Key takeaways Navitas unveils 24kW power board targeting 97% efficiency Nvidia collaboration eliminates an intermediate power conversion stage New technology targets AI data center electricit
Key takeaways
- Navitas unveils 24kW power board targeting 97% efficiency
- Nvidia collaboration eliminates an intermediate power conversion stage
- New technology targets AI data center electricity losses
Navitas Semiconductor (NASDAQ: NVTS) has unveiled a 24-kilowatt power-delivery system developed with Nvidia, targeting 97% efficiency as the artificial intelligence industry searches for ways to reduce electricity losses in increasingly powerful data centers.
The new board converts 800 volts of direct current directly into 12 volts, eliminating an intermediate conversion stage traditionally used in AI computing systems. According to Navitas, the technology could improve power density, reliability and electrical efficiency while reducing the space occupied by power equipment.
The announcement, made October 12, highlights how AI infrastructure investment is expanding beyond GPUs into the electrical systems supporting them.
Nvidia Partner Navitas Targets 97% Power Efficiency
According to Navitas's official announcement, the new power-delivery board can provide 24kW to an entire compute tray while targeting 97% efficiency at full load.
Its power density exceeds 2,000 watts per cubic inch, and the board fits Nvidia's 400mm-by-60mm form factor, allowing installation close to AI processors.
The design uses 16 gallium nitride transistors from Navitas's GaNFast portfolio, reducing power-conversion complexity through a single-stage architecture.
At the stated efficiency target, approximately 3% of the electricity entering the board would be lost during conversion. For a hypothetical 24kW output, that represents roughly 742 watts of losses, assuming 97% efficiency is achieved.
Navitas plans to showcase the technology within Nvidia's AI Factory MGX ecosystem at the October 2026 OCP Global Summit.
Why AI Data Centers Need More Efficient Power Systems
Electricity is becoming an increasingly important constraint on AI infrastructure expansion, particularly as companies deploy larger computing clusters.
Even companies with sufficient Nvidia GPUs may face delays caused by unavailable grid connections, transformers and electrical equipment. These problems have already raised questions about whether AI data centers can expand quickly enough to meet demand.
Higher-efficiency power systems cannot solve electricity shortages, but they can reduce the energy wasted between incoming power supplies and computing equipment.
Navitas previously introduced a 20kW board converting 800V directly into 6V, targeting 97.5% peak efficiency. The new 24kW design expands its portfolio by supporting a different voltage architecture rather than simply replacing the earlier product.