Who the Customer Is

The line sits at the end of the segment note. In its most common form, the 【セグメント情報等】 — segment information — note in the consolidated financial statements of a Japanese annual securities report closes with 関連情報, related information, printed in three parts: sales by product and service, sales and fixed assets by geography, and 主要な顧客ごとの情報, information by major customer. Where a single external customer accounts for 10% or more of the revenue in the consolidated income statement, the third part prints the customer's name, the sales to it and the segment those sales belong to; where no customer reaches the threshold, it prints a sentence saying so. Other filers head the item 主要な顧客に関する情報, information on major customers, and one, Nippon Steel, 主要な顧客に対する売上収益, revenue from major customers. A second table sits much earlier in the same document, in 第２ 事業の状況 among the production, order and sales figures: 主な相手先別の販売実績及び当該販売実績の総販売実績に対する割合, sales to principal counterparties and their share of total sales, and that table prints a percentage the note does not.

This desk read the segment note in the June 2026 reports of thirty-five of the largest filers, and the counterparty table wherever the note named a customer. Thirty-one of the thirty-five print a statement on major customers. Eight report a customer at 10% or more: seven name it, and one, HOYA, describes a customer group without naming it in the note while its counterparty table names Seagate Technology LLC. Twenty-three state that no customer reaches the threshold. In the remaining four — Toyota, SoftBank Group, Honda and Central Japan Railway — the segment note, as extracted from the filed PDF, carries no customer item among its parts.

The customers named across the eight reports are these. Tokyo Electron prints Samsung Electronics Co., Ltd. and Taiwan Semiconductor Manufacturing Company Ltd.; Advantest prints the NVIDIA group and Taiwan Semiconductor Manufacturing Co., Ltd., and for the year before, TSMC and the Samsung group; DENSO prints the Toyota group; FANUC prints SHANGHAI-FANUC Robotics CO., LTD.; Daiichi Sankyo prints McKesson, Cencora and Alfresa Holdings with its group companies; Takeda prints McKesson and Cencora; Mitsubishi Heavy Industries prints 防衛省, the Ministry of Defense; and HOYA's counterparty table prints Seagate Technology LLC. TSMC is named in two reports and Samsung in two, McKesson in two and Cencora in two. One customer is a ministry, one is an affiliate in which the filer holds half the voting rights, and one, Toyota, is itself among the thirty-five filers.

Tokyo Electron's note, on page 127, prints two rows for the fiscal year ended March 31, 2026: Samsung Electronics Co., Ltd. at ¥368,079 million (US$2,247 million) and Taiwan Semiconductor Manufacturing Company Ltd. at ¥315,813 million (US$1,928 million). The prior-year rows on page 126 read ¥286,800 million and ¥280,618 million, and a note under each table says the sales include customers belonging to the same corporate group as the named customer. The counterparty table on page 23 prints the same four amounts with their shares of total sales: Samsung 11.8% a year earlier and 15.1% this year, TSMC 11.5% and 12.9%. The company reports as a single segment, so the note names no segment.

Advantest's note, page 113, is written as prose rather than a table. For the year ended March 31, 2026 the customer groups at 10% or more are the NVIDIA group and Taiwan Semiconductor Manufacturing Co., Ltd., both mainly in the test-system business, at ¥243,735 million (US$1,488 million) and ¥124,922 million (US$763 million); for the year before they were TSMC and the Samsung group, at ¥96,158 million and ¥87,734 million. The counterparty table on page 42 prints companies rather than groups: NVIDIA INTERNATIONAL, INC. at ¥228,273 million (US$1,394 million) and 20.2% of sales, with a dash for the year before; Taiwan Semiconductor Manufacturing Co., Ltd. at ¥124,922 million and 11.1% (¥96,158 million and 12.3% a year earlier); and Samsung Electronics Co., Ltd. at ¥82,795 million and 10.6% a year earlier, with a dash this year. The TSMC figure is the same in both places. The NVIDIA and Samsung figures are not, and the difference between ¥243,735 million and ¥228,273 million is the difference between a group and one company in it.

DENSO's note, page 110, names the Toyota group as its major customer and says the revenue arises in every segment: ¥2,131,303 million in Japan, ¥1,056,980 million in North America, ¥141,270 million in Europe, ¥763,681 million in Asia and ¥41,348 million in Other, and the table beneath prints ¥3,903,121 million for the year ended March 2025 and ¥4,134,582 million (US$25,242 million) for the year ended March 2026. The counterparty table on page 35 prints a different name and a smaller number: トヨタ自動車㈱, Toyota Motor Corporation alone, at ¥1,991,604 million (US$12,159 million) and 26.4% of sales, against ¥1,902,231 million and 26.6% a year earlier. The same customer appears twice more. The related-party note on page 159 lists the Toyota group as an entity with significant influence, with sales of automotive parts of ¥4,134,582 million, purchases of ¥60,400 million and accounts receivable (売掛金) of ¥405,112 million (US$2,473 million) at March 31, 2026; and the list of affiliated companies on page 12 carries Toyota Motor Corporation as その他の関係会社, an other affiliated company, holding 22.25% of DENSO's voting rights.

FANUC's note prints a customer for the first time in the two years the report covers. The prior-year section on page 78 says no customer reached 10%; the current-year section on page 79 prints SHANGHAI-FANUC Robotics CO., LTD. at ¥89,639 million (US$547 million), with a note that the company is a single segment and so names none. The counterparty table on page 18 gives the share, 10.4%, and a dash for the year before with a line saying the prior-year share was under 10%. The customer is not an outside company in the ordinary sense: page 8 lists SHANGHAI-FANUC Robotics CO., LTD., of Shanghai, among FANUC's 関連会社, affiliates, with 50 in the column for voting rights held and directors serving concurrently, and page 58 says the equity method is applied to it and to BEIJING-FANUC Mechatronics CO., LTD.

HOYA's note, pages 104 and 105, says that the segment with a customer group at 10% or more of revenue is 情報・通信, information technology, and that revenue from that group totalled ¥92,776 million in the prior year and ¥111,393 million (US$680 million) in the current one. It does not name the group. The counterparty table on page 29 prints the same two amounts against a name, Seagate Technology LLC, with shares of 10.71% and 11.75% — the one table among the eight that prints its percentages to two decimals.

Two pharmaceutical filers name the same two American wholesalers. Daiichi Sankyo's note, page 127, prints three rows: マッケソン社, McKesson, at ¥203,461 million and ¥269,418 million (US$1,645 million); センコラ社, Cencora, at ¥207,389 million and ¥251,034 million (US$1,533 million); and アルフレッサ ホールディングス株式会社及びそのグループ会社, Alfresa Holdings and its group companies, at ¥221,814 million and ¥232,433 million (US$1,419 million). Its counterparty table on page 43 puts the shares at 12.7%, 11.8% and 10.9% (10.8%, 11.0% and 11.8% a year earlier). Takeda's note, page 171, prints McKesson Corporation and its group companies at ¥592,323 million and ¥539,890 million (US$3,296 million), and Cencora Inc. — the note adds its former name, AmerisourceBergen Corporation, in parentheses — and its group companies at ¥577,017 million and ¥470,295 million (US$2,871 million); the counterparty table on page 35 puts the shares at 12.9% and 12.0% for McKesson and 12.6% and 10.4% for Cencora.

Mitsubishi Heavy Industries' note, page 99, names 防衛省, the Ministry of Defense, as the customer at 10% or more, attributes the revenue mainly to the 航空・防衛・宇宙 segment — aircraft, defense and space — and prints ¥704,181 million for the year ended March 2025 and ¥1,006,022 million (US$6,142 million) for the year ended March 2026. The counterparty table on page 25 puts the shares at 16.1% and 20.2%, and adds that the amounts exclude consumption tax. Of the thirty-one statements this desk read, Mitsubishi Corporation's, on page 146, is the one whose negative names the category: no single customer, customer group or government agency exceeded 10% of revenue in either year.

The twenty-three negatives are not one sentence. The formula six of them share — Keyence, Shin-Etsu Chemical, Daikin, SMC, Oriental Land, East Japan Railway — is that no counterparty accounts for 10% or more of the sales in the consolidated income statement, and so the item is omitted or not stated; Nintendo drops the clause about omission and prints that there is no such counterparty. MUFG's item reads 該当事項はありません, not applicable, and nothing more. Sumitomo Mitsui Financial Group and Mizuho Financial Group measure the threshold against 経常収益, ordinary income, in place of sales. KDDI states it the other way round: sales to any specific customer were under 10%. Elsewhere Kyocera writes of a single external customer, Panasonic Holdings of a single counterparty exceeding 10%, Murata Manufacturing of a counterparty at 10% or more, Nippon Steel of a customer that does not exist, Recruit Holdings of a counterparty in either year.

