Infineon used its 10th annual OktoberTech event to frame semiconductor demand around physical AIartificial intelligence, automotive electronics and data-center power. The sessions centered on robotics, automotive systems and power delivery, with the company arguing that AIartificial intelligence infrastructure is already changing both product requirements and manufacturing needs.
On the edge-AIartificial intelligence side, Infineon’s edge systems chief Stefan Zizala described a shift away from the earlier IoTinternet of thingsThis term describes connected devices and systems that exchange data over networks. expectation that massive numbers of devices would simply stream data to the cloud. His argument was that newer microcontrollers now have enough memory and compute to run AIartificial intelligence models locally, reducing dependence on cloud processing for functions such as motor control and predictive maintenance. He also cited cost, connectivity, latency and data-protection constraints as reasons to keep more processing at the edge.
The same local-processing logic was applied to software-defined vehicles. Peter Schiefer, president and CEOchief executive officerCEO identifies the company’s top executive who made the statement. of Infineon’s Automotive Division, said only about 10% of vehicles have a software-defined architecture today, with the share expected to rise to around 40%. He argued that legacy OEMoriginal equipment manufacturers will need to consolidate the roughly 100 distributed control boxes in a vehicle into a central computer if they want to manage increasing software complexity.
Infineon also pointed to the growing power requirements of AIartificial intelligence data centers. In October, the company announced a 27 kW three-phase PSU reference design, an 800 VDC1. volts direct-currentElectrical quantity used to describe or test circuit operating behavior.2. volts direct currentVDC denotes a direct-current voltage level used for power distribution and conversion. and ±400 V digital hot-swap controller, a first 900-V GaNgallium nitrideGaN is another material platform mentioned for use in different power delivery roles. bidirectional switch for Enphase’s IQ solid-state transformer, and an 800 V to 50 V, 12 V and 6 V GaNgallium nitrideGaN is another material platform mentioned for use in different power delivery roles. intermediate bus converter collaboration with Nvidia. These announcements were presented as part of a broader move toward higher-voltage distribution and more efficient conversion closer to the load.
That strategy is tied to manufacturing as well. Infineon opened its €5 billion Smart Power Fab in Dresden in July, a 300-mm facility for power and analog/mixed-signal devices and the company’s largest single investment to date. The fab is not yet in full production, but ramp-up is planned for late 2026, with Infineon presenting the site as capacity for the AIartificial intelligence buildout.
For the longer term, Adam White, president of Infineon’s Power and Sensor Systems division, said first prototypes or production systems for 800 V DCdirect currentElectric current that flows in one direction. could appear in 2028. He described the destination as a path from single-phase to three-phase supply and eventually from high-voltage AC1. alternating currentElectric current that periodically reverses direction.2. activated carbonPorous carbon material that adsorbs impurities from gases and liquids. to medium-voltage DCdirect currentElectric current that flows in one direction. at the facility level, with the goal of reaching 1 MWmolecular weightMass of a molecule, commonly expressed as relative molecular or molar mass. racks. He also warned that certification and safety requirements, including UL certification, remain major hurdles.
The company’s 800 V to 50 V and 800 V to 12 V intermediate bus converters, announced in March 2026, use 650-V CoolGaN on the primary side. Infineon also said it is developing an 800 V to 6 V conversion stage, extending the role of GaNgallium nitrideGaN is another material platform mentioned for use in different power delivery roles. in data-center power delivery as current levels rise sharply closer to the processor core.
White added that the balance of power delivery inside AIartificial intelligence systems is changing as well. GPUgraphics processing unitA GPU is a parallel graphics processor widely used for AI acceleration.-heavy training racks still dominate, but inference and agentic workloads are shifting the mix toward CPUcentral processing unitMain processor that executes program instructions and performs data operations.s, and Infineon sees customer interest moving toward CPUcentral processing unitMain processor that executes program instructions and performs data operations.-to-GPUgraphics processing unitA GPU is a parallel graphics processor widely used for AI acceleration. ratios as low as 1:1 in some systems. That shift keeps power delivery architecture in flux, especially as current paths shorten and losses rise when voltage is stepped down toward core levels.