News

Rising AI Data Center Power Demands Make Efficient Power Devices Essential

Publication Time:2024/4/18 14:26:18 Reading Volume: Source: HERTZPOWER SEMICONDUCTOR

The rapid development of generative artificial intelligence and high-performance computing is driving continuous growth in data center computing demand. As server performance increases, the power consumption of individual devices and racks is also rising, placing greater pressure on the efficiency and thermal management of data center power systems.

High-Power Racks Create New Power Supply Challenges

Traditional data center power systems require multiple stages of energy conversion to transform grid power into the operating voltages required by server chips.

Each conversion stage generates energy loss and heat. As server power rises rapidly, even a one-percentage-point difference in power efficiency can lead to significant differences in electricity consumption and cooling costs.

Reducing the number of conversion stages, increasing supply voltage, and adopting highly efficient power devices are therefore becoming important development priorities for data center power systems.

SiC and GaN Expand Application Opportunities

SiC and GaN devices offer fast switching speeds, low losses, and strong high-frequency performance. These advantages can help power systems operate at higher frequencies while reducing the size of transformers, inductors, and cooling components.

SiC devices are well suited to high-voltage, high-power front-end energy conversion and can be used in power factor correction, high-voltage DC conversion, and backup power systems.

GaN devices are suitable for high-frequency, high-power-density medium- and low-voltage conversion stages and can be applied in server power supplies, telecommunications power supplies, and board-level power modules.

Silicon-based MOSFETs still offer mature technology, a broad product range, and controllable costs, and will continue to be widely used throughout data center power systems.

Power Devices Affect Overall Computing Efficiency

In artificial intelligence infrastructure, computing chips determine computing performance, while the power system determines whether that performance can operate stably and efficiently.

As server power continues to rise, power devices will become increasingly important. Switching losses, on-resistance, thermal performance, and package parasitic parameters all affect power efficiency and system stability.

In the future, data center power architectures will evolve toward higher voltage, greater efficiency, increased power density, and modular design. Power semiconductors will gradually move beyond their traditional role as supporting components and become key foundational devices influencing data center energy efficiency.


Home

Home

Products

Products

Applications

Applications

Contact Us

Contact Us

中文中文