Infineon is expanding its collaboration with Nvidia to address one of AI infrastructure’s most pressing engineering challenges: delivering power efficiently as GPU consumption climbs. The German semiconductor company will supply gallium nitride (GaN) power semiconductors and control solutions for Nvidia’s emerging 800V direct current (DC) power architecture. The technology targets high-voltage intermediate bus converters (HV IBCs), which step down electricity from server racks to voltages that server boards can use. Announced Oct. 21, the expanded initiative builds on a partnership unveiled in 2025 that focused on protection circuits for 800V server racks.

Infineon Targets Nvidia’s 800V Power Architecture

The expanded collaboration reflects a shift in data center power design as AI workloads push conventional infrastructure toward its limits. As operators deploy increasingly powerful GPU systems, rack-level power requirements are climbing toward megawatt levels, putting greater pressure on electrical distribution and conversion efficiency. Nvidia is developing an architecture that delivers 800V DC directly to server racks alongside conventional alternating current (AC) and low-voltage DC approaches. Moreover, the strategy aims to reduce power losses and wiring requirements as data center operators prepare for denser computing deployments.

An intermediate bus converter plays a critical role in this architecture by bridging the high voltage entering a server and the lower voltages required by its electronics. After the initial conversion, additional power stages bring electricity to the precise levels needed by individual components, including the graphics processing unit (GPU) core. Infineon supports configurations that step down 800V to 50V, 12V and 6V, giving system designers multiple options for distributing power across server platforms. These conversion stages will become increasingly important as infrastructure teams balance electrical efficiency, thermal performance and the physical constraints of high-density AI systems.

GaN Chips Could Raise AI Server Power Density

Infineon is bringing GaN technology into the power conversion chain to improve performance within the limited space available inside AI servers. Compared with silicon-based power semiconductors, GaN devices offer faster switching speeds and lower conversion losses, helping engineers design compact, efficient power systems. Infineon said its technology can achieve switching frequencies of up to 1MHz, allowing smaller peripheral components, including transformers and inductors. As a result, this creates opportunities to increase power density without expanding the physical footprint of the power delivery system.

The benefits extend beyond component size. Consequently, faster switching and lower conversion losses can help reduce heat generation, an important consideration as AI servers concentrate more computing capacity within increasingly constrained spaces. Lower losses can also support more efficient power delivery, although overall system performance will depend on the design and operation of the complete electrical architecture. Therefore, data center operators, the challenge is no longer simply supplying enough electricity; it is delivering that power reliably and efficiently to increasingly demanding computing hardware.

Integrated Power Solutions Support Next-Generation Data Centers

Infineon plans to supply an integrated intermediate bus converter system combining its GaN power semiconductors, PSoC Control C3 controllers and EiceDRIVER product family. By bringing these components together, the company aims to support the control and conversion requirements of Nvidia’s 800V architecture. Moreover, the integrated approach connects semiconductor switching performance with the control and gate-driving functions needed to operate power conversion systems. It also positions Infineon to participate in the broader technology ecosystem emerging around high-voltage AI data center infrastructure. “GaN-based system solutions are a key technology for enabling next-generation data centers,” said Infineon chief executive officer Jochen Hanebeck. “Infineon is helping deliver the power required for the AI era, from grid to core.”

Why 800V Power Matters for AI Infrastructure

Neverthless, the expanded partnership highlights how power delivery is becoming a central design consideration in the AI infrastructure race. Higher rack densities increase the demands on electrical distribution, conversion equipment and thermal management, making power architecture a strategic decision rather than a back-end engineering detail. Nvidia’s 800V approach and Infineon’s GaN-based solutions target these interconnected challenges through higher-voltage distribution and more compact conversion stages. Furthermore, as AI deployments scale, the ability to deliver power efficiently from incoming electricity to individual processors could become an increasingly important differentiator for data center infrastructure.