Qt Group: The Universe Sent AI, but the Invoice Still Says Perpetual

A 27% headline, an 11% margin with an asterisk, and a CEO now pre-announcing a consumption-pricing revolution he said four quarters ago had no timeline.

QTCOM.HE · Earnings Call · 2026-10-29

A quarter that is really two companies

Qt Group's Q3 2026 headline read like a growth company: net sales of €51.7 million, up 27% (26% at constant currency), and CEO Juha Varelius opened by calling it “a pretty good, okay quarter” — Juha Varelius, CEO · 2026-10-29, noting that Q3 is structurally the group's slowest. But the arithmetic underneath is doing a lot of work. Maintenance revenue rose 155% — that is IAR, acquired last year and absent from the base. Distribution licenses returned to +18%, which the CEO immediately disowned as "timing." And the reported EBITA margin of 11% only exists after €5.3 million of restructuring charges; strip those out and margin was closer to 20%, with year-to-date EBITA at 17.9% versus 17.5% last year. Meanwhile IAR itself — the ~€42 million-revenue compiler business bought for its functional-safety exposure — is shrinking. Asked directly, Varelius gave the number: “Yes, decline, yes. So between 5% and 10%.” — Juha Varelius, CEO · 2026-10-29 That is the direct cost of forcing a subscription change onto a perpetual-license customer base. So the 27% is a blended optical illusion: an acquired asset that grew ~5% standalone, now contracting as its revenue model is re-cut, wrapped around a legacy core doing roughly 10%.

The pricing pivot is no longer hypothetical

This is the genuinely new thing in the report. For several quarters the consumption based pricing idea was floated and then walked back. In February, Varelius was explicit: “I did not say that we're going to do that change, and I did not say that we're going to do that change this year, but I said that, that's what I see... no time line, definitely not this year, next year or so.” — Juha Varelius, CEO · 2026-02-26 By Q3 it has a shape, a rationale, and an actual customer conversation. He now says the company will start introducing it, that some consumption-linked pricing already exists, and that customers using AI have been asking how they will be charged for framework usage. The logic is the interesting part, because it is a concession dressed as an opportunity:

I said that over time, maybe Qt can be like a EUR 300 million business and where do we grow from there. And then the universe send us AI. And now AI actually means that the -- as we go further develop this, the more people will be able to do coding and using Qt framework through AI.
Juha Varelius, CEO · 2026-10-29

That is the whole thesis in one breath. Qt's growth ceiling is that there are only so many C++ developers on earth — five to six million of a 25-30 million developer population, by his own count — and AI code generation is supposed to expand that population by lowering the skill floor. It also converts a per-seat developer licence into something usage-linked. Varelius is careful to promise neutrality: “I'm not expecting to have a huge revenue increase through that, but I don't want to see any decrease either,” — Juha Varelius, CEO · 2026-10-29 and the rollout will be staged — a limited market, new customers only — so a mistake stays containable. This is not a Qt-only idea. The global theme tape carried outcome based pricing among the quarter's market-level keywords, and the sector broadly is repricing software around AI agents and token consumption. Qt is late, but it is finally in the room.

The reassurance that has become a ritual

The most telling feature of the call is what is said identically every quarter. Varelius again spent minutes arguing that AI is not cannibalising developer seats: no change in churn, renewals holding, customers merely using AI as a tool. He reached for product recall risk, hardware memory efficiency, certifications and addressable market mechanics — the same architecture he used in February, when pressed that annual developer-team optimisation was a structural threat and he replied flatly that he did not see it that way. Notably, the keyword functional safety, a top-ranked pillar two quarters ago, has faded from the presentation even as it underpins every sentence about why embedded software is AI-resistant. Meanwhile the U.S. remains the wound that will not close. “...we have lots of room to improve in the U.S.,” — Juha Varelius, CEO · 2026-02-26 was the February line; May brought management changes and "people changes"; October repeats that Europe and APAC are fine and that "I'm not that happy how we're doing in the U.S." Three consecutive calls, the same diagnosis, still no inflection. That is an execution problem wearing a macro costume. One prop has quietly vanished. multiyear contracts used to be cited as a growth tailwind; this quarter the CFO told analysts the opposite, that maturity mix was a headwind in Q3 — “the multiyear contracts are in Q3 and as a share of the total in Q3, they are lower” — Ann Littorin, CFO · 2026-10-29 — and that no shift to longer contracts was padding the P&L. A year ago the same mix shift cost €3-5 million in a single quarter. Removing that prop explicitly is a small act of candour worth crediting.

What to watch into 2027

Guidance is unchanged — 10%+ net sales growth and a 15%+ operating margin, framed as a floor rather than a range after two profit warnings in the prior year. The €20 million cost programme is real: headcount sat near 1,000 at end-September, roughly flat year-on-year despite absorbing 200-plus IAR staff. Management is explicit that the savings will be partly reinvested, with hiring already underway in APAC and in AI capability. The investor's problem is the sequencing. The profit recovery is being funded by cutting capacity in the one region that needs fixing, while the growth story leans on a pricing model that does not yet exist commercially, a design tooling product the CEO says he does not expect to generate "huge revenue numbers," and an acquired asset in deliberate decline. The AI-led expansion — the part that justifies the multiple — is, in his own framing, a three-to-five-year story. That is a long time to hold a thesis that today rests on a 27% growth number with a timing caveat and an 11% margin with an asterisk.