MOSFET for AI servers offers industry-leading SOA performance and low ON-resistance

ROHM has released a 100V power MOSFET—RY7P250BM—optimised for hot-swap circuits in 48V power systems used in AI servers and industrial power supplies needing battery protection to the market.

As AI technology rapidly advances, data centres are facing unprecedented processing demands, and server power consumption continues to increase annually. In particular, the growing use of generative AI and high-performance GPUs has created a requirement to improve power efficiency while supporting higher currents simultaneously. To meet these challenges, the industry is shifting from 12V systems to more efficient 48V power architectures. Furthermore, in hot-swap circuits used to replace modules while servers remain powered on safely, MOSFETs are needed that offer wide SOA and low ON-resistance to protect against inrush current and overloads.

The MOSFET delivers these critical characteristics in a compact 8080-size package, helping to decrease power loss and cooling requirements in data centres while improving overall server reliability and energy efficiency. As demand for 8080-size MOSFETs increases, this new product offers a drop-in replacement for existing designs. Notably, the device achieves wide SOA (VDS=48V, Pw=1ms/10ms), ideal for hot-swap operation. Power loss and heat generation are also minimised with an industry-leading low ON-resistance of 1.86mΩ (VGS=10V, ID=50A, Tj=25C), approximately 18% lower than the typical 2.28mΩ of existing wide SOA 100V MOSFETs in the same size.

Wide SOA tolerance is crucial in hot-swap circuits, particularly those in AI servers that encounter large inrush currents. The device meets this demand, attaining 16A at 10ms and 50A at 1ms, enabling support for high-load conditions that conventional MOSFETs struggle to handle.

The company's new product has also been certified as a recommended component by a leading global cloud platform provider, where it is anticipated to gain widespread adoption in next-generation AI servers. Particularly in server applications where reliability and energy efficiency are mission-critical, the combination of wide SOA and low RDS(on) has been highly evaluated for cloud infrastructure.

Typical Application Examples include:

• 48V AI server systems and power supply hot-swap circuits in data centres
• 48V industrial equipment power systems (i.e. forklifts, power tools, robots, fan motors)
• Battery-powered industrial equipment such as AGVs (Automated Guided Vehicles)
• UPS and emergency power systems (battery backup units)