ROGER FIBREOPTICAL NETWORK DOSSIERS Build a Fiber Brief

Power supplies for computing servers and AI servers

AI servers use high-capacity, high-efficiency power supplies ranging from 3 kW to over 12 kW, leveraging silicon, silicon carbide (SiC), and gallium nitride (GaN) technologies to meet extreme power density and efficiency requirements. Power Requirements and CapacityAI servers demand significantly more power than traditional servers due to dense GPU clusters and continuous high-performance workloads. A fully populated AI server rack with multiple GPUs and CPUs can consume 12–15 kW per rack, with AI data centers requiring 50–150 kW per rack in some cases.

Power ratings for individual AI server PSUs are evolving from 3 kW to over 10 kW, with projections toward 12 kW and beyond, often using three-phase AC input designs to handle higher loads.

Efficiency and Standards

Efficiency is critical to minimize energy waste and cooling requirements.

Modern A

I server PSUs aim for over 97% efficiency, with compliance to standards like 80 PLUS Titanium, which mandates over 90% efficiency across light to full load and a minimum power factor of 0. 95. High efficiency reduces heat generation, enabling higher power densities and more compact form factors.

Semiconductor TechnologiesAI server PSUs increasingly use wide-bandgap semiconductors such as SiC MOSFETs and GaN FETs, which allow higher switching frequencies, lower conduction losses, and smaller magnetic components. Traditional silicon MOSFETs are still used in some stages for cost efficiency, but SiC and GaN devices dominate high-power, high-efficiency designs. These devices support synchronous rectification and resonant converter topologies like LLC, phase-shift full-bridge, and Dual Active Bridge (DAB) to maximize efficiency and power density.

Power Supply Topologies

To meet AI server demands, PSUs employ advanced topologies: Interleaved boost converters and totem-pole PFC for AC-DC conversionThree-level Vienna rectifiers for high-voltage input stagesFull-bridge LLC and phase-shift converters for DC-DC conversionSynchronous rectification in low-voltage outputs to reduce losses Control and MonitoringModern AI server PSUs integrate Digital Signal Controllers (DSCs), such as dsPIC devices, to manage high-speed PWM, power factor correction, and predictive maintenance. These controllers enable real-time monitoring, fault detection, and firmware updates, ensuring reliable operation under extreme loads.

Smart power stages with integrated current and temperature sensing further enhance efficiency and reduce component count.

Cooling and Form Factor Considerations

High power densities, sometimes exceeding 100 W/in³, require careful thermal management.

A

I servers often use liquid cooling, rear-door heat exchangers, or direct-to-chip cooling to handle the heat generated by high-efficiency PSUs and dense GPU clusters. Despite increasing power, PSU form factors remain constrained by rack space, driving innovations in compact, high-density designs. SummaryAI server power supplies are high-capacity, high-efficiency units designed to support extreme workloads.

They leverage Si, SiC, and GaN semiconductors, advanced AC-DC and DC-DC topologies, and digital control systems to deliver reliable, efficient power at high densities.

These PS

Us are critical for enabling AI data centers to operate at scale while managing energy consumption and thermal challenges.

Power supplies for computing servers and AI servers

POWER ICs FOR AI SERVERS Selector Guide

ited for AI server power architectures. Models such as the SiC461, SiC431, and SiC450 offer wide input voltage ranges, high current capabilities, and peak efficiencies up to 98 %, enabling optimized power

Meeting the Demanding Energy Needs of AI Servers with Advanced

Utilizing high-efficiency MOSFETs, sophisticated gate drivers and dsPIC DSCs equipped with high-performance and advanced peripherals enables the development of power supplies

How to Choose an AI Server Power Supply Unit (PSU)? AI Server

Explore the differences between general servers and FSP AI server power supply solutions. Learn how these advanced power solutions optimize performance for AI-driven workloads.

AI PSU | Infineon Technologies

Infineon explores the latest advancements in AI server power supplies, addressing the challenges driven by increasing power demands and the evolution of AI server rack architectures.

AI Data Center Power Solution

Consider your system''s power requirements, redundancy needs, and thermal conditions. Our modular lineup offers scalable options to suit a variety of environments.

FSP_How to Choose an AI Server Power Supply Unit (PSU)

In the context of high-performance AI servers, the PSU must deliver both ample and stable power to drive CPUs, GPUs, or AI accelerators. Additionally, PSU efficiency is crucial, as it directly influences

AI Server Power Supplies

AI power supply solutions for data centers and computing systems with efficient power management and performance.

Magnachip Launches New 6th-Generation 600V SJ MOSFETs for AI Servers

Magnachip''s new 6th-generation 600V SJ MOSFETs for AI server, EV charging and industrial power applications.

Data Center Liquid Cooling: The AI Heat Solution

“As AI has made racks denser and hotter, liquid cooling has become the de facto solution,” Karin Overstreet, president of Nortek Data Center Cooling,

2026 AI & Semiconductor Outlook: Memory, Power, Packaging Trends

Explore the 2026 AI and semiconductor outlook, covering HBM memory, AI servers, power architectures, advanced packaging, and supply chain shifts.

Green Energy-Efficient Servers: Recommended Low-Power Models

By 2026, green, energy-efficient servers have shifted from being “optional” to “mandatory.” When selecting low-power servers, prioritize four core metrics: power consumption,

We did the math on AI''s energy footprint. Here''s the story you haven''t

The emissions from individual AI text, image, and video queries seem small—until you add up what the industry isn''t tracking and consider where it''s heading next.

Worldwide AI Server PCB Market 2026 – PW Consulting

The Global AI Server PCB Market was valued at USD 12.65 Billion in 2025 and is projected to reach USD 67.84 Billion by 2032, growing at a CAGR of 32.1%.

XLS4-SX2-LAS1 MGX Server

Intel ® Xeon ® 6 is a new generation of processors that has tailored its CPU lineup to target all uses cases such as Cloud Computing, AI, Analytics, and Edge & IOT. Whether it''s high compute density

MX Stock Heads For 29-Month Highs — What''s The AI Server

Magnachip stated that it will showcase Medium-Voltage MOSFET solutions designed for server and data center power supply units, high-performance computing applications, and industrial

How Much Electricity Does a Data Center Use?

In 2023, U.S. data centers collectively consumed 176 TWh, equivalent to powering 16 million homes for an entire year. Why do AI applications use so

What we know about energy use at U.S. data centers amid the AI boom

Most of the electricity used by data centers – about 60% on average, the IEA reports – powers the servers that process and store digital information. This is especially true at AI-optimized

Servers Storage: Data Center Services & Solutions | Lenovo US

Lenovo offers data center solutions to power your business with the latest technology. Explore efficient servers, storage, networking, cloud, edge, AI & more. Call LenovoPRO at 866-426-0911

Related Video Reference

This video was associated with the source search result. Verify technical details against current product documentation and project requirements.

Technical note

This reference is intended for preliminary fiber and communications infrastructure research. Optical interface, fiber type, port or core capacity, electrical input and output, enclosure protection, cooling, installation limits and applicable standards must be verified for the specific project.

Still Have a Technical Question?

Use the inquiry form to describe the product category, application, interfaces, capacity and installation environment.

Start a Project Inquiry