Fewer Machines, More Money

The proposal, as the company's own centennial history tells it, was a small one: put a motor on a hand press, so that an operator could do 200 kilograms of pressing work on household current without putting it through their back. This was the early 1980s, inside Janome Electric, whose motor work until then had been OEM business — coffee mills, can openers. Somewhere in development the press acquired a computer. The sewing machine's universal motor gave way to a servo motor with an encoder resolving 800 pulses per revolution, a load cell became standard, and the finished machine controlled pressing speed, pressing force and position, about a hundred setting combinations across the three. It went on sale in April 1984 as the JP-20 electro-press at ¥980,000 — about US$6,000 at the single 2026 rate used throughout this piece, which is not what a dollar bought in 1984. The sales side, per the same history, said it would be hard to sell.

Janome's industrial equipment division describes itself, in its own words, as a veteran maker, headlining roughly forty years of servo presses and thirty years of desktop robots. Both are round headline numbers given without a base year, on a page that carries no publication or revision date when retrieved on 27 July 2026; the implied start points, 1984 and 1993, are inferred rather than stated. The 1993 end of that inference is at least anchored — the division says its first desktop robot, the JR500, was developed and put into volume production that year. Its account of where all this came from is unusually flat: while bringing sewing machine parts in-house, it says, it took up developing the equipment needed to make those parts stably and better. That is a sentence about vertical integration that turned into a product catalogue, and, like everything else on the page, it is the company describing itself.

Start with the name, because the name is a mechanism it borrowed rather than built. Standard lockstitch machines of Singer's Class 15 type carried a half-rotary hook whose bobbin looked, to people in the trade, like a janome — a snake's eye, the concentric ring pattern — and the trade used janome-type to mark those machines off from the older long-shuttle design. By the mid-1930s the term was shorthand for the newest household machines and belonged to nobody. The company, whose own March 1929 prototype, the Pine 100-series Model 30, was a machine of exactly that class, put the name on its products from 1934, applied to register it that year, and was granted the mark in November 1935. In the same month it renamed itself Imperial Sewing Machine. Its own history credits the founder not with coining the word but with trademarking a generic one before anyone else thought to. The mark became the corporate name in 1949, took a longer form in 1954, and was shortened to Janome in October 2021, a hundred years after the founding of a workshop that took its brand from the pine in two founders' surnames.

The hook is the part of a sewing machine that is seldom seen and finely timed: it holds and releases the lower thread in time with the needle. The company's published lineage runs long-shuttle; then, from 1929, domestic production of the vertical half-rotary hook, a mechanism it did not originate; then a full rotary hook for home machines — research opened in 1964, completed at the end of 1967, first fitted to an export model in May 1970 and to a domestic machine in April 1971, which also widened maximum stitch swing from the 5 millimetres of rival machines to 7, by the company's account. In 1976 the head of its research institute took the Science and Technology Agency Director-General's Award for that hook, the agency being the one folded into the education ministry in the 2001 reorganisation. In 1979 a horizontal hook arrives, adopted in the Memoria 5001, with no bobbin case and easier to clean; the company's own history handles it in a separate section rather than as the next generation.

The other lineage is control. Zigzag research began in a factory research room in 1952 and reached a fully automatic domestic model in 1961. Then, in April 1979, the Memoria 5001: a computer sewing machine listed at ¥188,000 — roughly US$1,150 at the 2026 rate, though the yen figure is a 1979 list price adjusted for nothing — which took the motion of the needle and the feed dog, previously set by dials and levers, and put it under program control. The stepping motor software was written, according to the history, with access over a communications satellite to a General Electric mainframe in the United States. By 1983 the machine had passed 500,000 units.

Here the accounts diverge, and both of them are the company. The corporate history page calls the Memoria Japan's first computer sewing machine. The centennial book keeps that framing and adds a wider one: the world's first machine you could program to sew. The same book records that in 1975 a top global sewing machine maker, unnamed, had already announced a machine generating stitch patterns by computer, and argues that such machines did no more than simplify existing operation, so Janome's was a different thing. For the book's two firsts the source notes no third-party certification; the corporate page's Japan-first is likewise the company describing itself. Anyone repeating the world-first version in English is repeating the more generous of two in-house accounts.

So: a needle whose position is computed, and a hook that has to arrive in the right place at the right moment. The company's own explanation of the desktop robot is that it applied positioning control developed for the embroidery function. Its explanation of the robot frame is that it applied sewing machine casting technique to get very high rigidity, which it presents as the precondition for precision — no stiffness figure, no tolerance, the claim left qualitative. Its integrated report, published in October 2025, puts the whole thing in a line on page 7: technique cultivated in sewing machines, applied, widening the business.

