AMPulse

Japan and Korea, full-census demography and patent estate

Two Ways to Be Quiet

Japan and Korea get filed together as the quiet Asian pair, and Japan as the laggard of the two. Counted whole, in AMPulse's census, they are two conversions from different industries: the Japanese patent estate sits inside heavy industry and materials houses founded before the process existed, the Korean one inside dental implant makers and a bioprinting cohort founded after it.

AMPULSE DATA REPORT · 2026Data snapshot: Sep 7, 2026

The pair nobody hears from

A 2023 survey of 327 manufacturers across three countries, commissioned by Materialise, put Japan last on producing end parts with additive: 36% against 60% in the United States and 43% in Germany, with the finding written up as Japan's "reluctance to onboard 3D printing for end-part production." The reading has stuck, and it survives on absence: no Japanese additive brand that English-language trade coverage returns to, and little published in English. Korea sits in the same sentence, and the pair gets filed as the quiet hardware-first Asian markets, with Japan the quieter of the two.

The quiet part is what our pipeline records. Across the last 12 months our news pipeline linked 917 stories to United States companies and 687 to Chinese ones. The Japanese and Korean censuses account for 250 between them. Neither produces story flow in the sources the pipeline reads, and those sources include nothing published in Japanese, so the counts say how visible each census is to this pipeline and not how busy each industry is.

The second half, that the two are the same kind of market, is where the census disagrees. Quiet is what they share on the surface. What follows takes the two censuses apart six ways: by age, by which industry the patents came out of on each side, by what the patents are filed as, by when the incumbents first filed on additive, by what the young companies were founded to do, and by who is buying machines.

One census is old, one is young

The median company in AMPulse's Japanese additive census was founded in 1995. The median Korean one was founded in 2014. Of Japan's 137 dated companies, 74, 54%, existed before 2000, and 63 of them existed before 1990; the SLA-1, the first commercial additive system, was commercialised in 1987. Korea's census holds 23 pre-2000 companies, 20.7%, against 68 founded since 2010, 61.3%.

The Japanese names carry it. Mitsubishi Heavy Industries dates to 1884, Toray to 1926, Matsuura to 1935, Kyocera to 1959. Each of them entered this census by adding a process to an existing business. Japan's additive-native cohort exists and is young: AeroEdge, the aerospace parts maker founded in 2015, is one of 55 Japanese census companies founded since 2010, and holds no patent link in our index.

Korea's old layer is thinner and its patent-holding members are mostly outside the machine-building tradition: Sindoh, founded in 1960, is a printer maker; DIO, founded in 1988, is a dental implant maker; Hanwha Aerospace dates to 1977. The companies doing additive as their whole business are recent: ROKIT Healthcare was founded in 2012 and T&R Biofab in 2013.

Value-chain mix says the same thing from a second angle, and not the thing the hardware-first shorthand predicts. Korea is the hardware-coded country, 51.8% of its categorized census against Japan's 30.9%; Japan's largest single block is services at 33.1%, against Korea's 21.4%. Listed companies are 23.9% of the Japanese census against 13.2% of the Korean one.

Where the recorded estates sit

Patents are where a quiet industrial base leaves a trace, and on the national figure Japan looks twice as deep. 94 of Japan's 138 census companies hold at least one active patent link in our index, and they hold 9,507 between them, 101.1 per holding company. Korea's 88 holders hold 4,053, 46.1 each. Patent ingestion has been attempted for 82.6% of the Japanese census and 85.1% of the Korean one. Similar attempt rates say nothing about how complete each retrieval was, so the national figure is reported as what the index holds and nothing more.

It is also not a statement about either country's additive research. Split the same holders by founding age and the per-company gap mostly disappears.

Among companies founded before 2000, the median Japanese holder carries 72 links and the median Korean holder 65. Among 2000s foundings the order reverses, 50 for Korea against 7 for Japan. Among 2010-or-later foundings the medians are 13 and 7, both low.

What differs is how many old companies each country brings. Japan's census contains 66 patent-holding companies founded before 2000; Korea's contains 19. By the share of its own census that files at all Korea is ahead: 77.2% of Korean companies hold a link against 68.1% of Japanese ones.

