Infineon’s AI power upside is real, but the print says margins are still paying for idle fabs

The print changes the debate from “is revenue bottoming?” to “who owns the economics of the recovery?” because the Street was positioned for €3,702.4 million of revenue and €0.49 of EPS, while the actual street-comparison basis came in at €4,619.4 million of revenue and €0.43 of EPS. That separation matters. A +24.8% revenue surprise should normally settle the cyclical recovery argument, but the -12.2% EPS surprise says the incremental revenue is arriving through a cost structure still burdened by underloading, mix friction, and investment drag. The market may be mispricing this event if it capitalizes the AI power target as though it drops through like a software attach rate; this is still an IDM with idle costs, front-end construction, and automotive cyclicality. The actionable conclusion is not bearish on demand, it is stricter than that: Infineon has found a credible growth pocket in AI data-center power, but the equity deserves credit only as the margin bridge becomes visible in reported gross margin and EPS, not merely in revenue guidance.

That tension is visible when the street-comparison print is kept separate from the company’s own reported basis. On the call, Jochen Hanebeck anchored the company accounts by saying, “In the fourth quarter, group revenue increased to EUR 3.943 billion.” In the quarterly history, Q4 FY2025 revenue was €3,944.0 million, up +6.5% QoQ and +0.6% YoY, with diluted EPS of €0.16 and gross margin of 38.1%. Those are not the same top-line basis as the €4,619.4 million used for the Street surprise, and mixing them would flatter the operating story. On the company basis, revenue is no longer falling, but gross margin at 38.1% is below Q3 FY2025’s 40.9% despite the +6.5% QoQ revenue rise. That is the core of the print: volume recovery without margin recovery. A PM who only sees the revenue beat misses that Q4 FY2025 EPS of €0.16 was below Q3 FY2025’s €0.23 and below Q4 FY2024’s €0.29, even as Q4 FY2025 revenue of €3,944.0 million was above Q4 FY2024 revenue of €3,919.0 million.

The history makes the operating leverage problem harder to dismiss as a one-quarter anomaly. Infineon’s gross margin was 46.6% in Q2 FY2023, 44.5% in Q3 FY2023, and 43.6% in Q4 FY2023, before sliding to 38.7% in Q2 FY2024 and again 38.7% in Q2 FY2025. The latest reported trajectory does not show a clean snapback: 40.9% in Q3 FY2025, 38.1% in Q4 FY2025, 39.9% in Q1 FY2026, and 38.7% in Q2 FY2026. Revenue did improve from €3,424.0 million in Q1 FY2025 to €3,590.0 million in Q2 FY2025, €3,704.0 million in Q3 FY2025, and €3,944.0 million in Q4 FY2025, but EPS moved from €0.19 to €0.18 to €0.23 to €0.16. That is not a normal upcycle earnings pattern. The market had reason to price in revenue stabilization after Q2 FY2025 revenue QoQ of +4.8% and Q3 FY2025 revenue QoQ of +3.2%; what surprised was that the Q4 FY2025 +6.5% QoQ revenue move came with gross margin down to 38.1% and EPS down to €0.16.

The company’s own explanation points directly to the culprit investors need to underwrite rather than wave away. Management quantified the drag as “just under EUR 1 billion in terms of idle costs corresponding to roughly 600 basis points, so 6% point margin headwind,” according to an Unknown Executive. That sentence earns attention because it frames the margin ceiling numerically, not rhetorically. If idle costs are just under €1 billion and roughly 600 basis points, then the path back above the 20% segment result margin that an attendee explicitly asked about is not just a demand question; it is a utilization and timing question. Hanebeck said Q4 segment result was €717 million and segment result margin was 18.2% compared with 18.0% in the previous quarter, so the sequential improvement in segment result margin was only from 18.0% to 18.2% even with the company-reported Q4 revenue at €3.943 billion. For FY2025, revenue was €14.662 billion and segment result margin reached 17.5% after 20.8% in the 2024 fiscal year, which quantifies the damage from the downcycle and the investment base. The surprise is not that Infineon has cyclical slack; the surprise is how much of the revenue recovery is still being consumed before EPS.

