Server rack power management

Powering next‑gen AI racks with efficient, scalable solutions across PSUs, IBCs, and VRMs

About

While traditional racks operated at tens of kilowatts, today’s AI racks exceed 100 kW, with future designs moving toward 600 kW and megawatt-scale power levels. This rapid increase in GPU power demand is pushing conventional server rack architectures to their physical, electrical, and thermal limits.

For design engineers, delivering more power is only part of the challenge. Increasing power and thermal density must be managed across PSUs, IBCs, and VRMs, often within smaller system footprints. Higher distribution voltages and lower currents are also becoming essential to reduce power losses and improve server power efficiency across the rack.

How does Infineon enable efficient power delivery across AI server racks?

Infineon’s broad server rack power management portfolio helps engineers address these challengesacross power conversion and distribution. Our portfolio combines Si, SiC, and GaN power switches with: 

  • Gate drivers
  • Integrated power stages
  • Auxiliary power solutions
  • Current sensors
  • Digital controllers
  • Microcontrollers

This system-level approach enables scalable power delivery, optimized efficiency, and seamless integration for next-generation server racks.

As AI server racks push towards megawatt-class power, traditional PSU architectures must undergo a redesign. Next-generation racks adopt high-voltage DC (HVDC) conversion, shifting from ~400 V AC to 800 V DC. To support these transitions, PSU switch ratings must scale from 650 V to 1.2 kV, enabling advanced three-phase topologies such as B6 or Vienna PFC for higher efficiency. 

Power density is only part of the challenge. PSU form factors remain tightly constrained, making effective cooling and space utilization critical to server rack power management. Sidecar placement can free up vertical rack space and provide greater flexibility for cooling high-power systems.

How do wide-bandgap semiconductors improve PSU efficiency?

Infineon’s CoolGaN™ and CoolSiC™ devices reduce switching losses in the AC-DC stages of AI server PSUs. Combined with gate drivers, digital isolators, auxiliary power solutions, and microcontrollers, they support single-phase and three-phase designs across multiple topologies.

Furthermore, modular and SMD packages, together with reference designs, help engineers improve server power efficiency, accelerate development, and reduce time to market.

BBUs provide immediate power when the main PSU or grid supply is interrupted, maintaining reliable server rack power management during outages. For example, to meet 48 V system requirements and a demand for 90–240 seconds of backup time, BBUs must deliver high efficiency and a scalable architecture. 

Infineon’s patented BBU technologies use 40 V and 80 V MOSFET-based designs to improve server power efficiency and power density while reducing BOM costs. 

The scalable architecture supports multiple power levels within constrained voltage ranges, helping engineers meet the backup power requirements of next-generation AI server racks.

IBCs provide efficient voltage conversion between the PSU and point-of-load (POL) converters. In high-density AI server racks, they must combine high power density, strong thermal performance, and scalable conversion efficiency.

Infineon supports reliable IBC designs with CoolGaN™ switches, gate drivers, digital controllers, microcontrollers, current sensors, and modular power stages. With strong support for wide bandgap devices and reference designs optimized for both HV and LV conversion paths, we help design engineers improve efficiency and power density and accelerate time to market.

How does power-path protection safeguard AI server racks?

Power-path protection is also essential to reliable server rack power management. Infineon’s hot-swap and eFuse solutions provide controlled power ramp-up, protect the IBC, and help maintain system stability in high-density AI server racks.

Infineon offers a range of hot-swap and eFuse devices for medium-voltage protection, including devices from the XDP700 series. For high-voltage protection, the portfolio includes hot-swap solutions based on SiC JFET devices.

Effective server power management requires precise, low-voltage, high-current power delivery to support GPUs, CPUs, TPUs, LPUs, ASICs, and custom AI accelerators under demanding computational loads. 

Voltage regulator modules (VRMs) step down the output from the intermediate bus converter (IBC) to meet the specific voltage requirements of each processor.

Meeting the power and thermal demands of AI processors

As compute performance increases, power delivery networks must provide tighter voltage regulation, faster transient response, and higher current density. These capabilities help maintain processor performance, energy efficiency, and reliable operation as workloads change.

Increasing power density also creates significant thermal challenges within limited board space. Effective xPU and ASIC power solutions therefore combine the following to deliver stable power under demanding AI workloads: 

  • High-efficiency VRM design
  • Optimized power-stage integration
  • Advanced thermal management 

By extending server rack power management from the IBC to the processor, Infineon supports efficient and reliable power delivery for next-generation AI servers.