Five of the thirty-one put the sentence in the body of the note without a heading of its own: Sony Group's, on page 161, follows the geographic information and says there were no sales to a single customer that was significant in fiscal 2024 or 2025; Mitsui & Co.'s, page 193, and ITOCHU's, page 146, say there was no significant revenue from a single customer; NTT's, page 130, that no operating revenue of 10% or more of total revenue arose from transactions with a single external customer; Mitsubishi Corporation's is the sentence quoted above. Hitachi heads its item 顧客別情報, information by customer, on page 130; FUJIFILM Holdings heads its 主要顧客及びその他情報, major customers and other information, on page 195. A reader searching the PDF for 主要な顧客 finds twenty-four of the thirty-one; the other seven are found by searching for 単一, single, for 主要顧客 without the particle, or for the 10% itself.

For the five companies this fragment was asked about, the answers divide three to two. Keyence's note, pages 53 and 54, prints the formula for both years, and its sales section on page 15 says the same thing in the older wording — no counterparty at ten hundredths or more of total sales, out of sales of ¥1,169,289 million — for a fiscal year that ends on March 20. Murata Manufacturing's note, page 126, and its sales section, page 52, both say no counterparty reaches 10%. Kyocera's note, page 102, says no single external customer accounts for 10% or more of sales. Tokyo Electron and Advantest print the names set out above. The threshold is the same in all five reports, so what the two names and three sentences describe is where the customers sit against it in the year reported, and not the shape of the customer list below it.

The four reports whose segment note carries no customer item are read differently. Toyota's note 5, pages 145 and 146, has three parts — an overview of the reportable segments, the segment figures, and 地域に関する情報, information by region — and closes with a sentence saying that the segment-split financial statements, the regional information and the revenue by customer location formerly in note 5 are, from the fiscal year ended March 2026, given in the management analysis in 第２ 事業の状況. SoftBank Group's note 6, pages 148 to 151, has three parts of the same kind. Honda's note 4, pages 138 to 141, has four, the fourth on the effect of the change in the EV market environment on the automobile business. Central Japan Railway's 関連情報 on page 116 has two numbered items, products and services and geography, before the next heading. In none of the four did this desk find a major-customer statement in the note; whether one sits elsewhere in the report is not something the extraction can rule out, and the text of this fragment does not claim it.

SoftBank Group's report speaks of customer concentration in another place. Its risk section, pages 29 and 30, says that most of Arm's revenue depends on a small number of major customers and can be affected by their business trends, that Arm develops several processors a year to soften the effect of a particular customer passing on a product, and that Arm is party to litigation with Qualcomm, described there as a current major customer, and with Nuvia. That is a risk factor about a subsidiary, written in the company's words, and not a figure in the segment note; the two should not be read as the same disclosure.

Where a report prints both the note and the counterparty table, the two can be read against each other. In six of the eight — Tokyo Electron, FANUC, HOYA, Daiichi Sankyo, Takeda and Mitsubishi Heavy Industries — the amounts match to the million yen and the table adds the percentage. In two — Advantest and DENSO — the table names a company where the note names a group, and the amounts differ. Neither is the other's correction: the note is answering the question of which customer group reached 10%, and the table which counterparty was billed. A reader who wants the percentage must go to the front of the report; a reader who wants the segment must go to the back; a reader who wants to know whether the customer is an affiliate, a shareholder or a ministry must go to the list of affiliated companies, the related-party note or the name itself.

The method was this. extract_customers.py opened the thirty-five PDFs with pdfplumber, removed all whitespace from each page's text and searched for 主要な顧客, keeping the page, the surrounding text and every table on the page that mentioned customers, counterparties or sales; that pass hit pages in twenty-five reports — twenty-four of them the statement itself, the twenty-fifth NTT's, where the words occur only in the auditor's key audit matter — and, in Nippon Steel's case, first found the four rows of its cross-shareholding table whose reason for holding begins 同社は、当社の主要な顧客であり, the company is a major customer of ours, before the segment note on page 131. A second pass over the remaining eleven searched for 単一顧客, 単一の外部顧客, 10% within sixty characters of 顧客, and the sub-headings of the segment note; it found seven statements and left four notes to be read item by item. The amounts and percentages quoted above were read from pdfplumber's tables by column heading, with whitespace kept, and compared with the page text. Yen amounts are as printed, in millions; the dollar figures are conversions at ¥163.8 to the dollar for reference only.