Enphase’s latest SST is rated for up to 1.25 MW AI load, supporting both 35 kV and 15 kV AC grid voltage
It offers a conversion efficiency and system availability of up to 98.5% and 99.999%, respectively
The company expects to demonstrate the product by the end of 2026, with commercial availability targeted for 2028
Power conversion solutions provider Enphase Energy, Inc. has developed a solid-state transformer (SST) platform, IQ SST, purpose-built for large-scale AI and hyperscale data centers.
The industrial pivot towards AI-led digital infrastructure has created exponential demand for data centers across the globe. The changing dynamics of data center workloads are pushing rack power densities from kW to MW-scale, demanding higher switching speeds, efficiency, and reliability. Enphase Energy CEO Badri Kothandaraman said, “AI is changing how power must be delivered to compute infrastructure,” where loads can swing multiple times per second from idle to full load.
Unlike conventional data centers, where grid AC is stepped down and converted to DC several times, the industry is migrating to 800 V or ±400 V DC-based designs (common DC busbar). It reduces intermediate conversion losses, saves space, and boosts conversion efficiency.
Reflecting this trend, the company’s latest SST, equipped with delta-connected 342 Gallium Nitride (GaN) based power modules, converts the medium grid voltage (35 kV or 15 kV) to 800 V DC. The company states that this single-stage conversion architecture supports both 800 V and ±400 V DC racks. While GaN-based switching enables faster, more efficient conversion than traditional counterparts, the proprietary nm-scale chips coordinate between hundreds of power modules. As a result, a single bidirectional SST rack is rated for up to 1.25 MW of full load power. The company claims that the SST can deliver a maximum conversion efficiency of 98.5% while keeping system availability at up to 99.999%. Furthermore, it eliminates the need for UPS systems and requires less floor space than conventional data center power distribution.
According to the company, this platform is expected to be demonstrated by the end of 2026, followed by a customer pilot project in 2027 and commercial shipments in 2028. It will be manufactured at the company’s US facility, helping meet the country’s annual AI data center demand of 11 GW by 2031.