Summary

  • Nokia's Q3 2026 disclosure prints net sales in euros on two columns for every segment and publishes a constant-currency definition precise enough that four of five growth percentages can be rebuilt from the printed amounts.
  • Demand travels in a different unit: AI & Cloud order intake is a bare 118 per cent with no euro base anywhere in the material reviewed, the only euro order figure is one prose sentence about EUR 1.5 billion of new orders never joined to that rate, and the sole euro backlog amount sits in an annual note dated 31 December 2025.

What the quarter printed in euros

Group net sales were EUR 5,154 million against EUR 4,700 million a year earlier, printed as a 10 per cent increase. A comparable column pairs the same EUR 5,154 million with EUR 4,705 million and calls that 9 per cent, labelled constant currency and portfolio. The same table carries four segments, each with both euro columns and both growth measures (https://www.sec.gov/Archives/edgar/data/924613/000110465926117972/tm2628115d1_6k.htm):

  • Network Infrastructure — EUR 2,158 million against EUR 1,850 million, +17 per cent reported, +15 per cent at constant currency, "led by IP Networks growing 43% and Optical Networks 18%".
  • Mobile Infrastructure — EUR 2,902 million against EUR 2,755 million, +5 per cent on both measures.
  • Portfolio Businesses — EUR 96 million against EUR 97 million, −1 per cent reported, −2 per cent at constant currency.
  • AI & Cloud — EUR 631 million against EUR 279 million, +126 per cent reported, +123 per cent at constant currency.

Two hundred and seventy-nine million becoming six hundred and thirty-one million is a small base becoming a small business, and it is printed to the million either way.

What the constant-currency sentence hands you

The convention is defined twice. The interim report: "When net sales growth is reported on a constant currency basis / adjusted for currency fluctuations, the exchange rates used to translate amounts in local currencies to euro, our reporting currency, are the average actual periodic exchange rates for the comparative financial period." (https://www.nokia.com/system/files/2026-10/nokia_results_2026_q3.pdf) The factsheet puts it shorter: "Constant currency reporting provides additional information on change in financial measures on a constant currency basis in order to better reflect the underlying business performance. Therefore, change in financial measures at constant currency excludes the impact of changes in exchange rates in comparison to euro, our reporting currency." (https://www.nokia.com/system/files/2026-10/nokia_factsheet_2026_q3_fi.pdf)

That is a definition with an address attached. Hold the comparative period's average rates, restate both periods at them, and the currency leg drops out. It is also why the gap between the two published columns can be named instead of guessed: two points for Network Infrastructure, 17 against 15; three points for AI & Cloud, 126 against 123; none for Mobile Infrastructure.

My own arithmetic, labelled as mine

Rebuilding the printed percentages from the amounts the same documents print for the same segment and period:

  • Group reported: (5,154 − 4,700) ÷ 4,700 = 9.66 per cent, against a printed 10.
  • Group comparable: (5,154 − 4,705) ÷ 4,705 = 9.54 per cent, against a printed 9.
  • Network Infrastructure: (2,158 − 1,850) ÷ 1,850 = 16.65 per cent, against a printed 17.
  • AI & Cloud: (631 − 279) ÷ 279 = 126.16 per cent, against a printed 126.
  • Mobile Infrastructure: (2,902 − 2,755) ÷ 2,755 = 5.34 per cent, against a printed 5.
  • Portfolio Businesses: (96 − 97) ÷ 97 = −1.03 per cent, against a printed −1.

Each lands inside a rounding step of the figure the report prints. That is what a reader-verifiable number looks like: the rule is published, both endpoints are printed, and the arithmetic closes.

One further fact of the table deserves stating, because it changes what the two group columns mean. The reported 2025 base is EUR 4,700 million; the comparable 2025 base is EUR 4,705 million. They differ by EUR 5 million, 0.11 per cent of the reported base. The 9 per cent column and the 10 per cent column therefore do not share a denominator, and the distance between them is not currency translation alone.

Where the demand channel runs on percentages

Order intake appears in one sentence in the material reviewed, sitting beside a net-sales growth rate it does not measure: "Group net sales to AI & Cloud customers grew 123% and order intake 118%." The same release carries the chief executive's line — "In AI & Cloud, net sales more than doubled year-over-year and we secured EUR 1.5 billion in new orders. Our strong order backlog also positions us for future growth." (https://www.nokia.com/newsroom/nokia-corporation-interim-report-for-q3-2026/) Across the release, the results report and the filed Form 6-K, that is the whole of the quarter's demand disclosure.

A growth rate is a complete sentence about direction and an incomplete one about size. Nothing reviewed prints AI & Cloud order intake in euros for the third quarter of 2025 or of 2026, for the group or for any segment.

The one euro order figure, and the line it never touches

The only currency amount attached to orders anywhere in these documents is the EUR 1.5 billion of new AI & Cloud orders in that chief-executive sentence, repeated identically in the release, the results report and the Form 6-K. No reviewed page joins it to the 118 per cent. Nothing states that the 118 per cent is growth in the quarter's new orders, that EUR 1.5 billion is the new-order amount rather than a component of it, or that the two share a perimeter.

Both numbers are exact as printed. What sits between them is unwritten, and the distance is not small: a euro flow and a euro stock are different objects, and a reader cannot turn one into the other without knowing which is which.

The backlog has an amount, and it is last year's

In the quarter the backlog is qualitative. The only euro amount is in the annual revenue note, dated to a different year-end: "At 31 December 2025, the aggregate amount of the transaction price allocated to partially or wholly unsatisfied performance obligations arising from fixed contractual commitments amounted to EUR 19.5 billion (EUR 20.0 billion in 2024)." (https://www.sec.gov/Archives/edgar/data/924613/000162828026015034/R15.htm)

The note is more precise than the quarter about what it counts, and it prints the schedule: "Management has estimated that these unsatisfied performance obligations will be recognized as revenue as follows: 2025 2024 Within 1 year 57% 53% 2-3 years 28% 27% More than 3 years 15% 20% Total 100% 100%".

Read as movement, EUR 19.5 billion against EUR 20.0 billion is minus EUR 0.5 billion, or −2.5 per cent — my subtraction, not a printed one. Within-one-year recognition rises four points, from 53 to 57, and beyond-three-year recognition falls five, from 20 to 15. Both shares are printed; the comparison is mine. What the note does not carry is a currency amount for any band: 57 per cent of EUR 19.5 billion is not a number this note prints, and applying a rounded share to a rounded total would produce a figure the document declines to state.

Where a stranger's arithmetic stops

Every revenue line in this quarter can be walked from a printed euro amount to a printed percentage and back. The demand line cannot be walked at all: a growth rate with no base, one euro order amount whose relation to that rate is unwritten, a qualitative backlog in the quarter, and a year-end backlog total whose schedule is percentages all the way down.

The documents were read as retrieval-extracted page text, so page numbers and column alignment are not observable here; the figures were cross-checked across the release, the results report and the Form 6-K. One naming point: the directory entry behind this piece is Nokia Oyj, while the documents carry Nokia Corporation — the name the same issuer signs on its release, its Form 6-K and its annual note (https://btw.media/en/directory/nokia-oyj-fi).

Four growth rates a reader can rebuild, and one that arrives as a bare percentage, is not a scandal. It is a hierarchy of evidence printed in the open. Knowing which number in a quarterly pack you can check and which you can only receive is most of the analytical work, and Nokia's own disclosure shows exactly where the line falls.