Full Breakdown
AI-Driven Design Accelerates Wireless Chip Innovation but Faces Funding Hurdles
By Drooid · · How we work
AI Inverse Design Breakthroughs
Researchers at Princeton University and IIT Madras have introduced an “inverse design” workflow that starts from desired electromagnetic performance and lets a convolutional-neural-network model generate circuit geometries. The approach has produced compact antennas operating on two distinct frequencies, multi-port RF filters and power dividers that outperform conventional, template-based designs. By pairing the neural predictions with optimization algorithms, the team reduced design time by at least a factor of ten compared with earlier AI-assisted methods, with some structures converging in seconds. The resulting layouts are often unintuitive—appearing randomly shaped—yet they deliver performance gains that human engineers had not anticipated.
Funding Setback and Program Uncertainty
In May 2025 the National Semiconductor Technology Center awarded nearly $10 million to a Princeton-led effort to develop AI-powered tools for wireless chips, with Kaushik Sengupta directing the work. A few months later the project received a stop-work order, and in August 2025 the U.S. Department of Commerce withdrew support from Natcast, the organization overseeing the programme, citing a dispute over its establishment. An August 2026 government audit confirmed Natcast’s cessation of operations in September 2025 and noted substantial cancellations across the federal semiconductor research portfolio. The current status of the original Princeton award remains unclear, though the research team has continued publishing results.
Risks, Transparency, and the Human Role
The AI-generated designs function according to electromagnetic laws, but their internal logic often remains opaque. Engineers cannot readily explain why a particular metal trace arrangement yields a given response, raising concerns about reliability under manufacturing variations, temperature shifts, or unforeseen operating conditions. The “black-box” nature could lead to hidden failures in critical systems such as medical devices or autonomous vehicles. Researchers stress that AI should augment—not replace—human designers, who must validate, troubleshoot, and decide which AI proposals merit fabrication.
Data & Statistics
- Funding: ~$10 million award (May 2025).
- Design speed: AI workflow cuts design time by >= 10×; some designs converge in seconds.
- Performance gains: Novel antenna and filter structures show measurable improvements over classical designs, though long-term commercial reliability has not yet been demonstrated.
Verbatim Quotes
- “Humans cannot understand them, but they can work better,” — Kaushik Sengupta, the lead researcher, a professor of electrical and computer engineering at Princeton
- “We are coming up with structures that are complex and look randomly shaped, and when connected with circuits, they create previously unachievable performance,” — Sengupta. The de
- “Classical designs, carefully, put these circuits and electromagnetic elements together, piece by piece, so that the signal flows in the way we want it to flow in the chip. By changing those structures, we incorporate new properties,” — Kaushik Sengupta, the lead researcher, a professor of electrical and computer engineering at Princeton
- “There are pitfalls that still require human designers to correct,” — Kaushik Sengupta, the lead researcher, a professor of electrical and computer engineering at Princeton