[CLSTR]
< Back to situation

[REVISION HISTORY]

Advancements in GaN technology for AI power systems

Updated 2 times since CLSTR started tracking revisions of this situation.

What changed

2026-09-01 09:17 UTC → 2026-09-03 01:31 UTC · added removed

Developments in gallium-nitride (GaN) technology are advancing to meet the power demands of AI infrastructure and high-voltage systems. Power Integrations announced the introduction of PowiGaN devices rated up to 2200V, a level intended to provide “substantial voltage margin for emerging high-voltage power systems” while competing with silicon-carbide devices in sectors like solar and HVDC infrastructure. Following this, Efficient Power Conversion (EPC) began mass production of the EPC2370, an 18V eGaN FET designed to support the transition toward 800 VDC power distribution in AI server power supplies. Expanding on high-voltage capabilities, researchers at EPFL’s POWERlab have developed a new GaN power-device design called an intrinsic polarization superjunction (iPSJ). This method utilizes GaN’s natural polarization properties to create both a two-dimensional electron gas (2DEG) and a two-dimensional hole gas (2DHG), rather than relying on chemical doping. This charge balance allows the electric field to spread evenly across the device, addressing the typical breakdown limitations of GaN electronics. Testing of these Schottky barrier diodes showed breakdown voltages exceeding 3.9 kV while maintaining low on-resistance and stability at temperatures up to 125°C. Such advancements may enable more compact The chip-sized iPSJ device, constructed from GaN layers on a low-cost silicon base, can withstand nearly 4 kV before breakdown, setting a record for this technology. This combination of high-voltage capability and efficient efficiency is considered essential for power conversion for electric vehicles, in renewable energy, energy systems and AI infrastructure. data centres.

Versions

  1. 2026-09-03 01:31 UTC Advancements in GaN technology for AI power systems
  2. 2026-09-01 09:17 UTC Advancements in GaN technology for AI power systems
  3. 2026-08-25 00:52 UTC Advancements in GaN technology for AI power systems

Only revisions since CLSTR began indexing content versions appear here. Select a version to see what changed compared to the one before it.