The AI power story is the legitimate counterweight, and it is the reason the stock should not be reduced to an auto-cycle short. Hanebeck said Infineon “almost triple[d]” revenues from power supply solutions for AI data centers in fiscal 2025, reaching over €700 million, and raised the FY2026 forecast from €1 billion to around €1.5 billion. That is the clearest positive revision in the event, because management is not merely describing AI exposure as an option; it is assigning a near-term revenue level of around €1.5 billion and saying the addressable market for Infineon can reach €8 billion to €12 billion by the end of the decade. The variant perception is to value that raise as a margin bridge only if it shows up in gross margin, not as an automatic multiple bridge. Power supply for AI data centers is a real revenue vector, but the same transcript includes idle costs of just under €1 billion, reported free cash flow of minus €1.051 billion for FY2025, and adjusted free cash flow of plus €1.803 billion. Those numbers can coexist, but they do not support a clean earnings acceleration narrative yet.

That cash-flow split is the second place where headline optimism needs discipline. FY2025 free cash flow was minus €1.051 billion, while adjusted free cash flow, excluding investments in large front-end buildings and major acquisitions such as Marvell’s Automotive Ethernet business, was plus €1.803 billion. Hanebeck said that adjusted figure corresponds to approximately 12% and 12.3% of revenue, and the company still plans to propose a stable dividend of €0.35 per share. For FY2026, reported free cash flow is expected to be around €1.1 billion and adjusted free cash flow around €1.6 billion. The improvement from minus €1.051 billion reported free cash flow to around €1.1 billion expected reported free cash flow is the best evidence that investment intensity should become less punitive, but the adjusted free cash flow guide of around €1.6 billion is below FY2025 adjusted free cash flow of plus €1.803 billion. That conflict is important: reported cash flow improves because excluded investment burden eases, while adjusted cash generation is guided lower. It argues for balance-sheet relief, not necessarily a better underlying conversion rate.

The guidance reinforces that the next quarter is a seasonal and China-auto test, not a victory lap. Hanebeck guided current first-quarter revenue of around €3.6 billion and said that would be a revenue decline of around 9% compared with the previous quarter, above typical seasonality. He also quantified a revenue reduction of around €400 million from the rule-of-thumb effect he described. In the quarterly history, Q1 FY2026 revenue is €3,627.7 million, with revenue QoQ of -8.0%, revenue YoY of +5.9%, gross margin of 39.9%, and diluted EPS of €0.19. That set of numbers is better than Q4 FY2025 on margin and EPS, but it is not yet a breakout. The China comment is the largest demand caveat because Hanebeck said electric vehicle growth in China is likely to slow after the share of electric vehicles in new car sales exceeded 50% and government subsidies are being reduced. Andreas Urschitz quantified exposure by saying the past fiscal year had a 38% share of revenue in China, with 29 percentage points achieved in Mainland China. A company with 38% China revenue share and 29 percentage points in Mainland China cannot offset an auto slowdown with AI power unless the AI ramp is accompanied by utilization improvement.

The Marvell Automotive Ethernet acquisition also deserves a specific read-through because the data pack gives no named Infineon customers and no named Infineon suppliers, which limits customer-channel conclusions. The only named transaction counterparty in the supply-chain-relevant material is Marvell’s Automotive Ethernet business, excluded from FY2025 adjusted free cash flow and tied by Sven Schneider to €200 million of revenue in 2026 with positive profitability. That has two implications. First, Marvell’s divested automotive Ethernet asset is not a drag in management’s framing because Schneider called it positive in profitability and assigned €200 million of 2026 revenue. Second, the absence of named customers and suppliers in the data pack means this print cannot be used to make a direct order-book claim for a specific customer or wafer supplier; the defensible read-through is instead competitive and end-market based. For automotive semiconductor competitors, the China exposure figure of 38% of revenue and 29 percentage points in Mainland China is a warning that China EV mix and subsidy normalization are still setting the slope of power and microcontroller demand.

