Infineon Converters Boost AI Data Center Power Efficiency
New gallium nitride devices enable the 800-volt distribution required for upcoming one-megawatt AI compute racks.
Updated on Oct. 21, 2026 in Data Centers
Live Poll
Do you expect new AI data center power technologies to lower energy costs for your household?
Infineon Technologies has released new gallium nitride power converters designed to manage 800-volt electricity for AI data center infrastructure. These components enable the high-efficiency power delivery required for upcoming high-density computing racks.
Why it matters
As data centers transition to 800-volt power distribution to reduce current and conductor demand, these converters address a critical bottleneck in scaling infrastructure for massive AI compute loads. This development supports the move toward the one-megawatt racks expected to go online in 2027.
The gallium nitride converters switch at frequencies up to one megahertz, achieving 98% efficiency at full load for 800-to-50-volt conversion. This high-frequency switching reduces the physical footprint of supporting passive electrical components.
The players
Infineon Technologies
A semiconductor manufacturer specializing in power systems and sensor technologies for automotive and industrial markets.
NVIDIA
A designer of graphics processing units and specialized AI hardware that leads the global market for data center compute.
A hyperscale data center operator contributing to open-standard hardware specifications through the Open Compute Project.
Microsoft
A cloud infrastructure provider currently developing 800-volt power requirements alongside industry partners.
The details
The devices function by stepping down 800-volt high-voltage supply lines to usable voltages of 50, 12, or six volts before power reaches processors. Gallium nitride (GaN) — a wide-bandgap semiconductor material — allows these devices to switch at high frequencies, which enables smaller supporting electrical components like inductors and capacitors. This voltage reduction is necessary to manage the immense power throughput required for modern GPU clusters while minimizing the thermal load on the power delivery chain.
Timeline
May 2025: Infineon and NVIDIA began their collaboration on power chain components.
March 2026: Official reference designs for the converters were released.
October 21, 2026: Infineon highlighted the new converter technology.
2027: NVIDIA targets the deployment of one-megawatt power racks.
The Tech Race
This release aligns with the power efficiency standards established by the Open Compute Project for next-generation AI infrastructure. It positions these converters as foundational for the industry-wide push toward higher-voltage architectures.
These components will be implemented in high-density AI data centers starting in 2027, primarily affecting cloud infrastructure operators and hyperscalers. The converters enable the operation of one-megawatt power racks that demand higher voltages to function efficiently.
The takeaway
The industry is shifting toward 800-volt distribution to handle the extreme power demands of modern AI compute. Watch for the mass deployment of one-megawatt racks in 2027 as the primary indicator that these high-efficiency power chains are meeting production requirements.
Further reading
For more on evolving infrastructure requirements, visit the Data Centers section.
Source note: This article includes information reported by TokenPost.
Live Poll
Do you expect new AI data center power technologies to lower energy costs for your household?