A link is one patent document, and the two estates are filed differently. Japanese holders' links are spread across five offices, Korean holders' links are three-quarters domestic, and the database's family table does not merge one invention's filings across offices. So a Japanese invention filed at home, in the United States, in Europe, under the PCT and in China can count five times against a Korean domestic filing counted once. The control is to count each company's filings at its own office only.

On that cut the age-band comparison changes direction: the median pre-2000 Korean holder has 92 filings at home against the Japanese holder's 35, and among 2000s foundings 42 against 6.5. On all offices Japan's old holders lead; at home Korea's do. These counts cannot estimate or compare invention rates. What they identify, on both cuts, is where each estate sits: 90.2% of Japan's home-office filings belong to companies founded before 2000, against 50.2% of Korea's. That is the finding the rest of the report rests on.

Korea's estate is a dental industry

The 2000s reversal in that table has a dental centre, and it reorganises the whole Korean side of the comparison. Read the descriptions of Korea's patent holders and pick out the ones whose business is dental: dentistry, orthodontics, the clear aligners and surgical guides that a dental 3D printer exists to make. The rule is dental, not implant: a wider rule that admitted spinal, cranial and facial implant makers would add 5 companies and 131 links, B-Trust Meditech, CGBio, FITme, Medyssey, Sigemedtech, and they are counted separately. There are 12 of them, and between them they hold 1,280 of Korea's 4,053 active patent links. That is 31.6% of the estate this census holds for Korea. Narrow to B33Y, the class that exists only for additive, and dental companies hold 31.9% of Korea's links. The Japanese census has 1 such company, holding 3.1% of Japan's estate.

By their product lines, the three at the top are implant makers that added additive. DIO, founded in 1988, sells the DIOnavi guided-surgery workflow and, since 2018, its own PROBO dental printer for surgical guides and provisional crowns. MegaGen, 2002, and Dentis, 2005, are implant houses that added printers and resins to a chairside workflow they already owned. Their estates are implant, abutment and guide patents accumulated over two decades; DIO holds 334 active links and 6 of them carry the additive class. Graphy, founded in 2017 and listed, is the one dental company in the top group that was formed for the process: its shape-memory clear-aligner resin is printed, and 70 of its 102 links are B33Y.

So the median-links reversal among 2000s companies runs through MegaGen, Dentis, GENOSS and Ray Medical, and it does not depend on them: the 5 dental holders of that band carry a median of 93 links, the 12 others 36, and Japan's 2000s holders 7. Dental holders have the higher counts; the Korean median stays above the Japanese one with them excluded. Read by age and dental together, Korea's additive-class links divide this way: companies founded since 2010 hold 60.9% of them, and among the post-2010 companies holding any patent link, the 5 dental companies hold 122 additive-class links against 412 for the 47 others, some of which hold none. Of the dental companies' additive-class links, 158 sit with companies founded before 2010, the implant houses and the dental printer makers Carima, Veltz 3D and Ray Medical among them, and 122 with post-2010 companies; the post-2010 majority of the whole estate comes mostly from companies that are not dental. And the "insurgent" reading of Korea needs the same correction the Japanese "laggard" reading needed. Korea's deepest additive patent holders are also converts. They converted from a different industry than Japan's did: dental implants rather than heavy machinery and materials. When either group converted is not something the founding years say. What the two countries do not share is which incumbent industry adopted the process, and that turns out to be the fact the patent figures were carrying all along.

Japan's estate is a materials industry

The same question put to Japan gets a different industry. Sort each country's active links by the census category of the company holding them, and the Korean estate this census holds is a hardware estate, 70.7% of its links in 45 companies the census files as machine or device makers, the dental implant houses among them. Japan splits: 43.9% hardware and 38% materials, the second share held by 24 chemical, metal and ceramics companies. Korea's materials holders carry 9.1%.

The names are the postwar chemical and metals industry. Toray (1926, 569 links), Kyocera (1959, 408 links), Asahi Kasei (1922, 399 links), Daicel (1919, 387 links), Mitsubishi Chemical (1933, 376 links), Sanyo Special Steel (1933, 257 links). The hardware side is not printer makers either: five machine-tool builders, DMG MORI, Mazak, Sodick, Okuma, Matsuura, hold 983 links between them, and the electrical and heavy-industry groups, Mitsubishi Heavy Industries, Mitsubishi Electric, DAIHEN and Nidec, hold most of the rest.