The executive responsible for development describes the transfer differently. The formulation there is that the second pillar was built on an easy-to-use user interface derived from sewing machines — not the mechanism, the interface. That is a quieter claim than the mechanical genealogy, and possibly the more accurate one. The official company chronology, read plainly, is a company steadily repurposing its own equipment engineering: stepping motors in 1979, a vacuum casting rig in 1980, the servo press in 1984. The desktop robot of 1993 comes from a different document, the product trajectory on page 11 of the 2025 integrated report, where the JR500 opens the robot line; the chronology page's robot entry could not be confirmed word for word.

What it sells now is two desktop robot series. The JR4000 comes in standard, board-depaneling, AP-D-option and automatic screw-tightening builds; the JR3000 in standard, camera-equipped dispensing, automatic screw tightening, twin-table, tangless-insert insertion, board depaneling, high-payload, and a benchtop multi-function inspection unit. The processes the page names — dispensing, assembly, screw tightening, board depaneling, soldering, inspection — are offered as examples rather than a closed list, and not every one has a matching machine in the lineup. The pitch is four-axis simultaneous control, teaching by dialogue entry with application-specific software, an option that adds up to two further axes without a PLC on models the page does not identify, and a teaching pendant that displays in eleven languages. Eleven is the only figure the page defines precisely. Precision is asserted without a tolerance, IoT and MES compatibility without a protocol.

The market it sits in is large and violently cyclical. Japan's robot industry association counted 2025 orders of ¥1,045,605 million — about ¥1.05 trillion, or US$6.38 billion — up 25.7 percent on 218,987 units, after two down years from the 2022 peak. Exports were 79.5 percent of shipment value and 82.1 percent of units. Domestic shipments fell 10.8 percent in value and have declined in units three years running. The headline recovery and the domestic contraction are the same year.

Inside those totals, the segment closest to a bench-top dispensing robot is close to unmeasurable. Electronic component mounting — a category counted separately from assembly, not inside it — was 17,192 units, 8.1 percent of units; the association discloses its value only in a supplementary trend chart, where export value alone is ¥296,435 million, about US$1.81 billion. The assembly subtotal was 25,936 units and ¥46,128 million, US$282 million, 12.3 percent of units and 4.6 percent of value. Below the subtotal line the association publishes units only: sealing and gluing 874, screw tightening 86, soldering 7. So no market size can be derived from this survey for the applications the product page lists; the 2024 reclassification broke the series; and how much of the desktop specialists' shipments the survey captures cannot be determined from it.

The home side, meanwhile, is mostly no longer made here. Japan produced 41,140 household sewing machines in 2025, worth ¥2,095 million, about US$12.8 million. That is a third below 2016 — 33.3 percent fewer units, 33.5 percent less value, both computed here from the association's table — though 2025 was itself a 13.5 percent rebound off the 2024 floor, so the decline is not a straight line. Imports ran to 620,756 units, roughly fifteen times domestic output, on trade data that is not strictly comparable with production data; they are also well below their 2020 peak, and 2025's 44 percent jump in units came with a 2.3 percent rise in value, which is to say cheaper machines. Of the sewing machinery association's 32 regular members, seven list household machines among their product categories — a tally taken from the member directory on a duplicate-count basis, not a figure the association publishes.

That company's own year, ended March 2026, went like this. Revenue ¥38,968 million — the securities report and the results release both state their unit as millions of yen — about US$238 million, up 7.2 percent. Gross profit ¥15,110 million, US$92.2 million, up 0.5 percent, so gross margin fell from 41.4 to 38.8 percent, a figure not printed on the page but implied by it. Operating profit ¥1,910 million, US$11.7 million, down 14.1 percent. Ordinary profit ¥2,097 million, US$12.8 million, down 7.2 percent. Net profit attributable to owners ¥590 million, US$3.60 million, down 67.1 percent. The dividend went from ¥40 to ¥55 per share, US$0.24 to US$0.34. The financial highlights page states none of the reasons.

The balance sheet barely moved: equity ratio 68.4 percent, return on equity down from 5.2 to 1.7 percent on the smaller net figure, overseas revenue 74.7 percent of the total. Capital expenditure fell 64.1 percent to ¥388 million, US$2.4 million, against depreciation of ¥903 million, US$5.5 million — a capex-to-depreciation ratio of 0.43, down from 1.06 the year before. Research and development held at ¥1,480 million, US$9.0 million, 3.8 percent of sales. Which is what the previous paragraph was about: that single research line runs to roughly seventy percent of the ¥2,095 million, US$12.8 million, of household sewing machines produced in all of Japan in calendar 2025. Different periods, different statistics, set side by side here for scale rather than as a like-for-like comparison.

