Six Thousand Companies—and the Scale Test: Japan’s University Startups Reach 6,220

Japan has identified a record 6,220 university-connected startups, 1,146 more than in the previous survey. The number confirms that company formation has become part of the nation’s research system. It does not yet show that enough discoveries are becoming products, global revenue or enduring institutions. Japan’s next test is scale.

A century before “deep tech” entered the policy vocabulary, Masatoshi Okochi was already trying to make Japanese science leave the laboratory. The third director of RIKEN created a company in 1922 to commercialize Adosoru, an activated carbon developed at the institute. By 1939, the network later called the RIKEN Konzern encompassed 63 companies and 121 factories. One of its descendants began selling sensitized paper in 1936 and became Ricoh.

That is not the direct ancestry of today’s university startups: RIKEN was a research institute, and the wartime industrial group belonged to another political economy. It is a useful prehistory. Japan has never lacked excellent science or isolated examples of invention becoming industry. Its recurring difficulty has been building a repeatable, transparent system that can move thousands of discoveries through ownership, proof, finance, management and markets without depending on an Okochi-like industrial organizer.

A new government count suggests that the front end of that system is finally broad. On June 12, 2026, the Ministry of Economy, Trade and Industry reported that it had identified 6,220 university startups as of the end of October 2025. The previous survey had found 5,074. Both the total and the increase of 1,146 were records. The University of Tokyo led with 595, followed by Kyoto University with 503 and Keio University with 473.

The headline is striking, but its meaning is narrower—and more interesting—than “Japan created 1,146 university companies in one year.” METI’s number is a stock, not an annual formation count. The difference includes enterprises that were newly found by the survey, companies that newly met its definition and changes in the population examined. It is broad enough to include a company that had no university relationship at birth but began joint research or received university technology during its first five years.

Six thousand two hundred twenty is therefore not a clean birth certificate for 6,220 laboratory spinouts. Nor is it a claim that all 6,220 are equally active, independent, funded or growing. It is a map of companies connected to Japanese universities through one of six routes. Read carefully, the number still records a profound institutional change: entrepreneurship has moved from a rare deviation in academic life toward a normal path by which research, people and technology enter society.

Japan’s university-startup question used to be: Can a professor form a company? The harder question now is: Can that company cross the valley of death, win customers and become global before its capital runs out?

6,220University-connected startups identified by METI as of the end of October 2025.

+1,146The survey-to-survey increase—not a count of newly incorporated firms.

595Startups associated with the University of Tokyo, the highest university total.

About 60%Of 1,093 locatable additions were founded outside Tokyo, up from about 54%.

¥761.3 billionJapanese startup equity funding recorded by Speeda for 2025, roughly flat year on year.

¥10 trillionThe government’s 2027 startup-investment ambition; its scope is not identical to the equity series.

What exactly did Japan count?

METI uses six categories. “Research-result ventures” commercialize inventions or new technologies created at a university. “Joint-research ventures” conduct joint work with a university within five years of formation. “Technology-transfer ventures” receive university technology within the same period. “Student ventures” are founded by students, while “faculty ventures” are founded by faculty members, staff or postdoctoral researchers. A final “related venture” category includes companies recognized or invested in by a university and certain companies created shortly after a founder left campus.

The categories capture a modern innovation ecosystem more accurately than a patent-only definition. A software company may derive its advantage from a founder’s research skill without licensing one patent. A company can acquire academic technology after it begins. A student startup may matter economically even when it does not originate in a laboratory. But breadth carries a cost: the total cannot be compared casually with a country that counts only companies formed to license university-owned intellectual property.

How to read the record

It is the number of companies METI could identify at the survey cutoff, not the number born during fiscal 2025.
The increase reflects new incorporations, newly identified companies, newly qualifying relationships and survey changes.
The definition includes research spinouts, student and faculty companies, later joint-research partners, technology recipients and other university-recognized firms.
The count does not by itself measure survival, revenue, exports, employment, follow-on capital or scientific impact.

Even the university league table is partly a table of discovery and institutional capacity. An office that searches alumni networks carefully, defines affiliated ventures consistently and maintains better records may show a jump without creating every additional company that year. Nagoya University’s total rose by 114 to 267; Kindai University’s rose by 78 to 196. Those changes can reflect genuine momentum, better identification or both. The ranking is valuable, but it should begin questions rather than end them.