What those companies file tells the same story from inside the patent office. Read every link by the classes its patent carries and the two estates are made of different things.

Japan's largest class is plastics shaping, 36.9% of its links, then welding and laser working at 21% and metal powder at 18.9%; the additive class itself is 19.4%. Korea's largest class is the additive class, 21.6%, and its second is dentistry, 15%; with implants and implant materials, 899 of Korea's 4,053 links carry a dental or implant class, against 79 of Japan's 9,507. Welding, the class a wire-arc or laser-deposition process files under, is 2.5% of Korea's estate and 21% of Japan's.

One cut inside the powder class matters more than the rest. Of Japan's 1,801 powder-class links, 856 are also filed under the additive class: powder patents written for a printer, from atomizers and steelmakers that were making powder before anyone printed with it. Korea holds 199. Count home-office filings only and it is 298 against 123. These are counts of co-classified filing documents, not of atomization capacity, qualified alloys or supply contracts, and on the single-office count the ratio is about two to one, against four to one on all offices; neither count is a measure of anything beyond documents. What they say is that the companies filing on printer powder in Japan are converted materials houses that an English-language reader would not name as additive companies, at a time when ISO/ASTM 52948:2026 gives metal powder-bed parts a shared imperfection vocabulary, SAE's AIR7359 and ARP7302 address powder feedstock and closed-loop powder sampling, and EOS has bought an atomizer, Metalpine, to hold its own powder supply. The class mix on the home cut is the same shape as on all offices: plastics shaping 29.5% and welding 19.8% for Japan, dentistry 13.9% for Korea, materials houses 37.7% of the Japanese estate against 8.2% of the Korean, dental companies 32.4% of the Korean against 1.9% of the Japanese.

This reframes the survey the report opened with. A questionnaire to manufacturers about end-part production measures the user side of additive. The estate this census holds for Japan is on the supplier side: the materials, the machine-tool cells, the welding heads. A supplier-heavy patent record does not show that Japanese factories run the process, and the survey does not show that they do not; the two measure different sides of the same industry, and the class mix is why the survey result and the patent record can both be true.

Who filed first

Founding year says how old a company is. It does not say when it started working on additive. The patent record can bound it. The additive-only class B33Y dates from 2015, and offices apply it to older documents on reclassification: this census holds 32 Japanese and 16 Korean B33Y links with application dates before 2015, the earliest from 2010. So the class reaches back unevenly, and the report dates by title instead: a company that filed on "three-dimensional shaping" or "積層造形" in 2005 filed on additive in 2005 whatever class the office later put it in. The table gives each holder's earliest retrieved additive-titled filing. It dates the recorded invention, not the current holder's entry: 6 rows hold a filing older than the company, an inherited or reassigned estate (Nabtesco, 1996 against a 2003 founding; Layerwise, 2014 against a 2020 founding; SEED, 2015 against a 2023 founding; Cubicon, 2016 against a 2017 founding; Mitsubishi Power, 2018 against a 2020 founding; SEKISAI, 2021 against a 2022 founding), and they are kept out of the named early-filer list below.

Two things sit in it. The first is how few of the old holders match at all: 34 of Japan's 66 pre-2000 holders and 7 of Korea's 19 have a linked patent whose title names the process, with medians of 2018 and 2020 among those that do. That dates the matched subset, late in the 2010s in both countries, and says nothing about the holders whose titles never name it.

The second is the early filers among the pre-2000 founders, and in this record they are Japanese. 9 pre-2000 Japanese holders have an additive-titled filing dated before 2015 against 0 Korean ones, and the earliest of them go back to the process's first decade. The oldest additive-titled filing in the Japanese census is from 1996, held by Nabtesco, a 2003 company, through Teijin Seiki, which licensed DuPont's Somos stereolithography technology in 1991 and brought its Soliform machine to market in 1992. Matsuura's first is from 2005, three years after it commercialised its first hybrid metal system, the M-PHOTON, in 2002, which it describes as the first machine to put a metal laser-sintering bed inside a milling centre; the LUMEX line followed in 2004. CMET, the stereolithography maker Teijin Seiki took control of around the turn of 2001, is at 2006. Korea's earliest are Carima in 2009, a resin printer maker, and ROKIT Healthcare in 2013, the year after it was founded. Japan has a 1990s stereolithography industry and a 2000s hybrid-metal one buried inside its census; Korea's additive patent record starts in the late 2000s and is mostly a 2010s record.