The reasons arrived a year later, in a bar chart with four bars. The company's own bridge runs from ¥2,224 million of operating profit, US$13.6 million, to ¥1,910 million, US$11.7 million: higher sales added ¥934 million, US$5.70 million, of gross profit; the cost ratio took back ¥300 million, US$1.83 million; selling and administrative expenses took another ¥300 million, US$1.83 million; and currency took ¥647 million, US$3.95 million. The bars sum to ¥313 million, US$1.91 million, against a printed decline of ¥314 million, US$1.92 million, which is rounding. What they do not do is separate volume from price from mix — everything sold, at whatever price, in whatever combination, sits inside the first bar. Currency gets its own bar and its own breakdown: ¥108 million, US$659,000, added to sales, ¥671 million, US$4.10 million, added to cost of sales, ¥84 million, US$513,000, added to expenses. The same twelve months appear below the operating line as a ¥116 million exchange gain, US$708,000, where the year before there was a ¥74 million loss, US$452,000, and in equity as an ¥1,811 million increase in the translation reserve, US$11.06 million. One currency year, three signs. The only rate the company prints is the dollar, ¥152.56 falling to ¥150.77, which is the yen getting stronger; the account given at the results briefing was that the currencies it prices in strengthened against the yen — the yen fell against them, which lifted sales — while the local currencies of its plants in Taiwan and Thailand rose as well, which lifted cost of sales and expenses. Which currencies those are, it does not say.

Now the number the year turns on. Janome sold 810,000 sewing machines against 890,000, on a table rounded to the nearest ten thousand that puts the fall at 84,000 — so somewhere between 9.0 and 9.4 percent fewer machines, depending on which of the company's own figures you use. Household equipment revenue went the other way, from ¥28,773 million, US$176 million, to ¥29,787 million, US$182 million, up 3.5 percent. Divide the one by the other, which the company does not: revenue per machine sold rose from about ¥32,300, US$197, to about ¥36,800, US$225, roughly 14 percent. It is an index rather than a price — the numerator is a segment, the denominator is machines — but the direction is not in doubt, and neither is what is absent. Price, mix and currency cannot be separated from this disclosure. Currency can at least be bounded: exchange movement added ¥108 million, US$659,000, to group sales, 4 percent of the ¥2,627 million, US$16.0 million, the group added. The company's account of the rest is two sentences: North America and Oceania sold high- and mid-grade machines well, and low-priced Chinese product made the CIS and Asia harsher. Segment operating profit fell anyway, from ¥2,159 million, US$13.2 million, to ¥1,854 million, US$11.3 million. The regional unit changes it prints — North America up 11,000, Oceania up 2,000, Latin America up 3,000, the Middle East down 5,000, Europe down 55,000, Asia down 56,000, Japan up 17,000, or up 10,000 if you subtract the printed levels — come to 83,000 against a stated 84,000, and where Africa is counted is not said.

Of the ¥2,627 million the group added, US$16.0 million, industrial equipment supplied ¥1,314 million, US$8.02 million; household equipment ¥1,013 million, US$6.19 million — the levels subtract to ¥1,014 million, US$6.19 million, the company prints ¥1,013 million; IT ¥366 million, US$2.23 million; and the property line gave back ¥67 million, US$409,000. So half the growth came from 15.8 percent of the revenue. Industrial equipment is also the one place where units and money are printed for the same objects, though not in the same table: the unit counts are in the securities report, the revenue in the securities report and the results deck alike. Robots: 1,295 to 2,855, up 120 percent. Presses: 865 to 1,045, up 20.8 percent. Together 2,160 machines to 3,900, up 80.6 percent, on revenue of ¥2,758 million, US$16.8 million, rising to ¥3,762 million, US$23.0 million, up 36.4 percent. Divide again — the company does not — and revenue per machine falls from about ¥1.28 million, US$7,800, to about ¥965,000, US$5,890, down 24 percent. That could be a lower price or a different mix and the file cannot tell you which: robots went from 60 to 73 percent of the units, and revenue is not published for robots and presses separately, so the average falls on mix alone even if nothing was discounted. What is certain is the shape. Robot shipments more than doubled, and robots and presses together collected 36 percent more money.

