Princeton University has been selected to lead a national initiative focused on developing advanced semiconductors for wireless communication and remote sensing, with an emphasis on using artificial intelligence (AI) to automate chip design. The project, funded by a nearly $10 million grant from the National Semiconductor Technology Center, aims to improve the efficiency and capabilities of chips that are vital for technologies such as next-generation wireless networks, satellite communications, autonomous vehicles, and smart health care systems.
Kaushik Sengupta, professor of electrical and computer engineering at Princeton, will direct the effort. His team is targeting the automation of microchip design for radio-frequency (RF) wireless communication—technology essential for connecting electronic devices both to each other and to their environments.
“Embracing AI for radio frequency design is paramount for maintaining the United States’ leadership in technological innovation,” said Deirdre Hanford, CEO of Natcast. “Leveraging AI not only accelerates our research capabilities but also ensures the U.S. remains at the cutting edge of communication infrastructure.”
Sengupta explained that designing specialized wireless chips currently requires significant manual labor and expert knowledge. “They are fundamentally handcrafted,” he said. “But if you could get to a point where the manual labor-intensive aspects of design can be automated out and you can start discovering new architectures or new functionality, there lies a window of opportunity.”
Unlike chips used in computers and data centers—which benefit from high levels of automated design—wireless chips face complex challenges due to overlapping forces and unpredictable environments. This complexity means that each stage requires input from multiple experts across different fields. According to Sengupta, this leads to longer development times and higher costs while limiting innovative solutions: “You’re sort of limited to the human imagination,” he said. “It’s a very bottom-up approach.”
Sengupta’s group is reversing this process by using AI-driven methods that begin with user demands before working backward toward optimal circuit designs. Their work has resulted in architectures that sometimes defy traditional expectations but outperform conventional chips. Graduate students Emir Ali Karahan and Zheng Liu presented these results at the 2022 IEEE International Microwave Symposium, earning top honors at the event—a milestone Sengupta credits with helping establish Princeton’s leadership in this area. The team later received further recognition with a Best Paper Award from IEEE Journal of Solid State Circuits in 2023.
Mengdi Wang, another Princeton professor involved in electrical engineering as well as statistics and machine learning, will support research into AI methods for chip automation. She noted two primary techniques behind their efforts: reinforcement learning—well known for excelling at strategy games like Go—and RFdiffusion models used in groundbreaking protein chemistry research.
The project includes collaboration with researchers from University of Southern California, Drexel University, Northeastern University; industry partners RTX, Keysight Technologies, Cadence; as well as an advisory board comprising leaders from Qualcomm, Skyworks Solutions Inc., Texas Instruments Incorporated (https://www.ti.com/), Nokia Bell Labs (https://www.bell-labs.com/), Ericsson (https://www.ericsson.com/en), and Maury Microwave.
The goal is to develop tools that lower costs associated with chip development while encouraging creativity within the field—a move expected to increase competition among semiconductor designers nationwide. The Princeton-led team was one of three groups chosen by Natcast during this round; others will be led by Keysight Technologies (https://www.keysight.com/us/en/home.html) and The University of Texas at Austin (https://www.texasadvancecomputing.org/).
