logo
Últimas noticias de la empresa sobre Solid-State Transformer Goes Commercial: The 800V DC Power Revolution for AI Data Centers

August 17, 2026

Solid-State Transformer Goes Commercial: The 800V DC Power Revolution for AI Data Centers

Solid-State Transformer Goes Commercial: The 800V DC Power Revolution for AI Data Centers

The 800V Future of AI Data Center Power

As AI cluster rack power climbs to 350–500 kW, the conventional power-transformer-plus-UPS supply chain is running into hard physical constraints. In the second half of 2026, the solid-state transformer (SST) — the architecture NVIDIA has described as the endgame for next-generation AI data center power — has moved out of the laboratory and into commercial deployment.

Sungrow, Zhiguang Electric, Xinfengguang, Delta and more than 40 vendors have launched products and signed demonstration orders in quick succession, building a clear industry consensus: 2026 is the first year of SST commercialization.

Why AI Is Forcing a Power Architecture Shift

AI model training keeps pushing data center energy demand higher. Forecasts suggest global data center electricity consumption could reach roughly 3% of worldwide consumption — about 1,500 TWh — by 2030. With super-node hardware accelerating into mass production, 350 kW and 500 kW racks are now a reality.

  • 48V dilemma: delivering 350 kW at 48V would draw up to 7,290 A — massive current that means heavy cabling, high losses and expensive distribution space.
  • 800V answer: moving to 800V high-voltage DC cuts current to 437.5 A, roughly 1/17 of the original — the key reason ByteDance's AI Rack 3.0 standardizes on 800V HVDC.

Yet 800V HVDC is only a transitional step. The SST is the real architectural breakthrough: by replacing the traditional iron core with silicon-carbide (SiC) power electronics, it converts 10 kV medium-voltage AC directly to 800V DC in one stage, raising system efficiency above 98.5% and cutting equipment volume by half. Sungrow's EnerNeo SST and Zhiguang Electric's 4.2 MW Coota SST both target this core capability.

Global Leaders Enter the Race

CompanyPositioningHighlight
ABBGlobal SST leader8×13.8kV/5MW SSTs (40MW) delivered to a Google data center
SiemensEuropean standard-setterLeads IEC 62800 SST standard; 12MW Deutsche Telekom order
EatonAI data center SST leaderMV SST 2.0: 10kV to 800V DC, 2.5MW, 98.5% efficiency
SungrowAIDC commercialization pioneerEnerNeo SST on sale; first 800V DC SST public tender
Heron PowerMulti-port SST newcomer40GW SiC SST gigafactory planned; 50GW early orders
DG MatrixSiC SST innovatorInterport power router, up to 98.5% efficiency

Challenges and the Road Ahead

Commercialization does not mean the technology is risk-free. Experts point to several hurdles that must be cleared for engineering-grade deployment: long-term reliability of high-voltage SiC devices, high-frequency insulation, and modular redundancy control. SiC devices alone account for 30–35% of overall system cost, making their price decline a key enabler for mass adoption.

Industry estimates project the SST market to expand from roughly RMB 10 billion in 2026 to over RMB 1 trillion by 2035, with global new installed capacity exceeding 80 GW by 2030. The consensus roadmap is clear: 2026 is the commercialization kickoff, 2027 brings volume ramp-up, and 2028 onward opens a decisive battle for market structure. The AI computing revolution has lit the fuse — but engineering capability, cost control and multi-scenario deployment will ultimately decide who leads.