Vinci4D applies physics AI to accelerate semiconductor design cycles

Semiconductor engineering faces mounting pressure as designs grow more intricate and their physical behavior becomes harder to forecast. Modern accelerators pack together multiple chips, memory and packaging in ways that create interdependent engineering problems across disciplines. Vinci4D Inc. has built a platform using physics AI to embed simulation directly into the design workflow that engineers already use.

The company's approach resolves a longstanding tension in the field, according to Dinesh Panneerselvam, head of technical solutions at Vinci. "Oftentimes, what I see in the field is engineers have to choose between accuracy or fast physics," he said. "They don't get to do both. What Vinci does is bring them both at the speed at which they're designing." Panneerselvam discussed the technology during an interview at SEMICON West with Christophe Bertrand of theCUBE Research.

Physics AI moves analysis upstream

Vinci's platform combines a physics foundation model with a GPU-native physics kernel through what the company calls Continuous Physics Reasoning. This combination delivers consistent outcomes while maintaining the precision of traditional solvers, Panneerselvam explained.

"In a product development world, the team that understands the physics — or the behavior of their product early on — they win the race," he said. "That's exactly what Vinci is providing. That means you design at your architectural level. We provide what the performance is."

The platform trains its model on Vinci's proprietary data rather than information from customers. By running behind each customer's firewall, the system keeps sensitive design details on the customer's own infrastructure, Panneerselvam noted.

"The way we have designed our platform, it works right into your workflow," he said. "You don't have to change any of your current workflows. We ingest your design files as is and start building the model. There is so much automation built in that there's very little manual input."

Extending physics AI beyond chips

Physics AI applications extend beyond semiconductors into physical AI systems such as robots and humanoid devices. Vinci's simulation capabilities could enable engineers to predict how such products will perform before settling on a final design, according to Panneerselvam.

"Customers are able to open doors for new innovation — things that they haven't imagined before," he said. "For example, you have an engineering intuition for a product, and it stops there. What Vinci and products like Vinci allow is basically [to] explore beyond that. We call it the art of the impossible, and you can move into that space."

Vinci is pursuing opportunities in automotive and aerospace engineering as it moves beyond its semiconductor roots. The company's $250 million Series B funding round will fuel expansion into additional physics domains and industries, with planned additions in electromagnetics and fluid dynamics, Panneerselvam stated.

https://www.youtube.com/embed/lDwtJsC7O-M?feature=oembed

"The bottom line is physics doesn't change," he said. "It's just the applications that we have to build. The physics remains the same."