A data center with a scale of hundreds of billions, how can inductors improve power efficiency?

1. AI Data Center Market and Power Supply Challenges

In recent years, the explosive growth in the demand for training AI large models has driven the continuous

expansion of the AI data center market. According to the "2025 China Artificial Intelligence Computing Power

Development Evaluation Report", the global market size of AI servers reached 125.1 billion US dollars in 2024,

and is expected to rise to 158.7 billion US dollars in 2025, and is expected to exceed 222.7 billion US dollars in 2028.

Among them, the proportion of generative AI servers will increase from 29.6% in 2025 to 37.7% in 2028.

However, the rapid expansion of data center scale has also brought new challenges, especially the power supply issue.

As AI applications such as large models continuously increase their demand for computing power, the computing power

of AI chips is also constantly rising, but this is accompanied by a significant increase in power consumption. Nowadays,

the power consumption of a single computing power chip has soared from around 300W in the past to 1000W, which poses

higher requirements for the power supply of data centers.

2. Upgrade Requirements for the Power Supply Unit

On one hand, data centers require more efficient power supply units (PSUs) to meet the increasing power consumption demands;

on the other hand, computing cards such as GPUs also need more powerful power supply designs to ensure their stable operation.

To enhance the energy efficiency of PSUs, in addition to using wide bandgap semiconductor power devices like SiC (silicon carbide)

and GaN (gallium nitride), inductors, as a key component for improving power supply efficiency and load capacity, also play a crucial role.

3. Definition and Working Principle of Inductance

An inductor is a passive component. Its core function is to store magnetic energy and exert an inhibitory effect on the current changes in the

circuit through the phenomenon of electromagnetic induction. Structurally, an inductor is a coil formed by winding insulated wires, usually

wrapped with magnetic materials such as ferrite or iron powder cores. Its shapes are diverse, including ring-shaped, cylindrical, and sheet-like.

In the power supply of AI servers, currently, large current inductors with flat coils are the mainstream.

Its working principle is based on Faraday's law of electromagnetic induction and Lenz's law. When current passes through an inductor coil,

a magnetic field is generated. At this time, electrical energy is converted into magnetic energy and stored; when the current changes, the change

in the magnetic field will generate an induced electromotive force (i.e., reverse voltage), and at the same time release electrical energy. Therefore,

when there is a sudden change in the current, the inductor can play a buffering role and stabilize the voltage.

4. Application of Inductors in AI Servers

Applications in GPU

In the power supply circuit of the GPU computing card, 12V voltage needs to be converted to 1V for power supply. Due to the low voltage and high power,

the current can be as high as several hundred amperes. When the GPU computing card starts up, the current demand surges, which may cause a sudden drop

in the power supply voltage. At this time, the inductor can release the stored energy to avoid the sudden drop in voltage, thereby ensuring the stable operation of

the GPU. Additionally, in a high-frequency environment, the inductor also needs to maintain low heat generation to prevent performance degradation due to temperature rise.

2. Application of the power supply section

Apart from being used in GPUs, inductors are also widely employed in the power section of data centers, such as in the DC-DC conversion part of the PSU. By storing and releasing energy,

inductors can stabilize the voltage and ensure the stable operation of devices like servers and GPUs.

3. Applications in Optical Modules

In optical modules, capacitance can filter out electromagnetic interference (EMI) in high-frequency circuits, improving the quality of signal transmission. This effect is particularly crucial in

high-speed optical modules.

5. The Development Direction of Inductance

With the continuous development of data centers, inductors have also shown three main development directions:

High power and high temperature resistance

With the continuous improvement of the computing power hardware power in AI data centers, the inductors need to handle greater power within

the limited space of the server racks and must possess higher heat resistance capabilities to ensure stability in high-temperature environments.

2. High Frequency and Low Loss

In data centers, wide bandgap semiconductor devices such as GaN and SiC are increasingly being used in power supplies. These high-frequency devices

require inductors to support higher operating frequencies and reduce core losses, in order to improve the overall system's conversion efficiency.

3. Miniaturization and Integration

In AI data centers, servers and AI acceleration cards tend to integrate more computing units within a limited space. Therefore, components including

inductors need to be further miniaturized. Currently, the size of surface mount inductors used in servers has been reduced to the millimeter level.

At the same time, in order to increase power density while reducing volume, integration has become an inevitable trend. For example, DrMOS integrates components

such as inductors, capacitors, and MOSFETs into a power module package.

As the computing density of data centers continues to increase, inductors will further evolve towards integration (embedded in the substrate) and

functional integration (co-design with ICs), becoming the core foundation for supporting the next generation of high-efficiency electronic devices.