Where each country's additive patents live

Narrow the same links to patents carrying a core additive or near-additive class, B33Y for additive itself plus the powder-metallurgy, polymer-shaping, welding and ceramics classes the pipeline treats as its core set, and the two estates sit in different generations of company. 93% of Japan's 6,803 core links belong to companies founded before 2000. 54.8% of Korea's 1,198 belong to companies founded since 2010.

Narrow once more, to B33Y alone, and the national medians meet: 6 links for the median Japanese holder of one and 6 for the median Korean holder, across 64 and 55 companies. On the additive class the positive-holder medians match while the totals do not: 1,848 Japanese B33Y links against 877. What the converted industrial base adds on top of that is the classes around additive: decades of powder, welding and polymer-forming filings that a machine-tool group or a materials house accumulates whether or not it ever ships a printer. Korea's B33Y links are majority-held by companies that did not exist in 2010, and a third of them by dental companies.

What the young companies were founded to do

The two post-2010 cohorts are of comparable size, 55 Japanese companies and 68 Korean, and they were founded for different work. 22.4% of Korea's cohort names a medical target market, 15 companies, and the deepest of them are bioprinters: ROKIT Healthcare and T&R Biofab, with Clecell, Aldaver, Innoregen and Holosmedic behind them. Japan's medical share is 11.1%. Where Japan's cohort leans instead is toward consumers, 18.5% against Korea's 4.5%, a share made of service bureaus, distributors and fabrication spaces rather than product companies.

The category labels understate one Japanese cluster. Four of the companies the census files under industrial or consumer hardware are construction printers: Polyuse, Serendix, V3D Asia and DigitalArchi. Serendix completed what it describes as Japan's first 3D-printed house in 2022, delivered a two-person unit in Suzu after the Noto earthquake in 2024, and has a capital alliance with JR West to print station buildings; it has the most recorded funding events in the Japanese cohort, 8 since 2021, and the most-covered, with 12 stories in the trailing 24 months. Ricoh 3D, with 41 patent links, leads the cohort's filers, and it is an incumbent's additive unit dated by the launch of the business rather than a startup; Sun Metalon, at 38, and Polyuse, at 33, follow. The named Japanese natives are building houses, prosthetics and cranial helmets, and recycling metal chips; the machine builders among its most-funded are construction printers, and none makes an industrial metal or polymer system. Korea's are printing tissue, aligners and rocket parts.

The patent record separates the cohorts the same way. 52 of Korea's 68 post-2010 companies hold a patent link; 21 of Japan's 55 do. The Japanese cohort holds more bureaus and distributors, and those are the rows with no link; whether they file elsewhere is not something a missing link says.

Funding: thin in both, and shaped differently

Neither country's ledger holds many events. Since 2021, the AMPulse ledger records 60 funding events across 28 Japanese companies and 82 across 39 Korean ones, counted over the census after its duplicate rows were merged on 2026-09-08, so every record attached to a former duplicate is in the count.

Korea's ledger has a 2022 cluster that Japan's never shows: 18 event records, which collapse to 18 financings, 16 of them equity by 16 companies. At this size that is a shape in a thin ledger, not a national venture cycle. Japan has never recorded more than 5 equity financings in a year, and its ledger is grants, debt and strategic money around a small equity core. The Japanese companies that raise are the construction and medical natives, Serendix, Instalimb, Sun Metalon, Polyuse and Japan Medical Company. The Korean ones are Graphy, INNOSPACE, Breezm and T&R Biofab, with a seed-and-series-A layer under them.

Who is buying machines

Patents and founding years describe capability. The AMPulse equipment ledger records purchases, and for this pair it holds 16 verified deals, 8 in each country. That is too few to trend and enough to name the buyers.

