NEW ALLOYS FOR POWER ELECTRONICS - Inductors for the AI age

Rapid advancements in artificial intelligence (AI) are causing a significant increase in the energy demands of state-of-the-art data centers and AI systems. AI servers now consume up to three times more energy than conventional systems. While the energy requirements of data centers are rising exponentially and already account for over 2% of global power consumption. The total power demands of some hyperscale data centers exceed > 20 MW. Today, high-performance components such as CPUs, GPUs and FPGAs require currents of well over 1,000 A at DC link voltage levels, while compact sizes and maximum energy efficiency are always expected. In this regard, conventional magnetic materials are being gradually pushed to their limits. Compact, highly efficient inductors that can easily deal with extreme currents and high switching frequencies are the main focus of ongoing technical developments.

The demand for higher processing power requires electronic components with special properties. However, higher performance also leads to increased parasitic losses in the components, thereby necessitating capacitive and magnetic components with extremely low losses. At the same time, these components must offer extensive reliability under harsh conditions. Since a large number of components are required to support AI processors, miniaturization of the components is also essential. However, components with extremely low losses are also in demand in high-performance applications in industrial environments or in the field of rail technology. Voltage levels such as 48 V, 400 V and 800 V are usually present in the said sectors.

High saturation magnetic flux density (B_S) and low coercive field strength (H_C) are required for the efficient performance of inductors in electronic devices. A high B_S value enables the inductor to cope with stronger magnetic fields without reaching saturation. This allows it to transmit higher currents and store greater energy. This is particularly important for applications that demand compact designs coupled with high performance, such as AI systems. A low H_C value means that the material requires less energy for magnetization and demagnetization. This helps reduce energy losses and improve overall efficiency. Conventional magnetic core materials, such as Mn/Ni-Zn ferrites or Fe-Si alloys, have long been applied for these applications but often fail to achieve the right balance between a high B_S value and a low H_C value.

Formability, integration and AC losses in high-frequency operation

The combination of high saturation flux density and low core losses has long been the goal for the developers of magnetic materials. Particular focus is placed on nanocrystalline Fe-B-P-Cu alloys (Fig. 1). 

The new material Nanomet is built on an Fe-based alloy. The Fe-B-P-Cu alloy produces a fine, spherical powder when rapidly quenched. Further, the alloy can be transformed into complex 3D shapes using the hot-press process. This minimizes stray fields and enables optimal adaptation to the printed circuit board layout. The spherical morphology of the powder particles promotes dense and uniform packing. The resulting cores offer uniform magnetic field distribution, reduced hot spots and optimized magnetic flux guidance. After targeted annealing, a nanocrystalline structure with optimized magnetic properties forms.

The result is magnetic cores with a saturation flux density of 1.55 T combined with a coercive field strength of just 23 A/m (Tab. 1). This combination of materials was not previously achievable with conventional ferrites or Fe-Si alloys. For comparison: Fe-Si-Cr-based powder core materials only achieve B_S = 1.28 T and H_C = 720 A/m.

The potential of the material is also clearly evident in high-frequency conditions. At 1 MHz, the AC losses of the new cores are up to 4.5 times lower than those of conventional metallic composite cores and only roughly 50 mW higher than traditional ferrites. Moreover, the alloy exhibits “soft saturating” behavior, which additionally boosts energy efficiency. 

Table 1: Comparison of magnetic parameters (source: Yageo Group)

Application of a buck converter

A typical voltage conversion example shows how the aforesaid material properties have a specific impact when applied: A buck converter that converts an input voltage of 6 V to 1.8 V at a clock frequency of 1 MHz (Fig. 2). This configuration utilizes a 70 nH inductor based on the new Fe-B-P-Cu alloy. It achieves magnetic saturation above 160 A, even at elevated temperatures. For comparison: Under exactly the same conditions, an identically sized ferrite core exhibits saturation behavior at around 50 A (at 100 °C).

The new alloy thus enables 40% higher energy buffering in an identical installation space. The formula for the calculation is:

E = 0.5 * L * I²; where E = energy (J), L = inductance (H) and I = current (A)

This property enables the use of smaller coils for the same power or, given a certain volume, the power density to be further enhanced.

In addition, the inductor exhibits stable behavior across a wide temperature range, offering a decisive advantage in thermally demanding applications such as GPU power supplies and other high-performance power supplies. The improved current carrying capacity results in fewer parallel circuits and thus a simplified layout, reduced component costs and potentially lower system costs.

Summary

The combination of high magnetic performance, low power loss and thermomechanical formability makes the new nanocrystalline Fe alloy an ideal material for next-generation AI power electronics. This can help achieve a considerable increase in efficiency, particularly for CPU/GPU power supply applications and high-performance power supplies, and make a decisive contribution to slashing the CO₂ footprint.

 

For more information and a direct ordering option, please visit our e-commerce platform at www.rutronik24.com.

Subscribe to our newsletter and stay updated.