Navitas Semiconductor has completed its acquisition of Claros, adding integrated voltage regulator (IVR) technology to its portfolio of gallium nitride (GaN) and silicon carbide (SiC) power semiconductors. The combination is aimed at a growing challenge in AI infrastructure: moving increasingly large amounts of power efficiently and precisely from the grid to the processors driving modern AI systems.
AI’s next bottleneck is increasingly a power problem
As AI processors become more powerful, data-center operators face a challenge that cannot be solved by compute capacity alone. Delivering power to those processors at the required voltage, current, speed and density is becoming a critical part of the AI infrastructure equation.
Navitas Semiconductor is positioning its completed acquisition of Claros around that problem.
The company has closed its previously announced acquisition of Claros, a power-management technology company developing integrated voltage regulator technology for AI data centers. The transaction gives Navitas an IVR portfolio designed to handle the final stage of power delivery to the xPU—the processors used for AI and other accelerated workloads.
The acquisition expands Navitas’ strategy beyond power conversion technologies such as GaN and high-voltage and ultra-high-voltage SiC. The company is now targeting the full grid-to-xPU power chain, from high-voltage infrastructure down to the processor.
That matters because the electrical architecture inside AI data centers is changing alongside processor power requirements.
Moving power conversion closer to the processor
Claros’ technology integrates power transistors, digital control, inductors and capacitors into an integrated voltage-regulator stack.
The objective is to move power conversion closer to the xPU, supporting architectures such as Vertical Power Delivery (VPD) and Embedded Power Delivery (EPD).
Claros has two product approaches based on the same technology stack: Packaged Integrated Voltage Regulators (pIVRs) and Embedded Integrated Voltage Regulators (eIVRs).
The closer power conversion moves toward the processor, the more important efficiency, transient response and power density become. AI accelerators can experience rapidly changing workloads, creating demanding power-transient conditions that traditional power-delivery architectures must accommodate.
Navitas says its Claros technology is designed to address those requirements while its modular PowerArray architecture can scale power delivery as processor requirements increase.
This is a different layer of the AI infrastructure stack from the accelerator itself. Rather than improving the model or increasing compute performance directly, IVR technology helps ensure the processor can receive the power required to operate effectively.
Why xPU power delivery is becoming strategic
The industry’s shift toward larger AI systems is putting pressure on nearly every layer of data-center infrastructure.
AI servers are moving toward higher-density configurations, while accelerator and CPU platforms require increasingly sophisticated power delivery. At rack scale, this can translate into substantially greater electrical loads and more complex thermal and power-management requirements.
Navitas CEO Chris Allexandre described power as a growing bottleneck for AI infrastructure, arguing that future systems will need to deliver thousands of amps to increasingly power-hungry processors.
The company’s “power wall” framing reflects a broader engineering issue: simply supplying more electricity to an AI rack is not enough. Power must be converted and distributed efficiently across multiple voltage levels while responding rapidly to processor demand.
This is where Navitas sees an opportunity to combine its existing semiconductor technologies with Claros’ IVR capabilities.
From 800V infrastructure to the xPU
Navitas has been developing GaN and high-voltage and ultra-high-voltage SiC technologies for emerging 800V DC data-center architectures.
GaN can support efficient power conversion at high switching frequencies, while SiC is used in higher-voltage power applications. Adding IVR technology extends that chain into the lower-voltage portion immediately surrounding the processor.
The strategic proposition is therefore less about a single semiconductor technology and more about controlling multiple stages of power conversion.
For AI data-center designers, that could create a more integrated approach to power delivery—from the incoming electrical infrastructure through rack-level systems and ultimately to the xPU.
It also puts Navitas into competition across several layers of an increasingly crowded AI power market. Semiconductor companies, power-management vendors and data-center infrastructure suppliers are all developing technologies aimed at supporting higher rack power and greater efficiency.
Navitas expects a larger AI infrastructure market
The acquisition also materially changes Navitas’ market opportunity.
The company expects the transaction to more than double its identified 2030 serviceable addressable market to more than $8 billion. Navitas estimates that VPD and IVR markets will contribute at least $3.5 billion of that opportunity.
The company separately identifies approximately $3.5 billion of existing SAM for GaN and high-voltage and ultra-high-voltage SiC, along with about $1 billion associated with new JFET technology.
Those figures are Navitas’ own market estimates rather than independent industry forecasts, but they illustrate the strategic rationale for the Claros acquisition: the company is seeking to participate in more stages of the power-delivery chain rather than remain concentrated in individual power-conversion components.
The rise of power-aware AI infrastructure
The acquisition comes as AI infrastructure increasingly becomes an exercise in balancing compute, power, cooling and physical density.
For AI chip designers, higher-performance processors are only useful if the surrounding electrical architecture can support them. That makes power delivery a potential constraint on the performance and scalability of future systems.
Navitas’ strategy is to address that constraint across multiple layers.
Its GaN and SiC technologies target higher-voltage power conversion, while Claros adds integrated voltage regulation close to the processor. The combination could give the company a broader role as AI data-center architectures move toward higher-voltage distribution and increasingly dense accelerator systems.
The real test will be adoption by hyperscalers, AI infrastructure providers and processor platforms. Integrated power delivery is ultimately a system-level decision, and winning that business requires more than semiconductor performance. Efficiency, thermal characteristics, reliability, packaging, software control and compatibility with evolving xPU architectures will all matter.
Still, the acquisition highlights an important shift in AI infrastructure economics.
The race to build faster AI systems is no longer only about GPUs, accelerators or networking. Power conversion and delivery are becoming part of the performance architecture itself.
By adding Claros, Navitas is betting that the companies capable of solving that power problem from the grid all the way to the xPU will have an increasingly important role in the next generation of AI data centers.
Market Landscape
AI infrastructure is moving toward higher rack densities, larger accelerator clusters and new power architectures, increasing demand for efficient power conversion at multiple stages.
Navitas’ strategy combines GaN, high-voltage and ultra-high-voltage SiC, IVR and emerging JFET technologies across the power chain. The Claros acquisition places the company closer to the processor, where transient response, efficiency and power density become particularly important.
The competitive landscape includes power-semiconductor and power-management companies such as Infineon, onsemi, Monolithic Power Systems, Vicor and others, alongside data-center infrastructure suppliers developing higher-voltage architectures.
The emerging 800V DC AI data-center architecture is particularly significant because it could shift power distribution closer to the high-voltage side while creating new requirements for downstream conversion and processor-level regulation.
Top Insights
- Navitas’ Claros acquisition adds IVR technology to its existing GaN and SiC portfolio, extending its AI infrastructure strategy directly to the xPU.
- Integrated voltage regulation moves power conversion closer to processors, targeting efficiency, transient response and power-density challenges in AI systems.
- Navitas is combining 800V DC power technologies with processor-level regulation to address multiple stages of the AI power-delivery chain.
- The company estimates its identified 2030 serviceable addressable market will exceed $8 billion following the Claros transaction.
- The deal underscores a broader shift toward treating power delivery as a core constraint on AI compute scalability and rack density.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
