Infineon Lifts Data Center Targets While Markets Price an AI Retreat

The market has decided artificial intelligence infrastructure is running out of road. A post-midterm tech selloff erased $2.5 trillion in a single session. Semiconductor indices fell twelve percent. Lawmakers floated construction moratoriums. Strategists at UBS warned that thirty percent of planned 2028 US data-center additions were suddenly at risk.

Then Infineon Technologies arrived with an entirely different message. The German chipmaker builds power semiconductors. These are the silicon switches that convert and move electricity. The utility grid delivers high-voltage alternating current. An advanced AI accelerator cannot touch that power directly.

Modern processors demand roughly one volt of direct current delivered at hundreds of amps. Between the utility substation and the chip die sits a dense stack of transformers and converters, and every stage burns off a slice of the incoming power as heat. Heat is waste. How efficient that chain is decides whether a data center runs or melts.

Think of the GPU as your phone and Infineon's silicon as the charger brick. Nobody brags about the brick in their backpack until it overheats and melts the outlet. Inside a server farm, every accelerator rack is a space heater that also does linear algebra, and Infineon makes the switches that stop it becoming only a space heater.

That positioning makes power silicon the newest gatekeeper of the AI buildout, because the bottleneck has quietly shifted. It is no longer about processor allocations. It is about usable grid watts. Without extreme conversion efficiency, cooling systems choke and rack densities collapse. Watts come first. You cannot scale frontier models without mastering the power stage.

Infineon just backed that shift with aggressive numbers. The company closed fiscal 2026 with €16.35 billion in revenue, up twelve percent. Data-center power revenue reached €2.2 billion, including €1.7 billion from AI servers. That AI slice more than doubled year over year. Data-center power is now roughly thirteen percent of group revenue.

For fiscal 2027, management raised its data-center power revenue target to close to €4 billion, including about €3.5 billion from AI racks, a forty percent increase over February's target. Group revenue should reach roughly €20 billion, delivering top-line growth above twenty percent, and chief executive Jochen Hanebeck guided to a five-point jump in segment profitability.

Yet public markets immediately faded the raise. Shares jumped as much as seven percent on the news. Then they surrendered the gains and finished less than one percent higher. The following session brought a broad chip rally, and Infineon traded up roughly six percent alongside peers. The stock still sits roughly forty percent below its June peak, trading near thirteen times forward earnings. By contrast, peers like Vicor and Monolithic Power command multiples near twenty-eight and thirty times.

Investors are still pricing Infineon as a sleepy automotive supplier. For the equity to re-rate, two things must happen together. Data-center power sales must actually land near that €4 billion target. At the same time, segment margins must climb those guided five percentage points into the mid-twenties.

That operating leverage is what awards premium multiples to structural winners, and it is the only thing standing between Infineon and a rerating. If profitability expands as promised, thirteen times earnings looks like an obvious pricing mistake. If the margins stall, the multiple is right. Then it is just an auto chipmaker.

The bear case argues this is a cyclical industrial buying an AI narrative with expensive fab steel. Infineon plans to pour between €3.0 billion and €3.5 billion of capital expenditure into Dresden, Kulim, and Villach. That spending left reported free cash flow at just €993 million for the year. Guidance promises only about €1.8 billion next year. That barely covers the cost of a single fab phase.

Bears also emphasize that the other eighty-seven percent of the business is still cars, factories, and consumer hardware. In fact, fourth-quarter net profit missed analyst expectations even as the operating segment beat. The €4 billion target also folds in roughly €500 million of flat traditional data-center sales. That is not pure AI growth. The bears win if fiscal 2027 data-center sales track below €3.5 billion. They also win if free cash flow falls under €1.8 billion.

Watch the fiscal first-quarter print next year. Management guided to approximately €4.6 billion in revenue and a segment margin near twenty-three percent, which will show whether enterprise power demand holds up against Washington's political noise. It also shows who was right.

Disclosure: The Signal holds no position in IFX or IFNNY. Positions may change. This is not financial advice.