PCIM 2026 Confirms Datacenter Power as the Semiconductor Industry's Most Pressing Challenge

800V DC, GaN vs SiC, and the Technologies Reshaping AI Datacenters

PCIM 2026 confirmed a major shift in the semiconductor industry: datacenter power architecture has become one of the most critical challenges of the AI era. As next-generation AI accelerators push beyond 6,000W, the industry's focus is expanding beyond compute performance to the infrastructure required to power it. Across the conference, discussions consistently centered on 800V DC distribution, advanced power conversion, and the semiconductor technologies enabling higher-density AI datacenters.

800V DC Emerges as the New Datacenter Priority

Just one week after major hyperscalers aligned around NVIDIA's 800V DC architecture, exhibitors showcased a growing ecosystem of technologies designed to support higher-voltage datacenter power distribution. Power Distribution Boards (PDBs) were a major focus, particularly architectures that convert 800V directly to the 6V rails increasingly required by AI accelerators.

Exhibitors showcased both conventional 800V-to-50V approaches and emerging 800V-to-6V designs, highlighting the industry's push to reduce conversion stages and improve efficiency closer to the processor. While this approach can improve efficiency and reduce infrastructure complexity, the extreme conversion ratio creates significant thermal and current-management challenges that remain a key engineering hurdle for next-generation AI systems.

Solid-State Transformers Gain Momentum

Another recurring theme was the rise of Solid-State Transformers (SSTs). Several vendors demonstrated modular SST architectures capable of converting medium-voltage utility power directly into datacenter DC distribution voltages, with many demonstrations targeting 12 kV AC-to-1.5 kV DC conversion derived from megawatt EV fast-charging designs.

Many of these designs leveraged silicon carbide (SiC) devices and reflected growing interest in simplifying power delivery between the grid and AI workloads. While standards and supply chains continue to evolve, SSTs are increasingly being viewed as a potential building block for next-generation AI facilities.

The Real Story: GaN and SiC Are Both Winning

A growing reality in power semiconductors was confirmed: this is not a GaN-versus-SiC market. Many of the showcased PDB and power conversion solutions relied on 650V GaN devices, whose high switching frequencies enable smaller magnetic components and higher power densities—an increasingly important advantage as AI compute trays continue to shrink. At the same time, SiC remains the preferred technology for higher-voltage applications, including SSTs, solid-state circuit breakers, and facility-level power distribution.

The contrast was evident throughout the exhibition floor. GaN is expanding rapidly in high-density power conversion, while SiC continues to anchor the infrastructure required to deliver power at scale. Rather than competing for the same opportunities, both technologies are becoming increasingly important as datacenter power architectures evolve.

What This Means for the Semiconductor Industry

Power architecture is becoming a key competitive battleground for the AI era. As 800V DC adoption accelerates, decisions around GaN, SiC, and next-generation power systems will increasingly influence technology roadmaps, investment priorities, and long-term market positioning.