported. person_ctry_code is blank for a large share of the Chinese applicant records in this scope. Geography in this report is filing authority, which answers where protection is sought, not where the work was done.
  • CPC coverage is partial. 26,948 of 33,424 families (80.6%) carry a CPC code in subclass C25B; CPC is assigned mainly by the EPO, USPTO and KIPO. Breakdowns that rely on CPC — the application-domain table and the Y-section figures — therefore under-represent purely national Chinese and Japanese filings.
  • Adjacent hydrogen routes are outside the subject. Steam methane reforming, biomass gasification, thermochemical cycles and purely photocatalytic water splitting are not part of this landscape.
  • No external benchmark. The family count was not cross-checked against an outside database in this analysis; it should be read as the result of the stated profile on the stated edition, not as an absolute size of the field.
Glossary — Patent Terms Explained
Patent Family (DOCDB)
A group of patent applications protecting the same invention in different countries. Counted once, so that international filing strategy does not inflate the numbers.
Earliest Filing Year
The lowest filing year among all applications of a family. Used to place each family on the time axis exactly once.
IPC / CPC
International and Cooperative Patent Classification. IPC is assigned by virtually every office; CPC is maintained by the EPO and USPTO and applied consistently across family members, but is not assigned everywhere.
Y02 / Y04 (Y sections)
CPC-only tagging schemes for climate change mitigation and smart grids, applied on top of the ordinary technical classification. They have no IPC equivalent.
Reclassification chain
The record of which classification symbol replaced which, per scheme edition. Without it, a young symbol produces a time series that measures reclassification progress rather than invention.
SOEC (Solid Oxide Electrolyser Cell)
High-temperature electrolysis of steam using a ceramic electrolyte. Classified in C25B 1/042 since the 2021 edition.
PEM (Proton Exchange Membrane)
Electrolysis using a polymer membrane electrolyte, operating at low temperature with fast load response.
Photoelectrolysis
Splitting water at an electrode driven directly by light. Classified in C25B 1/55 since the 2021 edition.
Grant Rate
Share of families with at least one granted application. Meaningful only on cohorts old enough to be largely examined.
PCT / WO
An international application filed at WIPO. It reserves the option of national protection later; it is never itself granted.
han_name
PATSTAT's harmonised applicant name. A first level of cleanup, not a corporate-group consolidation.

Report Files

This report consists of a metadata file (meta.json), a cover, a KPI strip, ten content sections and a contact section, plus queries.sql — the SQL actually executed for this analysis, carrying the PATSTAT and classification editions in its header. Anyone with access to PATSTAT Global on BigQuery can re-run those queries; on a later edition the figures will differ, which is why the edition is named on every page that carries a number.

Like what you see?

This report was built with a fully reproducible pipeline: EPO PATSTAT Global on BigQuery, a custom MCP server, and Claude AI for analysis and visualization.

Custom Patent Report Need a similar analysis for your technology area, company, or research question? We build tailored patent intelligence reports.
The Toolchain Want to run your own AI-powered patent analyses? We help you set up PATSTAT on BigQuery with MCP server integration.