Navitas, an Nvidia partner in power semiconductors, has introduced a 24kW AI power board aimed at 97% efficiency for data center applications. The design is positioned to address rising power demands from AI accelerators and tighten integration with Nvidia's ecosystem.
Overview
On 2026-11-21, CoinMarketCap / Coinpaper.com reported that Navitas, an Nvidia partner in power semiconductors, introduced a 24kW AI power board aimed at 97% efficiency for data center applications. The announcement is notable less for a single component specification than for what it signals about AI infrastructure: power delivery is becoming a first-order design problem, and semiconductor suppliers are positioning themselves closer to the Nvidia ecosystem as AI accelerators push data center power demands upward.
The available summary frames the board as a response to rising power requirements from AI accelerators and as a step toward tighter integration with Nvidia. It does not disclose the board's topology, semiconductor material, control architecture, production timeline, or named customer deployments. Those omissions matter because, in high-power data center systems, headline kW and efficiency figures are only the beginning of the engineering story.
Technical Details and the 97% Target
The 24kW rating places the design in the high-power segment of data center power delivery. Rather than addressing edge or consumer systems, the board is positioned for dense compute environments where AI accelerators dominate the load profile. In such environments, a power board must do more than convert voltage; it must deliver large amounts of power with tight regulation, thermal discipline, and reliability.
The 97% efficiency target is equally central. At 24kW, efficiency is not a marginal marketing metric. Even a small shortfall from the target translates into substantial waste heat that must be removed from the rack or row. A 97% target therefore implies a design optimized around loss reduction, thermal management, and high-density power conversion. The summary does not state whether 97% is a peak, typical, or end-to-end figure, and that distinction will matter once the product moves from announcement to deployment.
Integration with Nvidia's Ecosystem
Navitas is described as an Nvidia partner in power semiconductors, and the new board is positioned to tighten integration with Nvidia's ecosystem. That framing is significant. In AI systems, power delivery is increasingly co-designed with the accelerator platform rather than treated as a generic utility. Tighter integration can mean closer electrical and thermal alignment between the power board and the compute payload, though the summary does not identify specific Nvidia platforms or Navitas components involved.
The partner language also suggests a strategic shift. Power semiconductor vendors are seeking to move from component supply toward reference designs and platform-level engagement. For Nvidia, such partnerships can help broaden the ecosystem of validated power solutions around its accelerators. For Navitas, the association provides credibility in a market where data center operators demand proven reliability.
Market Context: AI Accelerators Reshape Power Design
The summary explicitly ties the board to rising power demands from AI accelerators. That is the core market driver. As AI workloads scale, accelerators consume more power, and the power delivery path must scale with them. The result is pressure on board density, conversion efficiency, and thermal design.
A 24kW board aimed at 97% efficiency is a response to that pressure. It suggests that Navitas is targeting the high end of data center power delivery, where the cost of inefficiency is measured in cooling capacity and rack-level power budgets. The summary does not quantify those budgets, but the direction is clear: AI power systems are being pushed toward higher power and higher efficiency simultaneously.
Industry Impact
The announcement reflects a broader competition in data center power delivery. Efficiency claims are becoming differentiators, and suppliers are eager to demonstrate that they can handle the higher power levels associated with AI accelerators. The 24kW and 97% figures are the headline metrics in this announcement, and they are likely to be compared against alternative power designs as the market matures.
At the same time, the summary provides no competitive benchmarks, pricing, or deployment data. That limits the extent to which the board can be assessed against rival offerings. What can be said is that Navitas is attempting to stake out a position at the high-power, high-efficiency end of the AI power system market, with Nvidia partnership as a central part of the pitch.
Implications for AI Data Centers
If the 97% target is achieved in production, the implications for AI data centers would be meaningful. Higher efficiency reduces waste heat and can ease cooling burdens, while a 24kW board can support denser power delivery. Those benefits are conditional, however. The announcement does not disclose availability, production timing, or customer deployments, so the immediate impact remains prospective.
The more certain implication is strategic. Navitas is using the 24kW board to signal that it can address the power requirements of AI data centers and to align itself with Nvidia's ecosystem. That positioning matters in a market where operators increasingly look for power solutions validated against specific accelerator platforms.
Future Outlook
The next milestones for the Navitas 24kW AI power board would be independent efficiency validation, production readiness, and design wins within Nvidia-based systems. Until those details emerge, the announcement should be read as a directional signal: AI data centers are demanding more power and better efficiency, and power semiconductor vendors are responding with platform-oriented designs.
The key question is whether 97% efficiency can be sustained in real operating conditions and at scale. If it can, the board could strengthen Navitas's position in the Nvidia ecosystem and contribute to more efficient AI data center power delivery. If it cannot, the announcement will remain a specification target rather than a deployed advantage.