A new era of AI infrastructure is unfolding as Navitas Semiconductor integrates GaN and SiC power technologies into NVIDIA’s 800-volt AI factory ecosystem, boosting energy efficiency, performance, and scalability for next-generation Rubin Ultra GPUs and Kyber rack-scale systems.

AMD and Navitas are taking complementary roles in this high-power AI era: AMD drives AI compute with new processors and GPUs, while Navitas enables efficient power delivery with GaN and SiC technologies for AI infrastructure.

The technology delivers higher power density and better thermal management, with peak PFC efficiency up to 99.3%, up to 30% lower power losses versus silicon solutions, and a potential 45% reduction in copper use and cooling costs.

Environmentally, the approach reduces energy consumption and cooling needs, contributing to green AI efforts and potentially expanding access to high-performance computing by lowering operating costs.

Navitas expects gradual growth in 2026 as high-power programs ramp, with a larger surge anticipated in 2027 when new AI power designs reach volume adoption.

The shift from 54V to 800VDC power delivery enables direct conversion from utility power to 800V, reducing conversion stages and boosting overall efficiency, while lowering cooling and copper requirements.

Overall, the push balances computational advances with energy efficiency to sustainably scale AI, cutting environmental impact while enabling breakthroughs in large-scale models and applications.

Navitas projects higher margins for high-power products due to their value and multi-generation demand, signaling a stronger long-term profitability trajectory versus traditional power electronics peers.

Beneficiaries include hyperscale cloud providers, enterprises, AI research labs, and startups that need higher compute density and sustainable power delivery, with open ecosystems and cost efficiency shaping market competition.

This transition to 800V HVDC architecture aims to cut the number of conversion steps, with NVIDIA projecting up to a 5% improvement in overall power efficiency.

Industry-wide implications include accelerated adoption of wide-bandgap technologies in data centers and a shift in competitive dynamics among AI infrastructure providers, impacting big tech players and startups alike.

The move could benefit major AI operators through more efficient infrastructure and may drive broader adoption of 800V architectures and GaN/SiC components across data centers and edge deployments.