Rank	University	2025 survey count	Previous count	Increase
1	University of Tokyo	595	468	+127
2	Kyoto University	503	422	+81
3	Keio University	473	377	+96
4	Osaka University	342	298	+44
5	University of Tsukuba	301	264	+37
6	Nagoya University	267	153	+114
7	Tohoku University	260	222	+38
8	Tokyo University of Science	257	226	+31
9	Institute of Science Tokyo	209	187	+22
10	Kindai University	196	118	+78

The wall between campus and commerce

Postwar Japan became an industrial and technological power through large corporations, lifetime employment, bank finance, supplier networks and patient internal research. The model produced formidable laboratories at companies such as NTT, NEC, Hitachi, Toshiba, Sony and the pharmaceutical groups. It did not give a university professor an obvious route to own an invention, assemble a team, raise risk capital and become chief scientific officer of a new enterprise.

Japanese universities traditionally treated publication and education as their mission while companies developed products. National-university researchers were civil servants before the 2004 incorporation reforms. Intellectual-property ownership could be fragmented or uncertain, conflict-of-interest rules were cautious and the transfer offices common at American universities scarcely existed. A professor who wished to enter business crossed not only a funding gap but a cultural boundary.

The United States had moved in a different direction. The 1980 Bayh–Dole Act let universities, small businesses and nonprofits retain rights to inventions made under federal funding, helping technology-transfer offices license research. The American system has weaknesses and should not be romanticized: most patents earn little, regional concentration is severe and university entrepreneurship can distort academic incentives. But it created a legible path from a federally funded invention to exclusive licensing, venture investment and a startup.

Japan’s institutional answer began to take shape in 1998. The Technology Licensing Organization Act created an approval and support system for TLOs that identify university inventions, seek patents, find licensees and return part of the income to researchers and institutions. Before that reform, METI notes, universities generally did not possess corporate-style organizations dedicated to patenting and licensing their researchers’ results.

The policy goal soon became numeric. In 2001, the “Hiranuma Plan,” named for economy minister Takeo Hiranuma, called for the creation of 1,000 university ventures in three years and a tenfold increase in university patents over a decade. MEXT found 916 university ventures by August 2004, up from 614 in its preceding count. Other historical series put the total above 1,000 by the end of fiscal 2004 and around 1,500 by the end of 2005, depending on definition and timing.

Then came a sobering lesson. Building the company was easier than building the market. Academic reviews of the first boom found that many drug-development startups disappeared or abandoned therapeutics, a field that can consume years and enormous capital before a first sale. The rate of formation slowed in the late 2000s. Support had increased the number of vessels leaving shore without making the ocean less dangerous.

Universities become institutions that can invest

The 2004 incorporation of national universities gave them more managerial autonomy and strengthened their ability to own and manage intellectual property. Universities built IP headquarters and collaboration offices. Licensing and joint-research volumes rose. Yet METI’s own later assessment found that value per collaboration and licensing income remained small. More doors had been installed; not enough traffic moved through them at commercial scale.

A second layer addressed the space between a scientific result and an investable business. The Japan Science and Technology Agency’s START program brings “business promoters” into university projects to conduct due diligence, shape intellectual property, build prototypes and recruit managers before incorporation. University ecosystem programs provide gap funds—money for the experiments, engineering and customer validation that research grants and private investors often leave unfunded.

Universities also gained dedicated venture investors. Osaka University Venture Capital and Kyoto University Innovation Capital were established in December 2014 as wholly university-owned firms. Kyoto iCAP’s first fund reached ¥16.001 billion in 2016, its second ¥18.14 billion in 2021 and a third announced in 2026 can reach ¥20 billion. The University of Tokyo formed UTokyo Innovation Platform in 2016; its first fund totaled ¥25.001 billion with the university and major banks participating.

These institutions do more than write checks. A university fund can recognize science too early or specialized for a general venture capitalist, finance proof-of-concept work, assemble a syndicate and remain patient. A TLO can protect the patent portfolio and structure a license. An incubator can provide wet-lab space. Entrepreneurship courses can introduce students to founders before a career choice becomes irreversible. Together, they turn startup creation from an act of defection into one legitimate form of academic impact.

Japan’s broader startup policy then shifted into overdrive. The national five-year plan adopted in November 2022 organized policy around talent and networks, startup finance and diverse exits, and open innovation. It set an ambition to increase annual startup investment roughly tenfold, from around ¥800 billion to ¥10 trillion by fiscal 2027. The target was deliberately galvanizing. It was also large enough to reveal how far capital supply still lagged behind the desired company population.

The valley between a paper and a product

A research result and a company are not adjacent stages. Between them lies an expensive sequence of proofs. Does the result reproduce outside the originating laboratory? Is there protectable intellectual property and freedom to operate around other patents? Can the technology work at manufacturing scale? Who is the first customer? What regulatory evidence is required? Can gross margins eventually support the cost of sales? A university startup must answer every question while the science itself may still be uncertain.