The three-year plan running to March 2028 asks for ¥43.5 billion of revenue, US$266 million, against ¥36.3 billion, US$222 million, in the year it was written from; a 9.2 percent operating margin; and 8.1 percent return on equity, against a self-assessed cost of equity of 8 percent. Home equipment is to go from ¥28.8 billion to ¥31.4 billion, US$176 million to US$192 million. Industrial equipment is to go from ¥4.8 billion to ¥7.9 billion, US$29 million to US$48 million — a 65 percent increase, and 43 percent of the entire planned revenue increment. The company's own review of the predecessor plan records revenue, profit and return on equity all short of target. All of it is a target rather than an outcome.

The segment carrying that increment is currently two businesses in one line item: robots and presses at 57 percent of its ¥4,841 million, US$29.6 million, and die casting at 43 percent. The integrated report discloses five consecutive years of segment operating losses through March 2025, at margins of −6.5, −2.1, −1.6, −4.7 and −8.7 percent, and the president writes that turning it positive is the commitment of the new plan. The report also lists what the company considers its own weaknesses in the segment: price competitiveness constrained by high prices, revenue volatility from dependence on autos and electronics, and a low in-house parts ratio. The division's origin story, as told on its own website and in the centennial history, is a decision to make Janome's parts inside Janome.

Split the segment further, because the company now does. Robots and presses: ¥3,762 million, US$23.0 million, of revenue and ¥88 million, US$537,000, of operating profit, against an ¥85 million loss, US$519,000, the year before, which is the first profit in the disclosed run. Die casting: ¥2,392 million, US$14.6 million, and a loss of ¥632 million, US$3.86 million, against ¥337 million, US$2.06 million — a margin of minus 26.4 percent, computed here. Together they lose ¥543 million, US$3.32 million, at minus 8.8 percent, a tenth of a point below the minus 8.7 before it: a fourth consecutive year of negative margin in the deck's own run — minus 1.6, minus 4.7, minus 8.7, minus 8.8 — and one more loss year after the five the integrated report lists through March 2025. The company's account of the die casting loss is a cost ratio that will not come down and freight, with raw materials and labour added in the annotation to the bridge chart, and no volume or price attached to any of them; it says price revisions and cost reduction have been started. The forecast for the year to March 2027 puts the segment at ¥240 million, US$1.47 million, of profit. That is a swing of ¥783 million, US$4.78 million, inside a planned group operating profit increase of ¥1,090 million, US$6.65 million, from ¥1,910 million, US$11.7 million, to ¥3,000 million, US$18.3 million: 72 percent of the improvement, from 15 percent of the revenue. The same segment forecast is printed as ¥6,145 million, US$37.5 million, on one page of the results deck and ¥6,200 million, US$37.9 million, on another. Meanwhile the IT business, 7.4 percent of revenue, earned ¥537 million, US$3.28 million, 28 percent of group operating profit, and is called a record without a date attached.

Against ¥590 million of net profit, US$3.60 million, the company declared ¥958 million, US$5.85 million, of dividends and bought ¥1,406 million, US$8.58 million, of its own shares: ¥2,364 million, US$14.4 million, in all. It prints the total return ratio as 433.7 percent; dividing the ¥2,364 million by the ¥590 million gives 400.7 percent, and the page does not reconcile the two. Operating cash flow was ¥1,780 million, US$10.9 million, down from ¥2,625 million, US$16.0 million, on a smaller pre-tax profit and a tax payment that roughly doubled. Investing turned positive at ¥170 million, US$1.04 million, because ¥1,607 million, US$9.81 million, of property was sold while ¥398 million, US$2.43 million, was spent on it. Financing drew ¥887 million, US$5.42 million, of net short-term borrowing, in the year after repaying ¥1,360 million, US$8.30 million. Cash rose ¥1,069 million, US$6.53 million, of which ¥540 million, US$3.30 million, is translation rather than anything that happened. The declared dividend is ¥55 and the plan is ¥60, US$0.34 and US$0.37; the payout ratio is 164.7 percent on the consolidated result and 295.5 percent on the parent company's; the dividend-on-equity ratio that policy sets at 3 percent or more came in at 2.8. On 12 May 2026 the board resolved to cancel 1,129,400 shares, 6.2 percent of those outstanding, with 29 May set as the date — after the year end, and outside the March balance sheet; the securities report filed on 16 June records the cancellation, ¥1,474 million, US$9.00 million, of it, in the period since.

