Overview

Navitas Semiconductor has introduced an ultra-high-voltage 2,300V SiC power module family together with a 24kW, 800V-to-12V DC-DC power delivery board built for NVIDIA MGX AI data center systems. The announcement, dated October 24, 2026, marks a full-scale entry by the company into the AI data center infrastructure power market. Critically, the products are described as moving beyond development concepts into deployed hardware, linking Navitas' roadmap directly to systems that are being fielded rather than merely planned. For a company that remains unprofitable, that distinction matters: hardware deployment is the mechanism through which a US$450 million design win backlog can begin converting into quarterly revenue.

Technical Details: SiC Modules and the GaNFast Board

The launch spans two distinct device technologies. On the high-voltage side, Navitas has created a 2,300V silicon carbide power module family, pushing well beyond the voltage classes typically associated with mainstream power conversion. On the board level, the company is offering a 24kW, 800V-to-12V DC-DC power delivery board based on its GaNFast gallium nitride technology. The architectural significance of the board lies in what it removes rather than what it adds: the design eliminates an entire intermediate voltage conversion step, taking 800V input directly to 12V output. Each removed conversion stage represents a potential reduction in conversion losses, component count, and board footprint.

  • 2,300V SiC module family: ultra-high-voltage silicon carbide devices for demanding conversion roles.
  • 24kW DC-DC board: 800V-to-12V delivery aimed at AI rack power.
  • GaNFast technology: gallium nitride platform underpinning the board's direct conversion topology.
  • Direct conversion: removal of an intermediate voltage step between the 800V bus and the 12V rail.

The Shift to 800V HVDC Rack Architectures

The new board expands Navitas' 800V HVDC power portfolio for AI rack applications. This is not an isolated product decision but a response to a broader architectural migration in data centers, where rack-level power delivery is being reorganized around higher distribution voltages. An 800V high-voltage DC bus reduces the current required to deliver a given amount of power, which in turn eases the copper and conduction-loss penalties that scale with AI rack densities. The remaining challenge is the final step down to the low-voltage rails that processors and memory actually consume. By performing that step directly from 800V to 12V, Navitas is targeting the part of the power chain where conversion stages have historically accumulated.

Market Context: AI Demand Against a Soft Industrial Backdrop

Navitas' positioning is deliberately asymmetric. On one side, demand for high-voltage SiC and GaN power semiconductors is rising as AI infrastructure scales, giving the US$450 million order backlog a plausible path into reported earnings. Analysts cited in the coverage project revenues of US$143.2 million and earnings of US$24.0 million by 2029, assuming annual top-line growth of 52.3%. On the other side, the company remains unprofitable, and its traditional downstream markets — electric vehicles, solar, and industrial — are described as sluggish, creating persistent pressure on short-term results and ongoing risks of restructuring and cost absorption. Execution on the AI platforms is therefore not merely a growth story; it is the variable that determines whether margin stability is achievable.

Industry Impact and Competitive Positioning

The linkage to NVIDIA-backed architecture is the strategic center of gravity here. A power component qualified into an AI rack platform is validated against a demanding set of electrical, thermal, and reliability requirements, and that validation carries weight with other system designers. Navitas is effectively using a high-density AI power socket to demonstrate that its SiC and GaN portfolio can serve the most demanding conversion roles, not just consumer or automotive applications. The announcement also signals that the power semiconductor contest in AI data centers is broadening beyond compute silicon into the delivery network that feeds it — an area where voltage class, conversion topology, and integration density are the competitive axes.

Implications for AI Data Centers

For operators, the appeal of a direct 800V-to-12V board is straightforward: fewer conversion stages in the path between the facility distribution bus and the load. In dense AI racks, every stage consumes board area, generates heat, and adds a failure point. Removing an intermediate step simplifies the power tree and can improve the overall efficiency of the delivery chain. The 24kW rating of the board also places it in the range relevant to high-power AI rack deployments, where the power delivery subsystem must keep pace with the aggregate draw of accelerators and their supporting components. The 2,300V SiC modules extend the same logic to higher-voltage roles within the broader system.

Future Outlook and Investor Considerations

The investment case now hinges on conversion rather than announcement. Navitas holds a US$450 million design win backlog, and the central question is how quickly AI-related programs translate into recognized revenue. The bull case is that AI power board demand offsets weakness in EV, solar, and industrial markets, lifting margins along the way. The bear case is that cost pressure and execution risk dominate at a still-unprofitable company. Investor sentiment captured alongside the analysis was decidedly cautious, with 28% bullish and 72% bearish across 393 participants. The metrics to watch are the quarterly proportion of AI-related revenue and the trajectory of margins, since those two data points will reveal whether the new hardware platforms are becoming the earnings driver that the 2029 projections assume.