That is the “valley of death,” but the phrase can conceal multiple valleys. The first is technical: turning a laboratory demonstration into a reliable prototype. The second is commercial: proving that a customer will pay for the solution rather than admire the science. The third is organizational: transferring tacit knowledge from a professor and a few students into a company that can hire, document, manufacture and sell. The fourth is financial: finding investors whose fund life and risk tolerance match the development clock.

The mismatch is especially severe in biotechnology, advanced materials, quantum hardware, semiconductors, fusion, robotics and climate technology. Software can reach users with modest initial capital. A drug may require toxicology, clinical trials and regulatory review; a new material may need a pilot line and years of qualification inside a customer’s product; a robot must survive real factories rather than controlled demonstrations. Failure is normal, assets are specialized and revenue arrives late.

METI’s drug-discovery support explicitly acknowledges that Japanese startups have greater difficulty than American and European peers in securing the capital needed for long development programs. Its program couples public subsidies with investment from certified venture-capital firms for preclinical work and early clinical trials. Regional METI work makes a parallel point: deep-tech companies outside the largest hubs face large early losses and difficulty financing research equipment and development.

Gap funding is therefore not a charitable bridge to venture capital. It is part of market design. Public money should absorb uncertainty that private funds cannot yet price, but it should buy evidence—not indefinite survival. The best programs use milestones, external technical review, customer discovery and follow-on discipline. The object is not to protect every laboratory company from failure. It is to let the technically strongest ideas fail or advance for the right reasons.

Capital has grown; the clock has grown faster

Japanese startup finance is far larger than it was a decade ago, but the 2025 market showed the limit of momentum. Speeda recorded ¥761.3 billion in equity funding, broadly flat from the prior year. That series should not be compared mechanically with the government’s ¥10 trillion goal because data providers, instruments and policy documents do not use perfectly identical boundaries. The directional gap is nevertheless unmistakable.

Early rounds alone cannot solve it. Deep-tech companies need larger Series B, C and later financings, venture debt, project finance, strategic customers and—in some fields—non-dilutive research and procurement contracts. Japan has often produced small initial public offerings that give founders and investors liquidity but leave the company undercapitalized for global competition. A thin late-stage market can turn the Tokyo Stock Exchange Growth Market into an emergency financing mechanism rather than the culmination of scale.

In January 2026, Reuters reported on a new late-stage fund being raised by Minerva Growth Partners against precisely this background: too little growth capital and pressure for startups to list before they are ready. One fund does not close the gap, but it reflects a changing diagnosis. The bottleneck is no longer only the seed round around a professor’s patent. It is the ¥5 billion, ¥10 billion or larger sequence required to internationalize manufacturing, clinical development, sales and acquisitions.

Financing design also affects who controls the timetable. A ten-year venture fund may be poorly matched to a platform technology that needs twelve years to regulatory approval. Corporate venture capital can contribute technical validation and customers, yet become strategically restrictive if a startup is perceived as captive to one industrial group. Government subsidies reduce dilution, but annual administrative cycles can interrupt a multiyear engineering plan. Global investors bring larger networks but expect governance, reporting and market ambition that a domestically oriented team may not yet possess.

The missing occupation: company builder

METI’s survey shows a genuine strength: university startups employ people with doctorates at higher rates than ordinary corporate research operations. The country worries that doctoral talent has too few attractive career paths; laboratory companies can convert advanced training into economic capacity. The survey also found many chief executives whose prior careers were in universities or research institutions.

Scientific leadership is essential, but it is not the same occupation as company building. A principal investigator is trained to identify new questions, publish evidence and win grants. A chief executive must narrow choices, allocate scarce cash, recruit across functions, negotiate licenses, hear “no” from customers, kill projects and finance the next two years before the previous experiment is complete. Some people can do both. An ecosystem cannot assume that every inventor should.

The strongest model often separates and connects three roles: the scientific founder who protects the truth of the technology, the entrepreneurial chief executive who builds the organization and the experienced board that can challenge both. Japan needs more people who have moved repeatedly among industry, investing and startups—commercial translators who understand a molecule, material or machine well enough to construct its market.

Universities can help by making entrepreneurial leave, conflict-of-interest rules and equity ownership predictable. Companies can lend experienced operators without smothering the startup’s independence. Investors can recruit leadership before requiring a pitch. Immigration and compensation policy matter, too: a global executive or regulatory specialist compares Japanese equity, tax treatment, language and career mobility with opportunities in Boston, London, Singapore or the Bay Area.

