GaN goes mainstream

Daniel Murphy, Senior Director of Product Management at Cambridge GaN Devices, discusses CGD’s collaboration with NXP Semiconductors to accelerate time to market in the data centre and automotive sectors.

The race to build the next generation of AI infrastructure is reshaping the power electronics industry. As hyperscale data centres push toward ever-higher power densities and electric vehicle manufacturers search for greater efficiency, compactness and performance, the limitations of traditional semiconductor technologies are becoming increasingly apparent.

Gallium nitride (GaN) has long been viewed as one of the industry’s most promising technologies, offering switching speeds and efficiency gains that conventional silicon and, in some cases, silicon carbide struggle to match. However, unlocking the full benefits of GaN requires more than advanced transistor technology alone. It demands complete system-level solutions that combine power devices, control electronics, software intelligence and robust design methodologies.

That thinking sits at the heart of a newly announced long-term collaboration between Cambridge GaN Devices (CGD) and NXP Semiconductors.

The partnership brings together CGD’s expertise in gallium nitride power devices and NXP’s extensive capabilities in processing, control, analog technologies and automotive-grade electronics. The objective is ambitious: accelerate the adoption of GaN-based solutions in some of the world’s fastest-growing markets, particularly AI-driven data centres and next-generation automotive platforms.

For Daniel Murphy, Senior Director of Product Management at Cambridge GaN Devices, the collaboration represents a natural alignment of complementary strengths.

A partnership built around complementary expertise
The power semiconductor industry is increasingly recognising that next-generation power architectures cannot be solved by individual device innovation alone. While semiconductor breakthroughs remain essential, end customers are increasingly looking for complete, validated solutions that reduce development risk and shorten deployment times.

Murphy sees the CGD-NXP relationship as a powerful example of that trend.

From CGD’s perspective, the company contributes deep expertise in gallium nitride technology, a field in which it has established itself as a significant innovator. As demand grows for higher-frequency, more efficient power conversion systems, GaN is emerging as a critical enabling technology for future AI infrastructure.

At the same time, NXP brings strengths that extend beyond power semiconductors. The company possesses extensive experience in microprocessors, DSPs, control systems, analog technologies and automotive electronics, along with a substantial global customer footprint.

The result is what Murphy describes as a highly complementary partnership. Rather than focusing solely on individual components, both companies can collaborate on complete power architectures that combine advanced switching devices with sophisticated control and management capabilities. For customers, that means faster implementation and reduced engineering complexity. For the partners, it creates opportunities to address larger portions of the value chain while accelerating adoption of their respective technologies.

Addressing the time-to-market challenge
One of the central themes of the collaboration is reducing time to market. This has become particularly important in AI infrastructure, where demand is evolving at unprecedented speed and power requirements are increasing almost continuously.

According to Murphy, discussions with hyperscalers have revealed that some existing power semiconductor technologies are struggling to keep pace with emerging requirements. Switching frequency represents one of the most significant challenges.

Certain segments of the AI data centre ecosystem are already pursuing operation around the 1MHz level, creating performance requirements that Murphy believes exceed the practical limits of silicon carbide in some power conversion stages.

This is where GaN becomes especially attractive.

Its ability to operate efficiently at significantly higher switching frequencies allows designers to create smaller, lighter and more power-dense systems.

However, high-performance GaN devices require equally sophisticated control systems. Higher switching speeds place increased demands on gate drive technologies, digital control platforms and overall system design. The collaboration therefore addresses both sides of the equation.

CGD contributes the GaN switching technology required to achieve these higher frequencies, while NXP provides the processing and control expertise necessary to manage and optimise device operation. Working together allows both companies to accelerate customer adoption and reduce barriers associated with implementing advanced power architectures.

Why AI data centres need GaN
Few markets are creating as much excitement across the power electronics industry as AI data centres. The relentless growth of artificial intelligence workloads is driving dramatic increases in rack power, system density and

overall facility energy consumption. These developments are forcing operators to rethink virtually every aspect of power delivery.

For Murphy, one of GaN’s greatest advantages lies in its ability to support higher switching frequencies without introducing unacceptable efficiency penalties. As power converters switch faster, designers can reduce the size of passive components such as magnetics and capacitors. Smaller passive components translate directly into more compact power supplies. In AI facilities, where physical space is increasingly valuable, the benefits can be significant.

