Navitas and NVIDIA link up on a 24-kW 800V-to-12V power board aimed at next-gen AI systems

Navitas Semiconductor’s partnership with NVIDIA is centered on power conversion hardware, an area that is increasingly critical as data centers and AI server designers try to deliver more performance per rack while controlling heat, energy use, and physical space.

According to a report by Yahoo Finance, the companies are working on a 24-kilowatt 800V-to-12V power board. In plain terms, the board is intended to take input power at an 800-volt level and convert it down to 12 volts for components that require lower voltage, such as processors and other electronics within AI systems.

The goal, as described in the same report, is to improve efficiency and reduce the amount of board space needed. That matters because AI server designs often have to pack more computing into the same or similar form factors, while also managing power distribution and cooling constraints that can become bottlenecks as power draw rises.

The report frames the effort as part of Navitas’ broader 800V strategy. Higher-voltage architectures, commonly discussed across the data center industry, are generally aimed at reducing current for a given power level, which can lower losses in power delivery and potentially simplify or reduce the size of certain power components. In practice, executing an 800V approach depends on having power conversion solutions that can reliably handle the voltage range and still meet tight performance and packaging targets.

For NVIDIA, which supplies the GPUs and related platforms that sit at the center of many AI systems, power electronics are not just a supporting cast. The GPU and accelerator ecosystem is highly dependent on stable power delivery across varying loads, and board-level power conversion design can influence system-level efficiency and thermal headroom.

Navitas, in this context, is positioning its power technology as a building block that can integrate into AI system power architectures. The reported focus on a specific board class, a 24-kilowatt 800V-to-12V solution, suggests the work is aimed at workloads and chassis power envelopes typical of next-generation AI deployments, where power distribution must scale and remain efficient as utilization patterns change.

What the disclosed information does not spell out is just as important. The report and related public materials referenced in the packet did not provide additional specifics such as production timelines, whether the board is offered as a customer-ready product or a reference design, pricing, or any measurable outcomes from testing at the system level.

It also remains unclear how broadly the NVIDIA partnership is expected to apply, for example whether it is limited to a particular NVIDIA platform, server partner design, or evaluation phase. Without those details, the partnership can be interpreted as a technical alignment step, aimed at validating that Navitas’ 800V-to-12V power approach can meet the integration and performance expectations of AI hardware builders.

Looking ahead, market participants will likely watch for follow-on announcements that move from component and reference design discussions to deployment indicates. That could include customer qualification timelines, any mention of standardized integration within AI server designs, and disclosed efficiency or packaging results tied to real system trials.