Exit is not an ending

A healthy venture system recycles experience and capital. When a company lists successfully or is acquired, early investors return money to new funds, employees become angel investors, founders mentor the next cohort and universities receive license or equity proceeds that can finance more research. The exit is not the last chapter of one company. It is the first financing event for the next generation.

Japan has historically leaned toward IPOs. METI’s May 2026 startup M&A guidelines argue for a genuine dual track: companies should be able to compare a public offering with a strategic sale, and large corporations should treat acquisitions as a route to new business rather than a procurement transaction that absorbs and neutralizes the team. A wider M&A market would give deep-tech founders more paths to manufacturing, distribution and global regulatory capacity.

The quality of the exit matters more than the label. A premature ¥5 billion listing with low liquidity can trap a company in quarterly fundraising. An acquisition can destroy the technology if key employees leave and the buyer lacks an integration plan. Conversely, a well-capitalized public company can become an acquisition platform, and an industrial buyer can accelerate a university invention through facilities the startup could never afford. Policy should reward durable scale, not celebrate the bell-ringing or contract signing in isolation.

Three different roads out of the university

PeptiDream shows one route: build a proprietary discovery platform and use global pharmaceutical partnerships to finance expansion. The company was founded in July 2006 by University of Tokyo professor Hiroaki Suga and entrepreneur Norikazu Kubota, initially working from university laboratory space under an exclusive license. It listed in 2013. By 2024 it reported ¥46.6 billion in revenue and ¥21.1 billion in operating profit while expanding a global discovery agreement with Novartis. Its product is not one molecule alone; it is a platform capable of repeatedly generating candidates and partnerships.

CYBERDYNE shows a regulated hardware route. Founded in 2004 around University of Tsukuba professor Yoshiyuki Sankai’s work, it developed the Hybrid Assistive Limb, or HAL, and listed on the Mothers market in 2014. Medical-device approvals, reimbursement and clinical use have taken it across national regulatory systems. Its history demonstrates both the symbolic power and the practical difficulty of turning an academic robot into an approved service delivered through hospitals and care systems.

TIER IV shows the leverage of open infrastructure. Shinpei Kato developed and released the Autoware autonomous-driving software at Nagoya University in 2015 and founded TIER IV the same year. The rights moved to the nonprofit Autoware Foundation in 2018. The open-source platform has been used across countries, vehicle types and hundreds of organizations, while the company builds commercial platforms and deployment services around the common code. The model seeks scale through a global developer and industry network rather than exclusive control of every layer.

These are not interchangeable recipes and not proof that the median university startup succeeds. They show why the metric must accommodate several forms of value: profitable platform licensing, regulated deployment, open-source standards, acquisition, manufacturing alliances and new scientific employment. A policy designed only to produce IPOs would misunderstand at least two of the three.

Global must begin before the company is ready

A company that waits for domestic success before considering the world often discovers that its product, evidence and organization were optimized for the wrong market. Clinical trial design determines which regulators will accept the data. Semiconductor specifications lock in supply chains. Enterprise software accumulates language, security and integration assumptions. The first board, first license and first customer can either preserve or foreclose international options.

Government programs increasingly recognize this. J-Startup selects companies for coordinated support. J-StarX sends entrepreneurs, students and investors into overseas ecosystems. Japan Innovation Campus opened in Silicon Valley in November 2023 to connect Japanese founders with American investors, companies and universities. In May 2026, Andreessen Horowitz said Tokyo would become its only overseas office, with plans to increase Japanese investment and eventually run a founder program.

Access, however, is not globalization. A delegation photo does not create a sales channel. University startups need bilingual commercial leadership, international patent strategy, globally credible boards, financial reporting investors can compare, and incorporation structures that do not create avoidable friction. Most of all, they need customers. The most useful overseas program may be the one that returns with no ceremony and a paid pilot.

International comparison requires the same discipline as domestic counting. In 2024, U.S. universities reported 962 startups created that year from licensed university technology—642 in the institution’s home state and 320 elsewhere—under a much narrower and annual measure. That number cannot be placed beside Japan’s broad cumulative 6,220 as though one country were ahead by simple division. It does show the industrial rhythm of a mature licensing system: American universities have recently generated around a thousand such companies in many individual years, alongside thousands of licenses and options.

The regional promise—and the regional financing gap

The latest survey offers one of its most encouraging signals beyond the total. Of 1,093 additions whose domestic location could be identified, 656—about 60 percent—were founded outside Tokyo, compared with roughly 54 percent in the preceding survey. University entrepreneurship is not merely reproducing the capital’s concentration. Growth at Nagoya, Kindai, Kansai, Kanazawa and Iwate points toward more varied local networks.