The Japanese buyers are the incumbents and the state. Hitachi Metals took a Sciaky electron-beam system, Kimura Foundry an ExOne sand printer, SWANY a 3D Systems pellet extruder, Nikon runs an NXG XII 600 at its own AM technology centre, and the public entries are AIST twice, a prefectural research institute, and an Air Self-Defense Force supply depot: eight deals, seven buyers. The Korean buyers are the state's research arms and its military: KITECH, the Korea Institute of Materials Science, the Marine Corps with a mobile Meltio system, a university with a BLT machine, alongside one valve maker and three small firms. In neither country does a purchase by a young additive-native company appear.

One line in the Japanese column deserves the weight the laggard reading denies it. Nikon completed its takeover of SLM Solutions in January 2023, a deal announced at about 622 million euros, and the German company that built the NXG XII 600 now reports to Tokyo. The machine in the ledger is one of the largest metal powder bed platforms on the market, installed at the parent. Nikon itself is not a census row, as the exclusion note above records; its additive position enters this report only through the ledger.

Two subsets of the same census

The census objection writes itself: AMPulse finds Japanese companies through channels that reach large established firms and Korean companies through channels that reach new ones, and the demography is the discovery method reflected back. The cuts below are subsets of the same census, not independent discovery channels, and they cannot speak for companies the census never found. What they test is narrower: whether the shape depends on the discovery-heavy tail of the census. Inside the patent-holding subset, and inside the news-linked subset that most favours young companies, it does not.

Inside the patent-holding population the gap widens to 70.2% against 21.6%, and the median Japanese patent holder was founded in 1970 against Korea's 2014. Read through news flow alone, Japan still carries 38.6% pre-2000 to Korea's 23.2%, and that cut runs against the finding twice over: news favours young companies everywhere, and our Korean-language sources surface young Korean companies that have no Japanese-language counterpart. That cut is shown because a reader will ask for it; with nothing behind it in Japanese, it carries no weight of its own.

The counter-thesis

The strongest case against this report is that its patent figures measure conglomerates, not additive. Mitsubishi Heavy Industries, Toray and Kyocera file on everything; a B22F powder-metallurgy patent from Toray has nothing to do with a printer, and counting it as Japanese additive depth inflates Japan for no reason. That is why the B33Y-only cut is given, and why on that cut the two national medians meet. Where the objection bites harder is the early-filer claim. Nine pre-2000 Japanese holders filing on additive before 2015 is a small number, and one of them, CMET, sits in the Teijin Seiki lineage whose 1996 filing now rests with Nabtesco, a 2003 company that falls outside the nine. A reader who wants to call that a footnote rather than an early industry is entitled to; the report's claim is only that the Korean record holds no equivalent footnote.

The second objection is that the dental finding cuts against the report's own frame. If Korea's deepest holders are implant makers who converted, Korea is not the insurgent and the two countries are the same story, incumbents adopting a process, with different incumbents. That is the reading this report now takes, and it is stated in the dental chapter: the difference between the two countries is which industry converted, not whether conversion happened. What survives of the original frame is the young cohort. Korea's 68 post-2010 companies hold 52 patent estates and raise equity; Japan's 55 hold 21 and mostly do not. The insurgent layer is real in Korea and thin in Japan, and it sits on top of a dental incumbency rather than in place of one.

The third is the one the report cannot answer: companies the census never found. If AMPulse systematically misses small Japanese job shops that bought a metal system in 2019 and never issued a press release, Japan's young cohort is undercounted and, since job shops are services, its service share understated rather than overstated. The instrument cuts show the shape holds inside the companies the census did find; they say nothing about the ones it did not. The equipment ledger, at eight deals per country, is the same kind of floor.

Two clocks

Additive has no single adoption clock. An aerospace part is qualified into a program; a dental guide or aligner is made to a prescription under a device pathway that existed before printers; a defense purchase is politically accelerated and qualification-bound; a consumer-electronics part moves once the economics unlock. These are different processes with different start and end points, so the report puts no duration on any of them and ranks none against another.

The hypothesis this report offers, and does not test, is that the two censuses look the way they do because each country's converts came from industries with different adoption processes. Korea's implant houses adopt under a device pathway with a prescription at the end of it, and Korea's estate carries a dental class share and a post-2010 majority of additive-class filings. Japan's machine-tool builders, steelmakers and heavy-industry groups adopt through qualification into someone else's product, and Japan's estate carries plastics-shaping, welding and powder classes held by companies that were filing in those classes before the process existed. The census is consistent with that reading. It does not establish it: the data describe who holds which patents, not when any company converted, how long its qualification took, or what it produces.

