Onsemi Bets on Physical AI After Automotive, Seeking More Semiconductor Content

Onsemi announced on the 30th that it will expand its business into the physical AI market based on its automotive business. The company said it will use the power and sensing technologies accumulated in automotive as core capabilities and apply its intelligent power and sensing technologies, developed in the automotive industry, to physical AI applications as it seeks new growth opportunities.

Sudhir Gopalswamy, president of Onsemi's Analog, Mixed-Signal and Intelligent Sensing Group, said at Investor Day that the intelligent power and sensing technologies that drove growth in the automotive business are emerging as foundational technologies for next-generation industrial systems and physical AI applications.

Gopalswamy said Onsemi is expanding its core business areas. He added that the company's next phase of growth will center on increasing semiconductor content and enhancing solution value.

He also outlined a plan to apply automotive innovation technologies across new industrial systems as a next growth direction. Onsemi said it is broadening the power and sensing capabilities it built in automotive to other applications, positioning physical AI as a new growth area.

Onsemi identified electrification, autonomous driving and SDV as the main growth drivers of its automotive business. At the center of electrification, it has placed silicon carbide (SiC) technology. The company is taking an approach focused on strengthening device performance and system-level solution competitiveness.

Onsemi is pursuing a plan to improve SiC performance by about 25% with each new generation. The aim is to support next-generation 1000V battery architectures. It also plans to expand its SiC roadmap up to 1400V. Through this, the company aims to help automakers optimize range, performance and system cost in combination.

Onsemi is using Treo, a 65-nanometer analog and mixed-signal platform. Treo is responsible for integrating intelligent system solutions for power, sensing and control. Based on this, it plays a role in increasing value in vehicle power management.

Onsemi is pushing ahead with an embedded power platform (EPP) as an expansion built on Treo. EPP is designed to deliver up to four times the power density of conventional inverter designs. It also aims to reduce power loss by 15%. This is intended to support smaller, lighter and more efficient inverter systems.

In autonomous driving, the company is pushing to strengthen sensing technologies for accurate recognition across a variety of driving environments. Real-world driving conditions include darkness, fog, rain and glare, and the importance of perception stacks that combine multiple sensing technologies is growing in order to respond to these conditions.

In response, Onsemi has built a sensing portfolio spanning image, ultrasonic and inductive sensing, and is expanding its technology into indirect time-of-flight sensor (iToF) and short-wave infrared (SWIR) sensing. The company plans to improve sensing performance by up to 20 times and reduce system costs through technology integration.

Gopalswamy said autonomy starts with perception. He said autonomous machines need to accurately understand their surroundings, and that the company is expanding vision technology by improving the speed, accuracy and reliability of understanding the physical world through a range of sensing technologies.

The automotive industry is shifting from a structure centered on connecting multiple ECUs to one that uses centralized computing platforms and Ethernet-based networks. This change is cited as the backdrop for SDV creating new growth opportunities.

As this structural shift advances, demand for intelligent sensing and actuation nodes is expected to grow. Onsemi said it expects automakers to adopt new networking technologies and redesign vehicle electronic systems around software-based platforms, and based on that, it sees a multibillion-dollar business opportunity in SDV.

Onsemi's technological capabilities in automotive are expanding into physical AI. Physical AI refers to technology that interacts with the real physical world beyond AI's digital domain, and the company points to autonomous vehicles, industrial automation systems, drones, warehouse robots and humanoid robots as application areas.

For physical AI to function, intelligent machines must perceive their surroundings, respond to changing conditions, make decisions and carry out real actions. This requires sensing for understanding the physical world, efficient power conversion, and precise motor control and actuation. Gopalswamy explained that to implement intelligent systems, machines must perceive, explore and interact with the physical world.

Onsemi believes the requirements of physical AI align with the company's technological strengths accumulated in automotive. Gopalswamy said physical AI is a natural extension of autonomous vehicles.

He also said that as machines become more autonomous, semiconductor content increases. He added that sensing, power management and actuation are additional technologies required as autonomy advances, and that connectivity is also needed to enable information transfer across the system.

Onsemi is already applying its key technologies to a range of physical AI-related subsystems. The areas of application include vision systems, power conversion and motor control.

Onsemi said new platforms such as humanoid robots will require organic integration of dozens of power, sensor and control devices, and it expects semiconductor content to expand as a result, viewing this as a new business opportunity. Accordingly, the company aims to increase the revenue potential of its core technologies and secure growth opportunities from rising semiconductor content in next-generation intelligent machines by using the same power and sensing platform across the automotive, industrial and physical AI markets.

Source: TECHWORLD · Park Kyu-chan
Original: https://www.epnc.co.kr/news/articleView.html?idxno=407534

References

• https://www.epnc.co.kr/news/articleView.html?idxno=407534