Infineon’s broad power semiconductor portfolio includes robust solutions for every stage of power conversion,from the utility grid to the processor core. By combining Si, SiC, and GaN technology with advanced packaging and a system-level approach, we are able to deliver the highest efficiency, density and robustness.  As rack power demands push toward 1 MW and beyond, Infineon leads the architecture evolution, shaping the next generation of AI data centers from grid to core.

Powering AI data centers from grid to core

The exponential data growth driven by digitalization and artificial intelligence is significantly increasing the energy and technology requirements of data centers.

Our innovative portfolio of power semiconductors includes solutions that address each power conversion stage from the utility grid to the data center core, the AI processor, leveraging the benefits of Si, SiC and GaN to achieve the highest efficiency, density and robustness. Examples of such applications include server rack power supply units (PSUs), battery backup (BBU) and capacitor backup (CBU) systems, intermediate bus converters (IBCs) and power-path protection solutions as well as high density DC-DC power delivery solutions including vertical power delivery (VPD) for next generation xPUs and SoCs for AI accelerator cards, server motherboards, network switches and smartNICs and storage platforms. Our breadth of solutions covers requirements for today’s traditional 48 V server rack architectures as well as the evolution toward high-voltage DC (HVDC) +/-400 V and 800 V distribution architectures used in high power AI server side car and IT racks, and the adoption of solid state transformers (SSTs) and solid-state circuit breakers (SSCBs) in DC microgrid implementations.

Additionally, with our novel power system reliability modeling solutions, data centers can maximize power supply reliability and uptime, enabling real-time power supply health monitoring based on dynamic system parameter logging.

lorem ipsum
lorem ipsum
lorem ipsum

Si, SiC and GaN: a hybrid approach to reduced TCO

Silicon (Si), silicon carbide (SiC), and gallium nitride (GaN) each address distinct challenges across the AI data center power architecture. Si provides a cost-effective and reliable foundation, particularly well suited for high-current, mature designs. SiC enables highly efficient power conversion in high-voltage stages such as AC-DC and intermediate bus architectures, improving thermal performance and reducing losses. GaN delivers exceptional switching speed and power density, making it ideal for high-frequency conversion and compact power modules.
Together, these complementary technologies enable higher efficiency, greater power density, and ultimately a lower total cost of ownership across the entire power delivery network.

Discover the whitepaper

The future of AI data center power architecture

The traditional architecture integrating power delivery, backup power and IT payload in a single rack is facing challenges in meeting the demand of high-performance computing. These power levels mandate a change in architecture with a move towards high voltage DC. This is a transition that we expect to happen in three phases as shown in the image below.

Architectural evolution beyond 250kW, and up to 1MW per rack
Architectural evolution beyond 250kW, and up to 1MW per rack
Architectural evolution beyond 250kW, and up to 1MW per rack

In the future we expect a new power architecture to emerge: the DC Microgrid. By integrating direct current distribution inside the data center, it enables seamless connection of diverse power sources such as solar, fuel cells and battery energy storage, reducing energy dependency on the utility grid. The reduction in the number of conversion steps eliminates the associated conversion losses to increase overall efficiency. With high-voltage DC (either +/-400V or +800V), it supports dense AI racks, liquid cooling, advanced backup solutions such as Uninterruptible Power Supplies (UPS), Battery Energy Storage System (BESS) and Solid-State Transformers (SST) and Solid-State Circuit Breakers (SSCB) for protection.

Looking ahead, the combination of renewable power source integration, energy efficiency and scalable high-power delivery makes the DC Microgrid a cornerstone of future-ready AI data centers.

We power AI - Data center power distribution map
We power AI - Data center power distribution map
We power AI - Data center power distribution map

Latest from Infineon on AI Data centers

We Power AI Podcast: Expert insights on AI Power

Dive into the key challenges AI brings and explore the different ways to address them, guided by our team of experts.

We power AI with Adam White

AI is the transformational technology of our time enabling many important use cases like Large Language Models, Health Care or Financial Applications. But we don’t talk enough about what it requires to power all these applications. Why is this topic of such importance? Behind the brilliance of AI lies a compute power and electricity-intensive process with a staggering CO2 footprint. We at Infineon are proud to work on the forefront of energy efficient, robust solutions to power AI from the grid to the core.

Dive into the key challenges AI brings and explore innovations and solutions to address them, guided by our team of experts.

Listen to our expert episodes

Powering AI at scale: grids, HVDC, and what comes next

AI is pushing our energy grid to its limits, but keeping up requires a sustainable approach, backed by the latest advancements in semiconductor technology. In this episode of our We Power AI series, Kelsey talks with Peter Wawer, Division President of the Green Industrial Power Division at Infineon, about what it really means to power AI sustainably, from the grid down to the server rack. They cover why battery energy storage, HVDC, and DC-powered data centers matter, and look ahead to what “physical AI” could mean for industry, robotics, and society. Learn more about how Infineon is powering AI.