That competitive read-through is uneven across IDMs because Infineon’s margin structure is not keeping pace with several peers, even as growth has reappeared. In the latest peer table, IFNNY shows revenue YoY of +7.9% and gross margin of 38.7%. That places its growth above INTC at +7.2% and below NXPI at +12.2%, TXN at +18.6%, STM at +22.8%, and RNECY at +25.4%. On gross margin, IFNNY’s 38.7% is below INTC at 39.4%, RNECY at 51.2%, NXPI at 56.2%, and TXN at 58.0%, while above STM at 33.8% and 6724.T at 35.0%. The point is not that all these companies have identical mix; they do not. The investable comparison is that Infineon’s +7.9% revenue YoY is not weak enough to explain why its gross margin is only 38.7%. The company-specific issue is underutilization and capacity cost, not just end demand. That distinction matters for pair trades: a broad auto and industrial recovery can lift revenue for the group, but Infineon needs its own idle-cost unwind before its earnings beta matches its revenue beta.

The call tone shows management leaned forward on guidance and AI, while the Q&A profile became less comforting. The tone history shows sentiment rising from 0.21 in Q1 FY2026 to 0.29 in Q2 FY2026, guidance_tone rising from 0.18 to 0.55, and tone_confidence rising from 0.58 to 0.59. Prepared_sentiment increased from 0.34 to 0.42 and ai_optimism from 0.53 to 0.55. But the delivery quality split is not clean: qa_sentiment fell from 0.14 to 0.01, uncertainty increased from 34.8 to 35.4, and qa_evasiveness rose from 6.0 to 39.7, with the call-over-call delta showing qa_evasiveness +33.8. That is exactly what a PM should expect when management has a better AI revenue message but still has to answer for the path to margins above 20%. The prepared remarks can support around €1.5 billion of FY2026 AI data-center power revenue; the Q&A has to reconcile that with gross margin at 38.7% in Q2 FY2026 and the cited idle-cost headwind of roughly 600 basis points.

That tonal split also helps separate what was priced in from what actually surprised. Priced in was a cyclical trough narrative: revenue had already moved from €3,424.0 million in Q1 FY2025 to €3,704.0 million in Q3 FY2025, and Q3 FY2025 gross margin had improved to 40.9%. Investors could reasonably expect the Q4 FY2025 company-reported revenue increase to €3,944.0 million and the Street revenue beat on the separate basis. What was not priced in, or at least should not have been, was the EPS disappointment of €0.43 versus €0.49 on the Street basis and the company-basis gross-margin fall to 38.1% in Q4 FY2025. Also surprising was the size and specificity of the AI power raise, from the prior FY2026 forecast of €1 billion to around €1.5 billion, after over €700 million in FY2025. The result is a stock debate with two simultaneous revisions: raise revenue credibility in AI power, but lower confidence that near-term EPS will follow revenue without a utilization inflection.

The next quarter should settle whether this is a transition quarter or a structural margin slog. The numbers to watch are Q1 FY2026 revenue around €3.6 billion or the history figure of €3,627.7 million, gross margin near 39.9%, EPS near €0.19, and whether the -8.0% QoQ revenue decline stays close to the guided around 9% decline rather than worsening. The thesis is confirmed if management keeps FY2026 AI data-center power revenue around €1.5 billion, reports progress against the just under €1 billion idle-cost burden, and shows gross margin moving sustainably above the 38.1% to 39.9% range seen from Q4 FY2025 to Q1 FY2026. It breaks if Q2 FY2026’s €3,874.7 million revenue, 38.7% gross margin, and €0.23 EPS become the ceiling rather than the bridge, or if China exposure, 38% of revenue with 29 percentage points in Mainland China, forces another above-seasonal reset. The date that matters is the next quarterly call after this 2025-11-12 event: Infineon must prove that around €1.5 billion of AI power revenue can absorb fabs and lift EPS, not just decorate a revenue line that the Street already underestimated.