That distribution matters because research advantages are already regional. Tohoku has materials and semiconductors; Tsukuba combines national laboratories and a research city; Nagoya sits inside an advanced manufacturing corridor; Kansai connects medicine, pharmaceuticals, electronics and multiple major universities; Hokkaido and Kyushu can link science to food, energy and climate challenges. A startup cluster is strongest when it grows from capabilities that cannot simply be copied elsewhere.

But counting a company outside Tokyo does not prove that its capital and customers are local. Regional founders still travel to the capital for investors, specialist lawyers and corporate headquarters. Wet-lab space, experienced executives and follow-on funds remain uneven. The answer is not to duplicate every Tokyo institution in every prefecture. It is to build strong, specialized regional nodes connected by national capital and global markets—and to measure whether money, talent and procurement actually reach them.

From a formation dashboard to a scale dashboard

Counts were useful when university entrepreneurship was culturally and institutionally blocked. A target could mobilize TLOs, incubators, grants and administrators. Once thousands of companies exist, the same target can become misleading. It may reward the identification of dormant entities, encourage incorporation before customer evidence and make a university’s innovation office optimize its league-table position instead of portfolio quality.

Japan should continue publishing the count; long time series and transparent definitions are public infrastructure. It should place a second dashboard beside it. How many companies are operating after five and ten years? How much follow-on private capital did each public yen attract? How many products reached a customer, a production line or regulatory approval? What share of revenue comes from outside Japan? How many doctorate holders built long careers? How much licensing, dividend and equity value returned to universities? How many exits recycled founders and capital?

Formation-era measure	Scale-era question	Why it matters
Companies identified	How many remain active after five and ten years?	Separates incorporation from institutional durability.
Seed rounds and grants	How much follow-on private, strategic and project capital arrives?	Shows whether evidence is strong enough for the next risk stage.
Patents and licenses	How many products, approvals, pilots and manufacturing lines result?	Connects knowledge transfer to use in the real economy.
IPO or acquisition count	Did the exit capitalize growth and recycle talent and returns?	Measures ecosystem renewal, not ceremony.
Domestic company population	What share of revenue, customers, capital and leadership is global?	Tests competitiveness beyond a protected home market.
University ranking	What value returned to research, regions and skilled employment?	Keeps institutional incentives aligned with public purpose.

What 6,220 changes

The record does not mean that Japan has solved university commercialization. It means that the problem has changed shape. In 1998, the missing institution was often the office that could claim a patent and negotiate a license. In 2001, the missing act was incorporation itself. In the 2010s, the missing bridge was proof-of-concept capital and a specialized first investor. In 2026, all of those gaps still exist in places, but the system’s hardest deficiencies appear later: commercial leadership, large risk capital, demanding first customers, strategic M&A and the ability to operate globally.

METI’s wider 2026 startup ecosystem study estimates that startups directly generated ¥13.66 trillion—about 2 percent of nominal GDP—and supported a much larger total effect through supply chains and consumption. That estimate uses a broader startup universe, not the university cohort alone. It nonetheless explains why the 6,220 matter. University companies are not an academic side project. They are one mechanism by which publicly supported knowledge becomes high-skill employment, industrial resilience and future tax revenue.

Public purpose requires selectivity as well as ambition. Not every research project should become a startup. Some knowledge should be openly published; some inventions will reach society faster through a license to an established company; some technologies need a consortium or standards body; some student firms will remain excellent local businesses; many experiments should end. A mature ecosystem offers the right route, not the same route, for every discovery.

Okochi’s RIKEN network turned one institute’s science into dozens of companies through centralized industrial will. Modern Japan is building something more distributed: thousands of founders, universities, TLOs, funds, corporations and public agencies making repeated decisions under uncertainty. It will look messier, and it should. The important question is whether the system learns fast enough to move capital and people toward evidence.

Six thousand two hundred twenty is a threshold, not a finish line. Japan has demonstrated that a professor, postdoc or student can create a company without leaving the national innovation story. Now those companies must show what they can create after leaving the campus: medicines used in other health systems, materials designed into global supply chains, software that becomes infrastructure, factories and skilled jobs in regional cities, returns that finance the next laboratory.

The next record should not be merely 7,000. It should be a record number of university discoveries that survive the distance between an elegant result and a useful, trusted, globally purchased product. That is the scale test—and it is the one count Japan cannot afford to confuse.

Primary sources and further reading

The 6,220 figure is METI’s count of university startups identified as of the end of October 2025. It is a broad, cumulative survey measure; the 1,146 increase is not equivalent to startups newly incorporated in one year. Funding datasets and international measures use different definitions and are not treated here as directly comparable.