Two things to monitor, which would be consistent with the hypothesis without testing it. In Korea, the medical natives are entering a different process from the one the dental incumbents used: a device clearance is a starting gun, not a market, and the binding constraints sit downstream in distributors, hospital purchasing and reimbursement. T&R Biofab's cranial implant received its FDA clearance in July 2026, and that is where its clock starts; living-tissue bioprinting runs on a different regulatory path again and is not covered by the same reading. In Japan, procurement is one place a qualification-bound conversion can surface: the Air Self-Defense Force award of December 2025 is the ledger's first Japanese military purchase, a single example rather than a layer. From December 18, 2026, Section 849 of the FY2026 NDAA restricts DoD contracts to procure covered AM machines. Coverage follows supplier-list, domicile and designated foreign-control criteria, with purpose-specific exceptions. The restriction establishes no automatic preference for Japanese or Korean equipment. The equipment ledger records 569 United States installations, an OEM origin for 63 of them (24 American, 16 German, 9 Australian, 0 Japanese or Korean), and no defense flag on the population. It can record who installs what after the date; it cannot say whether a purchase happened because of the exclusion.

What each country is exposed to

Two censuses this different are not one market at two speeds, and the risks do not rhyme. The Japanese estate in this census is held by large, old groups in which additive is one line among many, which is a position from which a programme can continue unnoticed or be shelved unnoticed. 66 pre-2000 patent holders is a deep bench and a diffuse commitment; the Materialise finding about end-part production is compatible with every number in this report, because a converted base can hold the patents and still not run the parts. The Korean estate sits in two places with two different exposures: a dental incumbency whose additive work is one line inside an implant business, and 52 patent-holding companies founded since 2010, most of them formed around the process. Korea's 2022 equity cohort, 16 financings by 16 companies, is the one to watch; a young cohort's ordinary exposure is a selection cycle it has not been through, and that cohort reaches it first. The medical natives inside it carry the extra exposure of the clearance-to-revenue gap: a cleared device with no distributor is a cost centre on the clock the regulator does not measure.

What to watch

• Korea's 2022 equity cohort. 16 equity financings by 16 companies in one year, most at seed or series A. Which of those companies has raised again by the end of 2027 is the first selection-cycle reading the Korean cohort will give.
• Nikon SLM in Japan. The NXG XII 600 at Nikon's own centre is one installation. A large-format metal system installed at a Japanese customer other than Nikon itself would be the first external installation in the record; it would not by itself say whether the acquisition produced it.
• Dental filing in Korea. Graphy and DIO are the two holders whose filings would show a move from guides and provisionals toward printed final restorations first, if the subject matter of their next additive-class filings names it; the class alone does not.
• Japanese defence procurement. The Air Self-Defense Force award of December 2025 is the ledger's first Japanese military purchase. A second, or one from the maritime or ground services, would put Japan on the same buyer path Korea's Marine Corps is already on.
• A Japanese or Korean machine on a United States defense floor. The ledger records none among 63 origin-recorded American installations. The first such installation after the Section 849 equipment-procurement restriction takes effect would be the first in the record; by itself it would not show displaced demand.
• Construction. Serendix's JR West alliance and Polyuse's patent estate are the two Japanese native bets with a domestic customer. Building permits for printed structures, and any Korean entry into the same segment, are the tripwires.

Appendix: the company panel

The companies behind the figures, so a reader can check the claim against the row. Each country's panel is its twelve deepest patent holders, every company with a recorded funding event since 2021 or at least three collected stories in the trailing 24 months, and every patent holder the dental rule matches, with the matched term shown.

Public document examples checked on September 30: DMG MORI: US12138723B2 (Broad machine-tool estate; not an AM-specific example); ROKIT Healthcare: WO2022131461A1 (B33Y-classified medical modelling/AM document). These illustrate the counting scopes; they do not validate every linked document, first-entry year or legal status.

The September 30, 2026 correction replaces the earlier Section 849 interpretation with its enacted equipment-procurement scope. The frozen analytical data are unchanged.

Lee, S. (2026). Two Ways to Be Quiet. AMPulse Data Reports. https://www.ampulse.online/reports/japan-korea-quiet-two-ways-2026