Gunma Looks for Its AI and Semiconductor Suppliers by Mapping 2,000–3,000 Companies

The overlooked companies behind the chip economy

Japan's semiconductor ambitions are usually described through wafer fabrication plants, advanced chips and national investment programs. Gunma Prefecture has started at a different end of the problem: it wants to identify the businesses that may already possess relevant capabilities, even if they do not describe themselves as semiconductor companies.

On November 11, the prefecture advertised a contract for an AI and semiconductor supply-chain survey, with a budget ceiling of ¥2.97 million including tax. It is a procurement for research, not a factory announcement, a subsidy award or a commitment by any manufacturer to enter the market. That distinction matters. The proposed mapping exercise is preliminary, but it asks a question that large investment announcements often leave unanswered: who can actually make the parts, materials and systems a new industry needs?

The official Japanese name is “群馬県内におけるＡＩ・半導体関連サプライチェーン調査,” administered by the Technology Development Section of the Manufacturing Innovation Office within Gunma's Regional Enterprise Support Division. Its purpose is to inform future prefectural policy. No contractor or study findings had been announced as of this report.

A surprisingly specific request: screen 2,000 to 3,000 businesses

The technical specification reveals considerably more than the short procurement headline. The contractor is directed to search existing business databases using combinations of keywords, industrial classifications and customer relationships, producing an initial pool of roughly 2,000 to 3,000 companies with a presence in Gunma.

It must then examine the descriptions of those businesses manually and distinguish companies judged highly likely to have an AI or semiconductor connection. An initial search hit is therefore not proof of industry participation. A first-stage dataset and a screened company dataset are both required.

The deliverables go beyond names and addresses. Required fields include the location and branch status of each business, establishment date, principal products, capital, workforce, three years of sales and equity-ratio information, leadership and a credit rating or score. That design may help policymakers understand the depth and financial profile of a potential supplier base. It does not amount to an engineering qualification audit: the specification does not explicitly require factory visits, product testing or interviews with customers.

Nor does the specification establish how many companies will pass the second screening. The number 2,000–3,000 is a target for the initial search, not a published count of semiconductor suppliers in Gunma.

Why Gunma's industrial history matters

Gunma is sometimes introduced internationally through hot springs and mountain scenery. Its industrial history offers a more useful explanation of why the prefecture might conduct this research. At Tomioka, the Japanese government established an industrial model that imported machinery and expertise but also trained people to spread the new methods across Japan. The Tomioka Silk Mill began operations in 1872, bringing mechanized silk production into a system designed for industrial learning.

A later generation of manufacturing emerged around Ota. According to Subaru's corporate history, Chikuhei Nakajima founded an aviation research operation in 1917 in what is now the Ota area. Nakajima Aircraft became an important prewar enterprise; the postwar corporate evolution eventually led to Fuji Heavy Industries and today's Subaru. The company's official corporate profile dates Subaru's foundation in its present historical lineage to that 1917 research venture, the establishment of Fuji Heavy Industries to 1953 and the Subaru name to 2017.

Neither silk-reeling machinery nor automobile assembly automatically qualifies a firm to supply semiconductor fabrication equipment. The lesson is narrower and more meaningful: Gunma has repeatedly developed industrial skills around complex production, supplier relationships and manufacturing quality. Whether any specific expertise can be repurposed requires evidence, not nostalgia.

The economic weight behind the proposal

A prefectural investment briefing published in October 2025 reported Gunma manufacturing shipments of approximately ¥10.1485 trillion, based on an earlier industrial survey. Transport equipment accounted for ¥3.7532 trillion, by far the largest specified segment in the briefing. Food manufacturing, chemicals, electrical equipment and metal products also form important parts of the industrial base.

Those figures are historical manufacturing totals, not 2026 production, and they are not measurements of AI or semiconductor revenue. They nevertheless show why a small research contract might merit attention. If even a subset of existing suppliers has specialized precision machining, materials handling, industrial control or measurement expertise relevant to new buyers, the policy question is how to discover and evaluate it.

Gunma's challenge is partly one of industrial visibility. A firm may be categorized by the product it sells today, not by the applications its engineering could support tomorrow. Equally, a company may use semiconductor-related vocabulary in its marketing without being equipped to meet a demanding customer's purchasing and quality standards. A credible survey must guard against both errors.