Anyone taking numbers from more than one Janome document will meet the same metric twice. Overseas revenue is 74.7 percent of the total on the geographic table by customer location, where Japan is ¥9,876 million, US$60.3 million, and 75.0 percent on the revenue disaggregation in the segment note, where Japan is ¥9,748 million, US$59.5 million: same country, same year, ¥128 million, US$781,000, apart, because the two tables are cut differently. Set the note's Japan against the note's own total of ¥38,838 million, US$237 million — itself ¥130 million, US$794,000, short of the consolidated line — and the same share comes out at 74.9 percent instead. The results appendix prints 76 percent, which is the previous year and counts only household and industrial equipment, leaving IT out. Counted in machines rather than money the household business is 90 percent overseas, a fourth number wearing the same word. Consolidated revenue itself appears as ¥38,968 million, US$238 million, on one page of that appendix and ¥38,986 million, US$238 million, on another, ¥18 million, US$110,000, apart, and the page headed with this year's revenue prints last year's figures for two of the three segments. Capital expenditure is ¥388 million, US$2.37 million, on the highlights page and in the securities report's capital-expenditure line, ¥398 million, US$2.43 million, as payments for property in the cash flow statement, and ¥465 million, US$2.84 million, in the segment note: three definitions of one idea. None of these gaps is large. They are what happens when the same twelve months are published five times in five weeks — results release, presentation, appendix, a variance notice, and, five weeks after the first of them, the securities report.

What the file does not contain is as fixed as what it does. There are no orders: the order-book disclosure is omitted on the ground that production is principally to forecast, so the company's own sentence about firm orders in the industrial segment has nothing behind it in the filing. There is no customer above 10 percent of revenue, which is itself the answer to the concentration question, delivered as an exemption. There is no geographic profit — the geographic tables stop at revenue and fixed assets — so the margin on North America, 30.6 percent of revenue and the largest region, the one the company says sold its high- and mid-grade machines well, cannot be computed by anyone outside it. There is no revenue for the desktop robots on their own: the finest split printed is robots and presses together, and the appendix states that the product breakdown is not disclosed. In its thirty pages there are no shipment volumes at all and no unit price, only a pie chart of where the machines went. There is no assumed exchange rate in the forecast for the year to March 2027 and no sensitivity to one, in a group booking three quarters of its revenue abroad; the US tariffs appear once, in a clause about having responded quickly, with no rate, no product and no number. The forecast does carry a unit plan: 60,000 more machines in the developed markets and 150,000 more in the emerging ones, on 810,000 sold. The emerging block fell 59,000 in the year just reported, and the share percentages attached to the plan assume the market is the size it was in the year to March 2025, on the company's own estimate.

The Tokyo plant in Hachioji is the designated mother factory and builds household sewing machines and industrial equipment in the same place; the centennial history puts the top-end embroidery models there and the volume models in Thailand, designated the strategic factory. So the sewing machines and the positioning robots come out of one building, inside a group that employed 2,388 people on a consolidated basis as of 31 March 2025. Which is where the argument began. The company says it holds around 25 percent of the domestic home machine market — a figure it discloses without stating its basis, and the report does not say whether it comes from an internal estimate or third-party research; domestic, in any case, not global, whatever the English-language shorthand says. It counts eighteen overseas sales bases in one place in its report and sixteen companies in another. It passed seventy million machines built, cumulatively, in December 2019. All of them, in the end, descendants of a part that holds thread in a ring.

On conversion: yen figures throughout — from the company's financial highlights page, its 100th-term securities report filed on 16 June 2026, the results release and the variance notice of 12 May 2026, the results presentation and appendix dated May 2026, the results briefing of 27 May 2026, the mid-term plan disclosed on 9 May 2025, JANOME REPORT 2025, the centennial history volumes, and the industry statistics — are converted at ¥163.8 = US$1, the rate on 24 July 2026, applied uniformly to every period regardless of the rate prevailing at the time. Dollar figures attached to 1979 or 1984 yen prices are therefore arithmetic, not economics, and no inflation adjustment has been made. Significant figures follow the original yen disclosure. The securities report and the results release state their unit as millions of yen, and figures drawn from the highlights page are read on that basis; where they also appear in those filings, they agree. The presentation and appendix carry no day on their covers, only May 2026; the 12 May date belongs to the results release and the variance notice published that day. Ratios, percentages, multiples and counts are left as published or as calculated from published values, and where a margin, a ratio or a per-unit figure was computed here rather than printed at source, the text says so. Per-unit figures divide a segment revenue line by a unit count — machines rounded to the nearest ten thousand for the household business, exact units for robots and presses; they are indices, not prices.