“The demand is to increase switching frequency so that you can reduce system size,” Murphy explains.

Higher power density means more power supplies can be deployed within the same footprint, ultimately enabling greater compute density and processing capability.

According to CGD, silicon carbide performs exceptionally well across many applications but begins to encounter practical challenges as switching frequencies approach and exceed roughly 500 kHz in certain use cases. GaN, by contrast, can support operation at frequencies approaching 1 MHz. That distinction is becoming increasingly important as AI infrastructure evolves.

The industry’s pursuit of higher efficiency and greater power density is making faster switching not merely desirable but necessary.

Solving the energy efficiency equation
The energy demands associated with AI infrastructure have become one of the defining challenges facing the technology industry. With data centre power consumption projected to rise substantially over the coming years, even small efficiency improvements can translate into significant reductions in energy usage, cooling requirements and operational costs. Murphy argues that GaN will play a crucial role in addressing those challenges.

While higher switching frequencies enable converter miniaturisation, they also create the potential for increased switching losses if unsuitable semiconductor technologies are used. Attempting to achieve 1 MHz operation with less capable devices could result in efficiency penalties that outweigh the benefits of smaller form factors. GaN changes that equation.

The technology allows designers to push switching frequencies higher while maintaining efficiency levels that remain commercially attractive. As a result, data centre designers can simultaneously pursue greater power density and improved energy performance.

Importantly, Murphy does not frame the future as an either-or battle between semiconductor technologies. Instead, he sees complementary roles for silicon, silicon carbide and GaN. Each technology occupies a particular position within the evolving power conversion landscape. However, when it comes to the highest-frequency applications emerging within AI infrastructure, GaN increasingly appears positioned as the preferred solution.

The differentiator: CGD’s ICeGaN technology
While enthusiasm around gallium nitride continues to grow, competition within the GaN market is also intensifying. Manufacturers are pursuing a range of approaches to improve performance, reliability and ease of implementation.

For CGD, one of the key differentiators lies in its proprietary ICeGaN architecture and iDriver technology. Murphy highlights the company’s integrated iScan interface as a particularly important innovation. The technology incorporates a Miller clamp that enables extremely rapid turn-off performance while supporting unipolar operation. This means designers do not require negative gate-drive voltages during turn-off, simplifying system implementation and improving overall efficiency. The approach also helps minimise losses associated with third-quadrant operation that can occur in more conventional GaN architectures.

For engineers, these improvements can translate into both higher efficiency and simplified system design. In practical terms, removing complexity from GaN implementation remains an important objective. While the technology’s performance advantages are well understood, ease of adoption often determines how quickly new technologies move into mainstream applications. Solutions that simplify integration naturally accelerate market adoption.

Building complete power conversion platforms
Perhaps the most significant aspect of the CGD-NXP collaboration is its emphasis on system-level design.

Historically, semiconductor vendors often focused primarily on individual devices. Today’s customers increasingly expect complete solutions. That expectation is particularly evident within AI infrastructure.

Rather than supplying only power switches, CGD and NXP are working together on reference platforms that combine advanced GaN devices with processing, control and management technologies.

One example involves the development of high-voltage DC power bricks for next-generation data centre applications. These platforms combine current-generation CGD devices with upcoming products still under development, providing customers access to the latest semiconductor capabilities. At the same time, NXP contributes expertise in control architectures and processing technologies. The goal is to create validated reference designs that can be benchmarked, demonstrated and deployed more rapidly than individual component-based approaches.

This solution-oriented strategy offers several advantages.

Customers can evaluate complete architectures rather than assembling systems from disparate technologies. Development cycles can be shortened. Performance can be optimised across the entire platform rather than at the individual component level. As AI infrastructure continues evolving, such integrated approaches are likely to become increasingly influential.

Automotive: The next major opportunity
Although AI data centres are commanding enormous attention, Murphy is equally enthusiastic about the automotive sector.

The industry is currently undergoing a profound transition as manufacturers seek higher efficiency, improved performance and greater functional integration. Among the most promising opportunities is the evolution of onboard chargers.

Many automakers are exploring single-stage architectures capable of simplifying charging systems while improving efficiency. These topologies require bidirectional switching capabilities. According to Murphy, this is an area where GaN possesses a distinct advantage.

