U.S. SolarEdge to Deploy Solid-State Transformer Solutions by 2027 to Address High Power Density Challenges in AI Data Centers

en.Wedoany.com Report on Mar 27th, To support the increasing complexity of AI models, data centers are facing the challenge of rising power density. Companies like Nvidia are driving the development of computer racks towards higher densities. While most current data center racks operate at around 10 kilowatts, next-generation AI data centers may reach 100 kilowatts, with future demands potentially requiring about 1 megawatt per rack. Traditional alternating current (AC) infrastructure cannot meet this demand, prompting the industry to shift towards direct current (DC) systems, particularly 800-volt DC, to support next-generation computing.

Solid-state transformers are a key solution, integrating transformer and inverter functions and utilizing solid-state components for higher efficiency. For example, SolarEdge is collaborating with Infineon Technologies, leveraging silicon carbide chip technology to make solid-state transformers more compact and efficient. In data center environments, efficiency optimization is directly linked to computing power and operational costs, and the application of solid-state transformers helps improve energy utilization.

The design of solid-state transformers supports medium-voltage connections up to 34.5 kilovolts, enhancing system flexibility. Their manufacturing process draws on inverter technology, employing modular sub-components to facilitate large-scale production and rapid delivery. SolarEdge plans to deploy these solutions by 2027 to support next-generation AI equipment requiring 800-volt DC.

The shift to DC systems also facilitates the integration of solar power and energy storage. When combined with uninterruptible power supplies in data centers, this could enable more efficient energy management. This progress is seen as a natural evolution to unlock the advantages of DC power, promising to simplify data center infrastructure and enhance overall performance.