Specifications
TDM2454xx (3rd gen)
280 A quad-phase power module enabling true vertical power delivery (VPD)
Current density
2.0 A/mm2 — Infineon's highest-density power module for VPD
Package size
10 x 9 x 5 mm chip-embedded package tuned for module tiling under the processor
MOSFET technology
OptiMOS 6 trench technology co-packaged with new low-profile inductors for VPD
TDM2354xD / TDM2354xT (2nd gen)
160 A dual-phase power modules, VPD-optimised with benchmark electrical and thermal efficiency
TDM22545D / TDM22544D (1st gen)
140 A dual-phase modules co-packaging two smart power stages, two inductors and decoupling capacitors
Converter topology
Two independent synchronous buck converters per dual-phase module
Thermal rating
Module specifications characterised at coldplate temperatures of 80 C and below
HV IBC reference designs
800 VDC-to-50 V and 800 VDC-to-12 V intermediate bus converters powered by 650 V CoolGaN switches
Efficiency focus
Grid-to-core conversion efficiency to raise compute per watt and lower data-center TCO
Market drivers
Data centers already exceed 2% of global energy consumption, with AI-driven demand forecast to grow 165% between 2023 and 2030
Integration
Module-level tiling shortens current paths from the VRM to the xPU die
Applications
AI GPU and xPU VRM stages, OCP ORV3 servers, hyperscale training and inference racks
Overview
Infineon's OptiMOS power module family brings true vertical power delivery (VPD) to AI servers by placing the voltage regulator modules directly beneath the processor package, so current flows vertically into the die instead of travelling across the board. The portfolio spans three generations — 140 A, 160 A and 280 A — with the flagship TDM2454xx quad-phase module reaching 2.0 A/mm2 of current density.
The 280 A TDM2454xx combines OptiMOS 6 trench technology with a chip-embedded package and a new low-profile inductor design in a 10 x 9 x 5 mm footprint, which allows modules to be tiled under the processor for higher total current and shorter current paths. Two earlier dual-phase generations — the 160 A TDM2354xD/TDM2354xT and the 140 A TDM22545D/TDM22544D — remain in volume for mainstream AI and enterprise server designs.
At the front end, Infineon supplies high-voltage intermediate bus converter (HV IBC) reference designs built on 650 V CoolGaN switches that step 800 VDC down to 50 V or 12 V, supporting the industry's migration toward higher distribution voltages for AI racks.
Key Benefits
Industry-leading 2.0 A/mm2 current density in the flagship 280 A module enables true vertical power delivery under the highest-current AI processors. Co-packaged stages, inductors and decoupling capacitors cut component count and board area, while the chip-embedded package improves both electrical and thermal efficiency at coldplate temperatures up to 80 C. The 800 VDC CoolGaN intermediate bus converter reference designs give server architects a path to higher distribution voltages and lower grid-to-core losses.
Applications
AI GPU and xPU VRM stages in training and inference servers; vertical power delivery under high-TDP accelerator packages; OCP ORV3 and hyperscale rack power shelves; 800 VDC to 50 V / 12 V conversion in next-generation data-center power architectures; and enterprise server voltage regulation where density and thermal headroom are constrained.
Request a Quote — INFINEON OPTIMOS POWER MODULES — VERTICAL POWER DELIVERY VRMS FOR AI SERVERS
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →