Japan's machine tool orders rose 64.7% in August — and August was not the biggest month of the year
The Japan Machine Tool Builders' Association posted its August 2026 preliminary at 15:00 JST on 9 September 2026: total orders of ¥197,880 million, an index of 164.7 against August last year. That is the largest year-on-year figure printed in 2026. August was also not the largest month of 2026 — June was, at ¥203,515 million. Both statements come off the same association's own tables, and the distance between them is the whole story.
I pulled the figures out of JMTBA's Japanese preliminary PDF, 2026年8月分 受注速報, because there is no English version and there will not be one for weeks. The English notes for August are not due until the association's own next-publication schedule allows, and the English machine-tool-orders page still ends at the July final. If a figure below is not on that one-page table, I say where it came from.
What the one-page table actually prints
The preliminary is four lines and five columns. It has no commentary, no regional split, no industry mix and no outlook. It prints indices, not percentages: 164.7 means 64.7% above the same month last year, and 102.5 means 2.5% above July. Every trade report converting those into "surged" and "accelerated" is doing arithmetic the document declines to do for them.
Here is 2026 to date, month by month, taken from the eight monthly preliminaries in sequence.
| 2026 | Total orders (¥m) | vs prior month | vs same month 2025 | Implied 2025 month (¥m) | Foreign share |
|---|---|---|---|---|---|
| January | 145,587 | −8.2% | +25.3% | 116,190 | 77.6% |
| February | 146,784 | +0.8% | +24.2% | 118,183 | 74.7% |
| March | 193,506 | +31.9% | +28.1% | 151,059 | 73.9% |
| April | 188,971 | −2.3% | +45.1% | 130,235 | 73.9% |
| May | 176,833 | −6.4% | +37.4% | 128,700 | 74.5% |
| June | 203,515 | +15.0% | +52.8% | 133,190 | 71.5% |
| July | 193,102 | −5.1% | +50.4% | 128,392 | 72.4% |
| August | 197,880 | +2.5% | +64.7% | 120,146 | 74.0% |
Totals, monthly indices and the domestic/foreign split are JMTBA's. The "implied 2025 month" column is mine: the association publishes the year-on-year index but not the year-earlier level, so each figure is that month's total divided by its own index. It is arithmetic on published numbers, not a JMTBA figure, and I would not quote it as one.
The biggest percentage of the year sits on the smallest base of the year
Read the last two columns together. August 2026 is the second-largest order month of the year. August 2025, implied at ¥120,146 million, is the weakest month in the whole year-earlier row — below January's base, below February's, ¥31 billion below March's. A record percentage against a soft month is not the same event as a record month, and the August preliminary gives you no way to tell the two apart unless you build the comparison yourself.
This matters because the figure is about to be repeated for a month. The trade press quotes the preliminary, the machine builders quote the trade press, and by early October "+64.7%" will be sitting in a sales deck with no denominator attached to it. The honest version is shorter: Japanese machine tool orders are running at a very high level, they rose slightly in August, and the year-on-year comparison is flattered by a weak autumn last year.
速報 is not 確報, and the difference is a date
The document is labelled 受注速報 — preliminary orders. The revised figure is the 確報, and JMTBA prints the next publication date on the sheet itself: the next preliminary is due 9 October 2026 at 15:00. The two are not interchangeable, and the gap between them is where the detail lives. The July final carries the regional and industry breakdown; the August preliminary carries none.
The revisions are small, and they are real. Add the July preliminary's cumulative total of ¥1,248,243 million to August's monthly ¥197,880 million and you get ¥1,446,123 million. The August sheet prints ¥1,446,111 million. Twelve million yen on 1.4 trillion is rounding, and it is also the reason the word "preliminary" is on the page. Anyone writing a contract clause that triggers off a published index should be citing the final, naming the publication and naming its date — not a number lifted from a news story about a preliminary.
The two halves are walking away from each other
Foreign orders were ¥146,347 million in August, up 4.6% on July. Domestic orders were ¥51,533 million, down 3.2%, and that is the second consecutive monthly fall — July domestic was down 8.2%. Foreign demand is 74.0% of the August book and 73.9% of the eight-month book of ¥1,068,445 million out of ¥1,446,111 million.
For a buyer outside Japan that split is not trivia, it is your position in a queue. The eight-month run rate is ¥180.8 billion a month, up 40.9% on the same period last year. You are not competing for build slots against Japanese domestic customers, whose orders are shrinking. You are competing against the export book, which is the part that is growing.
The count nobody has verified
Two claims circulate with this series that the preliminary does not support. On the July notes, August makes the fourteenth consecutive month of year-on-year growth and the eighteenth consecutive month above ¥120 billion. Both counts are ours, built by extending the July final's own run; the August sheet asserts neither. Say "by our count" or check it against the notes PDF. Publishing a streak as though the association declared it is the kind of small mis-citation that gets repeated until someone puts it in a tender document.
What survives into a purchase order
Three things from this table are safe to rely on when you are negotiating for capacity, and one is not.
Safe: the level. ¥1.446 trillion ordered in eight months, averaging ¥180.8 billion a month, is a full order book by any reading, and full order books set delivery dates. Safe: the direction of the split, two months running. Safe: the publication calendar, because a price-escalation clause indexed to a statistic needs the statistic's own release date written into it, and JMTBA prints that date on every sheet.
Not safe: the 64.7%. It is arithmetically correct and it describes last year as much as this one. Quote it with the base attached or leave it out. If a builder's quotation cites accelerating demand as the reason your lead time moved from 26 weeks to 34, ask which month they mean and what the year-earlier month was. The answer is a one-page PDF away, and it is in the table above.
The wider point is duller than the headline and more useful. Japan's machine tool order book is very full, it is filling from abroad, and the month-to-month motion since March has been sideways at a high level rather than a climb. Plan lead times off ¥180 billion a month, not off a percentage.
If you need machining capacity quoted against a real delivery date rather than a market narrative, send us the drawing and the quantity and we will come back with a lead time we will hold.
Sources
- JMTBA, 2026年8月分 受注速報 (August 2026 preliminary orders, PDF, Japanese), published 9 September 2026 — every August figure above.
- JMTBA, July 2026 preliminary and the January–June preliminaries on the same path — the monthly series in the table.
- JMTBA, JMTBA Notes, July 2026 final (English, PDF) — regional and industry breakdown, July only.
- JMTBA news index and statistics index — the 2026.09.09 announcement entry.
- The mother of all indicators – What Japanese machines tell us, Meine Geldseite — the March 2026 record framing.
- Is the Automation Recovery Real? #AirTAC at 25x, #HIWIN at 36x—What Is the Market Actually Buying?, 阿將 Jonstyle — machine tool orders alongside automation component valuations.