Where an automotive supplier might intersect with semiconductors

The chip industry is much larger than chip design and the cleanroom itself. Equipment makers rely on complex assemblies; material suppliers serve tightly controlled processes; metrology, testing, automation, transport, temperature control and maintenance all create demand for specialized products. AI hardware adds computing systems and power requirements, while so-called physical AI brings machine perception and automated action into factories and other real-world environments.

These are examples of possible intersections, not evidence that Gunma has already secured business in each category. The prefecture has not published a definitive priority list of industrial subsegments or identified successful new entrants.

The difference between technical similarity and commercial qualification is substantial. Semiconductor customers may demand distinctive cleanliness conditions, traceability, process consistency, reliability testing and documented quality systems. A regional metalworking or electronics business could have promising capabilities and still face high certification costs or long sales cycles. Mapping a supplier is the beginning of that process, not the end.

A national strategy searching for a local base

The Ministry of Economy, Trade and Industry revised Japan's Semiconductor and Digital Industry Strategy in June 2026. Its explanation treats AI and semiconductors as part of a larger ecosystem spanning data, models, computing infrastructure, communications, electricity, talent and cybersecurity. Physical AI figures among the technological changes reshaping that agenda.

Gunma's proposal should not be confused with the national program's scale or authorization. The local survey is designed to inform prefectural policymaking. It does not, by itself, secure a national grant, allocate manufacturing subsidies, or announce a partnership with a particular chipmaker.

The possible policy sequence is straightforward, though not guaranteed: establish what industrial competencies exist, identify gaps, connect relevant suppliers with potential customers, and consider what technical support or workforce development might be useful. Each of those later stages would require its own decisions, resources and public accountability.

The practical limits of a database map

The specification's manual review requirement is an important safeguard against declaring every keyword match an AI company. But human review of business descriptions can still miss expertise that is commercially confidential, poorly documented online, or buried several tiers down a supplier network.

A hypothetical precision-machining company may manufacture components for industrial systems without mentioning semiconductor applications. Another firm might name semiconductor equipment on its website while having only a modest sales history in the field. Neither example describes a verified Gunma business; both illustrate why classification and capability are different things.

It is also unclear whether the eventual research will identify the connections between first-tier contractors and smaller subcontractors. Detailed supplier relationships may be confidential. A public company directory could be useful, but the specification includes financial information, credit scores and confidentiality provisions, making blanket public disclosure unlikely to be an automatic outcome. No such disclosure has been promised.

A procurement calendar, not a completed industry transformation

The application-to-participate deadline is November 18, 2026, at 5 p.m. Japan time. Full proposals are scheduled to be due November 25 at 5 p.m., and the announced selection timetable points to results beginning December 1. The contracted work would run until February 26, 2027. These are tender milestones; they are not publication dates for a completed industry survey.

As of the November 13 research cutoff, there is no verified survey winner, no completed shortlist of firms and no newly announced semiconductor plant attached to this contract. Prospective bidders should confirm all timetable details directly with the prefecture.

Another open question is what will be released to the public. The specification calls for electronic company-information files delivered to the prefecture; it does not promise a public dashboard, open dataset or list of qualified suppliers.

What a serious follow-up should measure

The first useful result will be how many of the initially identified 2,000–3,000 businesses survive manual screening, and why. A rigorous next step would ask whether the categories correspond to demonstrated products, customers or technical standards. Regional distribution may reveal clusters around familiar industrial centers, but that is an empirical question for the survey rather than a conclusion to write in advance.

A second set of questions concerns barriers: are potential entrants constrained by capital equipment, workforce shortages, quality assurance, export requirements or the difficulty of finding buyers? A third concerns public benefit: will the government use the map to build matchmaking programs, improve training or prioritize support for particular capabilities?

These questions are a reporting agenda, not promises made by Gunma officials. The difference between a list of businesses and useful industrial policy will be whether the results are checked against real market demand and followed by measurable action.

The industrial future is not announced; it is investigated

The story of Tomioka was not simply about importing a machine. It was about building the skills and institutions to operate it. The story of Ota was not simply about one factory. It was about sustained, complicated production networks that evolved through several industrial eras. The emerging AI and semiconductor economy poses another question about adaptation, but the historical analogy has limits.

At ¥2.97 million, Gunma's survey is small enough to disappear beneath the headline figures in semiconductor policy. Its significance may lie in whether it gives the prefecture a disciplined picture of what businesses can already do, and what they would need in order to serve future markets.

For now, the sound conclusion is modest: Gunma has opened a procurement to investigate its industrial links to AI and semiconductors. Which companies are genuinely positioned to participate—and what the prefecture will do with the evidence—remain the next stories to report.