Unlike traditional silicon devices, GaN lacks a conventional body diode, allowing current to flow effectively in both directions. That characteristic makes it particularly attractive for future charger architectures. The benefits extend beyond charging systems. Traction inverter applications are also attracting growing interest.

While silicon carbide currently dominates many high-voltage EV platforms, Murphy sees significant opportunities for GaN, particularly in vehicles operating on 400 V architectures. Increasing switching frequency enables cleaner motor drive waveforms and can reduce copper and iron losses within the motor itself. The result is not merely improved inverter efficiency but enhancements across the broader propulsion system. Such system-level gains are becoming increasingly important as automakers search for competitive advantages.

Reliability remains essential
Regardless of performance improvements, power electronics engineers consistently prioritise reliability. Whether the application involves a hyperscale data centre or a passenger vehicle, robust operation remains non-negotiable. Murphy believes the CGD-NXP collaboration strengthens both companies’ ability to address this requirement. NXP contributes extensive automotive credibility, particularly through its longstanding relationships with OEMs and Tier 1 suppliers. Its reputation in microcontrollers, gate drivers and automotive-grade electronics provides confidence for customers considering next-generation power architectures.

The companies are already collaborating on projects involving major automotive manufacturers. One example involves a GaN traction inverter programme with a leading German automotive OEM, utilising CGD’s advanced 9mΩ device technology alongside NXP gate-drive solutions. These projects provide valuable opportunities to validate GaN technology under demanding real-world conditions.

At the device level, CGD has also focused heavily on robustness. Murphy points to work conducted with Virginia Tech that demonstrated the resilience of the iScan architecture. Conventional GaN HEMTs often tolerate transient voltages only within limited ranges. Testing has shown that CGD’s technology can withstand significantly higher voltage stress than many standard devices. Those results provide additional confidence as engineers evaluate GaN for mission-critical applications.

The industry’s growing confidence in GaN
Perhaps one of Murphy’s most notable observations concerns the changing perception of GaN within the automotive industry. Only a few years ago, many vehicle manufacturers viewed GaN largely as an exploratory technology. Interest existed, but commercial deployment remained distant. That mindset has shifted dramatically. Over the past 18 months, CGD has observed increasing engagement from OEMs, Tier 1 suppliers and engineering teams evaluating GaN for future vehicle platforms.

Several factors are driving the change.

One is the expectation that GaN costs will continue to improve as production volumes increase and manufacturing ecosystems mature. Another is growing awareness of the broader system-level benefits offered by the technology. GaN’s lateral structure enables the integration of additional functions directly onto the semiconductor die. Features such as temperature monitoring and current sensing become easier to incorporate within future designs. These capabilities align well with industry trends toward smarter and more integrated power electronics.

As a result, discussions have evolved significantly. Where engineers once asked whether GaN might someday become useful, they are increasingly investigating how and when it should be deployed.

Looking ahead
The roadmap for the CGD-NXP collaboration is already taking shape. New products currently under development are expected to enter sampling programmes shortly, providing the foundation for a new generation of demonstrator systems and reference designs. For data centres, these platforms will be targeted at hyperscalers and power supply manufacturers seeking solutions capable of meeting future AI infrastructure requirements. For automotive customers, work continues across traction inverters, onboard chargers and advanced power modules. Murphy expects both sectors to deliver significant opportunities over the next several years.

What makes the partnership particularly compelling is its timing.

The AI revolution is increasing demands on power delivery systems at an unprecedented rate, while the automotive industry continues its shift toward more efficient, software-defined electric vehicles. Both trends require power electronics technologies capable of delivering greater performance without sacrificing reliability. By combining CGD’s gallium nitride expertise with NXP’s system-level capabilities, the companies are positioning themselves at the intersection of those two powerful market forces.

For the wider industry, the collaboration reflects a broader reality: the future of power conversion will not be determined by better transistors alone. It will be defined by complete ecosystems that unite advanced semiconductors, intelligent control, robust system architectures and scalable deployment strategies.

As AI infrastructure expands and vehicle electrification accelerates, GaN appears increasingly poised to move from emerging technology to mainstream solution. Partnerships such as the one between Cambridge GaN Devices and NXP may ultimately determine how quickly that transition occurs.