While traditional racks operated at tens of kilowatts, today’s AI racks exceed 100 kW, with future designs moving toward 600 kW and megawatt-scale power levels. This rapid increase in GPU power demand is pushing conventional server rack architectures to their physical, electrical, and thermal limits.

For design engineers, delivering more power is only part of the challenge. Increasing power and thermal density must be managed across PSUs, IBCs, and VRMs, often within smaller system footprints. Higher distribution voltages and lower currents are also becoming essential to reduce power losses and improve server power efficiency across the rack.

How does Infineon enable efficient power delivery across AI server racks?

Infineon’s broad server rack power management portfolio helps engineers address these challengesacross power conversion and distribution. Our portfolio combines Si, SiC, and GaN power switches with: 

  • Gate drivers
  • Integrated power stages
  • Auxiliary power solutions
  • Current sensors
  • Digital controllers
  • Microcontrollers

This system-level approach enables scalable power delivery, optimized efficiency, and seamless integration for next-generation server racks.

As AI server racks push towards megawatt-class power, traditional PSU architectures must undergo a redesign. Next-generation racks adopt high-voltage DC (HVDC) conversion, shifting from ~400 V AC to 800 V DC. To support these transitions, PSU switch ratings must scale from 650 V to 1.2 kV, enabling advanced three-phase topologies such as B6 or Vienna PFC for higher efficiency. 

Power density is only part of the challenge. PSU form factors remain tightly constrained, making effective cooling and space utilization critical to server rack power management. Sidecar placement can free up vertical rack space and provide greater flexibility for cooling high-power systems.

How do wide-bandgap semiconductors improve PSU efficiency?

Infineon’s CoolGaN™ and CoolSiC™ devices reduce switching losses in the AC-DC stages of AI server PSUs. Combined with gate drivers, digital isolators, auxiliary power solutions, and microcontrollers, they support single-phase and three-phase designs across multiple topologies.

Furthermore, modular and SMD packages, together with reference designs, help engineers improve server power efficiency, accelerate development, and reduce time to market.

BBUs provide immediate power when the main PSU or grid supply is interrupted, maintaining reliable server rack power management during outages. For example, to meet 48 V system requirements and a demand for 90–240 seconds of backup time, BBUs must deliver high efficiency and a scalable architecture. 

Infineon’s patented BBU technologies use 40 V and 80 V MOSFET-based designs to improve server power efficiency and power density while reducing BOM costs. 

The scalable architecture supports multiple power levels within constrained voltage ranges, helping engineers meet the backup power requirements of next-generation AI server racks.

IBCs provide efficient voltage conversion between the PSU and point-of-load (POL) converters. In high-density AI server racks, they must combine high power density, strong thermal performance, and scalable conversion efficiency.

Infineon supports reliable IBC designs with CoolGaN™ switches, gate drivers, digital controllers, microcontrollers, current sensors, and modular power stages. With strong support for wide bandgap devices and reference designs optimized for both HV and LV conversion paths, we help design engineers improve efficiency and power density and accelerate time to market.

How does power-path protection safeguard AI server racks?

Power-path protection is also essential to reliable server rack power management. Infineon’s hot-swap and eFuse solutions provide controlled power ramp-up, protect the IBC, and help maintain system stability in high-density AI server racks.

Infineon offers a range of hot-swap and eFuse devices for medium-voltage protection, including devices from the XDP700 series. For high-voltage protection, the portfolio includes hot-swap solutions based on SiC JFET devices.

Effective server power management requires precise, low-voltage, high-current power delivery to support GPUs, CPUs, TPUs, LPUs, ASICs, and custom AI accelerators under demanding computational loads. 

Voltage regulator modules (VRMs) step down the output from the intermediate bus converter (IBC) to meet the specific voltage requirements of each processor.

Meeting the power and thermal demands of AI processors

As compute performance increases, power delivery networks must provide tighter voltage regulation, faster transient response, and higher current density. These capabilities help maintain processor performance, energy efficiency, and reliable operation as workloads change.

Increasing power density also creates significant thermal challenges within limited board space. Effective xPU and ASIC power solutions therefore combine the following to deliver stable power under demanding AI workloads: 

  • High-efficiency VRM design
  • Optimized power-stage integration
  • Advanced thermal management 

By extending server rack power management from the IBC to the processor, Infineon supports efficient and reliable power delivery for next-generation AI servers.