The breakthrough efficiency and density enable AI data centers to scale power while cutting cooling costs and carbon emissions, a critical advantage as AI workloads explode.
The exponential growth of generative‑AI models is driving data centers toward ever‑higher power envelopes. Traditional silicon‑based converters struggle to keep up with the voltage and thermal constraints of 800 V DC architectures that promise lower distribution losses and smaller cabling. Industry analysts predict that high‑voltage DC (HVDC) will become the backbone of next‑generation AI facilities, but the transition hinges on converters that can deliver megawatt‑scale power without sacrificing efficiency or footprint. This shift creates a fertile market for wide‑bandgap semiconductors that can operate at higher frequencies and temperatures.
Navitas Semiconductor’s new 10 kW DC‑DC platform answers that need with a full‑brick module measuring just 61 × 116 × 11 mm. Leveraging 650 V and 100 V GaNFast FETs in a three‑level half‑bridge topology, the converter achieves a peak efficiency of 98.5 % and maintains 98.1 % at full load, while a 1 MHz switching rate drives the power density to 2.1 kW per cubic inch. The design also integrates auxiliary power and control circuitry, simplifying board‑level integration for both 800 V‑to‑50 V and ±400 V‑to‑50 V configurations.
The platform’s performance metrics translate into tangible business benefits: reduced cooling requirements, lower total‑cost‑of‑ownership, and a smaller carbon footprint—attributes that align with the sustainability goals of hyperscale operators. Early collaborations with leading data‑center owners suggest rapid adoption, and Navitas plans to showcase the solution at APEC 2025. As AI workloads demand 100‑ to 1,000‑fold compute increases per query, the ability to scale power delivery efficiently will become a competitive differentiator, positioning GaN‑based converters as the preferred choice over legacy silicon solutions.
Comments
Want to join the conversation?
Loading comments...