Data center power solutions

Powering the gigawatt era: The shift to DC grids and solid-state infrastructure in AI data centers

In this episode, host Kelsey Markl welcomes Head of Systems at Infineon’s Green Industrial Power Division Mehrdad Baghaie Yazdi back to the show, where they discuss the massive power demands of gigawatt-scale AI data centers and the necessary shift toward efficient DC grid architectures. They explore how solid-state transformers and advanced semiconductors are replacing traditional infrastructure to reduce energy losses and improve grid resilience.

Data center power solutions

We power AI with Adam White

AI is the transformational technology of our time enabling many important use cases like Large Language Models, Health Care or Financial Applications. But we don’t talk enough about what it requires to power all these applications. Why is this topic of such importance? Behind the brilliance of AI lies a compute power and electricity-intensive process with a staggering CO2 footprint. We at Infineon are proud to work on the forefront of energy efficient, robust solutions to power AI from the grid to the core.

Dive into the key challenges AI brings and explore innovations and solutions to address them, guided by our team of experts.

Listen to our expert episodes

Powering AI at scale: grids, HVDC, and what comes next

AI is pushing our energy grid to its limits, but keeping up requires a sustainable approach, backed by the latest advancements in semiconductor technology. In this episode of our We Power AI series, Kelsey talks with Peter Wawer, Division President of the Green Industrial Power Division at Infineon, about what it really means to power AI sustainably, from the grid down to the server rack. They cover why battery energy storage, HVDC, and DC-powered data centers matter, and look ahead to what “physical AI” could mean for industry, robotics, and society. Learn more about how Infineon is powering AI.

Data center power solutions

Powering the gigawatt era: The shift to DC grids and solid-state infrastructure in AI data centers

In this episode, host Kelsey Markl welcomes Head of Systems at Infineon’s Green Industrial Power Division Mehrdad Baghaie Yazdi back to the show, where they discuss the massive power demands of gigawatt-scale AI data centers and the necessary shift toward efficient DC grid architectures. They explore how solid-state transformers and advanced semiconductors are replacing traditional infrastructure to reduce energy losses and improve grid resilience.

Data center power solutions
Key resources

FAQs

Infineon develops the power semiconductor solutions that manage and convert energy at every stage of an AI data center from the utility grid entering the facility to the processor core running the AI workload. Our portfolio includes Silicon (Si), Silicon Carbide (SiC) and Gallium Nitride (GaN) technologies, with solutions for every part of the power path: power distribution, power supply units (PSU), battery backup units (BBU), capacitor backup (CBU), intermediate bus converters (IBC), power path protection and high density voltage regulation for xPUs and SoCs. This hybrid technology approach enables AI data centers to achieve the highest levels of efficiency, power density and reliability across the full power delivery chain.

High-voltage direct current (HVDC) is the next generation of power distribution architecture for AI data centers. Traditional data centers operate on 48 V distribution at the rack level, which is no longer sufficient to support the power levels required by modern AI processors that are expected to exceed 1 MW per rack within this decade. HVDC architectures, operating at either +/-400 V or +800 V, reduces the number of power conversion stages, lowers conversion losses and enables the higher power densities required by dense AI server racks. Infineon is actively developing 800 V HVDC solutions in collaboration with the main industry players, including NVIDIA, to accelerate this transition

The DC Microgrid is a data center power architecture that replaces the traditional mix of AC and DC distribution with a unified direct current network across the entire facility. By distributing power at high-voltage DC levels (800 V and potentially higher), the DC Microgrid eliminates unnecessary AC-to-DC conversion steps, reducing energy losses and improving overall efficiency. It also enables seamless integration of diverse energy sources including solar panels, fuel cells and battery energy storage systems, reducing dependency on the utility grid. For AI data centers operating at gigawatt scale, the DC Microgrid represents the most scalable and sustainable power architecture available and Infineon is building the semiconductor solutions to make it possible.

Infineon anticipates that the transition to higher voltage architectures will happen in three phases. The first phase optimizes the existing 48 V distribution architecture, improving efficiency and power density within current infrastructure. The second phase introduces a transition toward three-phase HVDC with sidecar racks, disaggregating power delivery and backup power from the IT payload to improve scalability and efficiency, enabling rack power levels beyond 500 kW. The third phase completes the shift to a future architecture based on a hybrid DC microgrid with centralized power delivery and backup infrastructure, minimizing conversion stages while supporting rack power levels exceeding 1 MW. Across all three phases, Infineon provides the Si, SiC and GaN semiconductor solutions needed at each conversion stage from the utility grid to the processor core.

Explore Infineon’s Data Center Power Solutions

Infineon is the only semiconductor company offering advanced solutions across silicon (Si), silicon carbide (SiC), and gallium nitride (GaN), covering the entire power path - from the grid to the processor core - for both current and emerging architectures. Explore our AI Data Center Power Solutions hub to navigate our offering: interact with the block diagram, discover key functional blocks, and find recommended products and technologies for every stage of the power flow.

Visit application page