Open in interactive viewer → charts, metric popovers & call review

Infineon Lifts Its AI Power Number to €4 Billion — and Quietly Retires the Backlog

Guidance jumps again while the AI-infrastructure tape de-rates, two price hikes land, and management stops talking about the two things analysts ask about most
IFX.DE · Earnings Call · 2026-11-10

The number that moved

Infineon's 2026 annual press conference, staged at its electronica booth in Munich rather than in a conference-room studio, was built around one arithmetic reveal. Fiscal 2026 closed with revenue of €16.353 billion, up 12% reported and 15% at constant currency, a segment result margin of 19.7% against 17.5%, and adjusted free cash flow of €1.9 billion — 11.8% of revenue, which the CEO pointedly noted is "in line with our target operating model." Then the forward number: roughly €20 billion of revenue for fiscal 2027, more than 20% growth, with the segment result margin climbing into the mid-20s.

The engine inside that is power supply for data centers. Jochen Hanebeck disclosed that AI data-center power revenue reached €1.7 billion in fiscal 2026 — "more than doubled" again — with roughly €500 million more from classical data centers, for €2.2 billion combined. For 2027 the company expects that combined line to reach "close to €4 billion."

That is the story. In August, on the fiscal Q3 call, the guide for the following year was deliberately vague but bracketed: the comparable figure would go from €2.5 billion to something like €3 billion. Four months later it is €4 billion. The single most important line in this transcript is “The combined revenue volume of these agreements, both those already signed and those currently under negotiation has now exceeded EUR 10 billion.” — Jochen Hanebeck, Board Member/Executive · 2026-11-10 That is a bookings statement dressed as a capacity remark, and it is roughly five years of the current AI power run-rate.

The conservatism is worth noting because it cuts against the bubble narrative. Fourth-quarter AI revenue alone was €800 million. Annualise that and you are at €3.2 billion before any growth — so a €4 billion target for next year implies something under 30% growth off the exit rate. CFO Sven Schneider said as much: “from the run rate, we didn't assume extreme growth in AI, but we did assume solid growth.” — Sven Schneider, Board Member/Executive · 2026-11-10 This is not a company extrapolating a mania. It is a company extrapolating a supply agreement.

From a price war to a price list

The second, less-heralded change is on price. For roughly two years Infineon's language on pricing was defensive — Chinese IGBT competition, aggressive silicon-carbide entrants, "annual price declines," and a Chinese price war that management repeatedly described as a market it would simply stop serving at unattractive price points. In May 2026 Hanebeck was blunt: “This market, especially in China, is under substantial price pressure.” — Jochen Hanebeck, Chairman of the Board · 2026-05-06

On Tuesday, COO Alexander Gorski described the opposite environment: scarcity spilling out of AI and grid infrastructure into adjacent product families, "order backlogs increasing incessantly," and two enforced price increases in calendar 2026, in April and July, with 2027 negotiations underway.

In 2026, we have already seen 2 price increases, which we have enforced on the market, as you all know, one being in April and the other one being in July. These increases were for power semiconductors, first and foremost, and we're now in volume and price negotiations with a lot of our major OEMs. In 2027, we will see a positive price environment.

Alexander Gorski, Board Member/Executive · 2026-11-10

That is a genuine regime change for a company whose margin recovery has, for six quarters, been attributed almost entirely to cost — the Step Up programme, idle-cost absorption and manufacturing footprint work. Cost-led margin expansion is respectable; price-led margin expansion is a different multiple. Note the vehicle: power semiconductors and the wider power infrastructure build-out, not the microcontroller franchise that historically carried the pricing power.

Two things management stopped saying

The first is the order backlog. Joachim Hofer opened the Q&A by asking how the €30 billion figure — a number Infineon itself had promoted for quarters — had developed. Schneider's answer was a disclosure retreat dressed as sector convention: “We have decided in line with the other companies of our sector to no longer report these figures because when viewed isolatedly, this KPI can only show you a limited picture.” — Sven Schneider, Board Member/Executive · 2026-11-10 He then added, almost as an aside, that if he were to report it, it would be higher than the prior quarter. Retiring a KPI in the same breath as saying it improved is a choice, and it removes the cleanest third-party check on the €10 billion capacity-agreement claim.

The second is the share price. In August, Schneider engaged the topic at length, volunteering that €60–65 was “not a bad share price at all” — Sven Schneider, Chief Financial Officer · 2026-08-05 and framing the AI build-out as "the biggest growth driver of our corporate history." On Tuesday, with the stock having fallen roughly 30% from a June peak above €80 and back above €50, the response was a shrug: “I have become accustomed to not comment on what the market says because the market is always right.” — Unknown Executive, Board Member/Executive · 2026-11-10 Consistent with that, the share price keyword, which was the single highest-momentum term in the fiscal Q3 keyword set, has vanished entirely from the current top-30. When a management team stops litigating its own valuation, it usually means the argument did not land.

The tape is voting the other way

Here is the tension worth holding. Infineon raised its AI power target by roughly a third in one quarter and added €10 billion of contracted or near-contracted volume — and the market's AI-infrastructure complex spent the last 30 days de-rating. Across the global price movers, AI data centers fell more than 20% with nine advancers against fifty-seven decliners; data center capacity, high-bandwidth memory and co-packaged optics all sit in the 30-day decliner list, alongside PCIe Gen 6 interconnect names. The 30-day advancer list is airlines, home improvement, insurance and consumer staples — a rotation out of the compute build-out, not into it.

So the company is guiding up into a tape that is repricing the theme down. Hanebeck's answer to the political backlash question — that data centers will diversify geographically, to India, the Arabian Peninsula, Australia, Europe and China — is the right long-term answer and also the one a CEO gives when the marginal buyer has stopped believing the US hyperscaler capex line. His own framing of the volatility was unusually candid: the capital market is "trying to identify the profit pools of the future, which are not necessarily going to be the same ones as today."

What to watch

Three things. First, the 2027 price negotiations: Gorski promised to announce results, and the mid-20s margin guide embeds something for them, but a positive price year in power semis would be the first genuine pricing cycle since 2022 and would validate the mix shift toward AI and grid. Second, the €4 billion number: it is only ~25% above the Q4 annualised run-rate, so the risk is not that it is a stretch — it is that it is sandbagged, and that the 2028 number, when it comes, is the one that reprices the stock. Third, the second-order bets: the AI data franchise is now roughly a tenth of group revenue, while Physical AI and robotics — showcased through the Agile Robots partnership and the new RISC-V AURIX RC1 platform — remain narrative optionality with no revenue line attached. The 300mm gallium-nitride ramp in Villach and the 300mm silicon-carbide milestone matter because they are the cost curve under the €4 billion, not because they are products.

Infineon is no longer a cyclical recovery story with an AI kicker. It is an AI power-supply company growing into a €10 billion order book that just told you it will stop telling you the order book. The gap between those two facts is where the next two quarters get decided.