Skip to content
Cardan-AI
Back to news
Cross-sector30 October 2026

Infineon and NVIDIA push 800 VDC power conversion for AI data centers

Infineon and NVIDIA unveiled, at GTC Berlin on October 21, a new gallium nitride (GaN) power-conversion architecture for 800 VDC AI data center racks. The goal: absorb per-rack power climbing toward the megawatt range without blowing up footprint or heat output.

Infineon and NVIDIA announced on October 21, 2026, at GTC Berlin, a deepening of their collaboration on 800 volt direct current (VDC) power-conversion architectures for AI data centers. The two companies are building on work begun in May and extended in October 2025 on serviceable, hot-swappable 800 VDC server racks, and are now extending the partnership to power-path protection, with hot-swap controller functionality that will let future server boards run directly on 800 VDC.

At the center of the system is a high-voltage intermediate bus converter (IBC), which steps the main 800 V bus down to an intermediate voltage, itself converted downstream into the core and point-of-load rails AI processors need. Infineon pairs its GaN devices with its PSoC Control C3 and EiceDRIVER product families to offer a full system solution. GaN switches faster and with lower losses than silicon, enabling IBC switching frequencies of up to 1 MHz — shrinking magnetics and lowering board height. Supported conversion ratios span 800 V to 50 V, 12 V, and 6 V.

The challenge is fundamentally one of space and heat: with per-rack power heading toward the megawatt range, floor space is scarce and every conversion loss shows up directly as heat to remove. GaN's low conduction and switching losses limit heat buildup as power per rack rises, and the multi-stage conversion approach cuts distribution losses while meeting constraints on board area, thermal budget, and uptime.

Cardan-AI analysis: the constraint now slowing AI deployment is no longer only the external power grid or chip availability — it is moving inside the rack itself, into power conversion. One more physical bottleneck to track for any company sizing AI infrastructure investment on GPU availability alone.

Analysis by

Cardan-AI Intelligence

Our research and analysis unit, dedicated to applied AI for business, industry and regulatory compliance.

Let's talk about your next competitive edge

Thirty minutes to identify the two or three use cases in your operations that pay for themselves within the first year.