Summary
- Infineon and Enphase have completed initial testing of 900V monolithic gallium nitride bidirectional switches.
- The devices are being integrated into Enphase's developing solid-state transformer, with production qualification still outstanding.
- A separate Enphase laboratory demonstration converted 13.8kV AC to 800V DC using 132 power modules; commercial deployments have not begun.
German semiconductor manufacturer Infineon Technologies and US energy technology company Enphase have completed initial testing of 900V gallium nitride power switches intended for solid-state transformers serving AI data centres.
The companies are integrating the new monolithic bidirectional devices into development builds of Enphase’s IQ Solid-State Transformer, a power conversion system designed to supply high-density computing equipment. Production qualification remains ahead, and neither company has confirmed commercial deployment of the new components.
The development extends Infineon’s gallium nitride switching technology beyond the 650V class already used in Enphase products. The higher voltage rating is intended to simplify converter designs and reduce the number of semiconductor components needed in selected electrical architectures.
Infineon and Enphase describe the devices as the first 900V monolithic gallium nitride bidirectional switches of their kind. That industry-first claim comes from the manufacturers, while their confirmed technical milestone is successful initial testing followed by integration into development hardware.
AI computing changes electrical requirements
Data centres contain extensive electrical infrastructure between the incoming power supply and the processors performing computing tasks. Electricity must be transformed, converted and distributed at voltages suitable for the equipment being operated.
Conventional arrangements frequently involve several conversion stages, with transformers and power electronics performing different functions. Every stage occupies space, requires supporting components and introduces some energy loss.
High-density AI equipment places greater pressure on those arrangements because substantial computing loads must receive stable power despite rapid changes in demand. Power delivery also influences cooling requirements, equipment layout and the amount of electrical capacity available for processors.
Solid-state transformers offer an alternative architecture combining semiconductor switching with high-frequency magnetic components. They can perform functions associated with voltage transformation and power conversion within a more integrated system.
The potential benefit lies in delivering electricity through fewer stages while maintaining control over voltage and current. Whether the approach improves efficiency and cost depends on the complete design, including protection equipment, cooling, insulation and performance under different operating conditions.
For operators, reduced conversion losses could lower both electricity consumption and the heat that must be removed from power equipment. The magnitude of any saving requires measured system performance rather than an assumption based on the characteristics of an individual semiconductor.
Bidirectional devices simplify converter designs
Gallium nitride is a wide-bandgap semiconductor material used in power electronics where high-frequency switching and reduced electrical losses can offer advantages over conventional silicon devices.
Infineon’s new device combines bidirectional voltage blocking and current conduction within a single component. In suitable circuit designs, that arrangement can replace combinations of separate switches performing related functions.
Reducing component counts can simplify connections and control arrangements, while higher switching frequencies can allow engineers to reduce the size of associated magnetic components. These characteristics are particularly relevant where electrical equipment must occupy limited space.
Moving from 650V to 900V also increases the voltage range available to designers, potentially allowing some conversion architectures to use fewer modules connected in series.
Those benefits depend on circuit requirements rather than the voltage rating alone. Engineers must account for switching behaviour, electromagnetic interference, heat dissipation and the ability of connected components to operate reliably under changing loads.
Enphase has already deployed gallium nitride technology in its IQ9 microinverters, providing operational experience with the material. The new 900V devices will nevertheless require separate qualification because they are intended for a different electrical environment and power conversion architecture.
Enphase demonstrates a separate transformer milestone
Alongside the semiconductor development, Enphase has demonstrated its IQ Solid-State Transformer converting a 13.8kV alternating-current input into a regulated 800V direct-current output at its research laboratory in Austin, Texas.
The demonstration used 132 power modules, arranged with 44 connected in series across each of three phases. Distributed controls maintained voltage sharing between the modules while the system responded to simulated changes in AI computing demand.
Each module was rated at 5.5kW and incorporated gallium nitride switching technology, Enphase’s fifth-generation Kestrel application-specific integrated circuit and a proprietary high-frequency transformer.
The successful test establishes progress in system-level power conversion, but Enphase has not stated that the newly tested Infineon 900V switches were incorporated into that particular demonstration.
The two announcements therefore describe related but separate milestones: development of a higher-voltage semiconductor component and a laboratory demonstration of the broader transformer architecture.
Enphase plans further testing at 34.5kV, followed by demonstrations of complete systems operating at megawatt scale. It aims to conduct customer pilots during 2027, subject to development and qualification progress.
These targets should be distinguished from commercial deployment. The company has demonstrated a particular electrical conversion arrangement, while sustained operation, reliability and performance under customer conditions remain to be established.
Higher-voltage distribution creates new opportunities
The transformer programme is intended to support 800V DC distribution within AI computing facilities, with longer-term development also addressing architectures operating at higher voltages.
Increasing distribution voltage can reduce the current required to deliver a given amount of power. In appropriate systems, that can reduce resistive losses and influence conductor sizing, although higher-voltage operation introduces additional requirements for electrical protection, insulation and maintenance.
Enphase’s proposed architecture also uses higher-voltage unidirectional switches on the DC side. Those devices perform different functions from the 900V bidirectional components and should not be treated as interchangeable technologies.
Infineon and Enphase also identify commercial solar inverters as a potential application, including three-phase equipment designed for 400V line-to-line systems in European markets.
This gives the technology a possible market beyond AI infrastructure, although commercial adoption will depend on cost, electrical performance, qualification and the suitability of the devices for individual converter designs.
Silicon carbide and conventional silicon technologies will continue to offer different advantages across power electronics applications. Higher-voltage gallium nitride expands the available engineering options rather than establishing that a single semiconductor material will replace the others.
The companies have completed initial testing of the 900V devices and begun integrating them into transformer development builds. Production qualification, larger system demonstrations and customer pilots will provide the next evidence of whether the component-level advances can translate into reliable commercial power infrastructure.












