Vinci raises $250M to build a physics foundation model for semiconductor and hardware engineering
Key Points
- Vinci raised $250M from AMD, Applied Materials, and Coastal Ventures to deploy a physics foundation model that automates thermal, mechanical, and electromagnetic simulation for semiconductor design at scale.
- The model trains on universal laws of physics rather than proprietary design files, enabling on-premises deployment inside chipmaker fabs without access to confidential designs or forward-deployed engineers.
- Vinci's go-to-market is strictly top-down, targeting VP-level conversations at major chipmakers anchored on time to market and performance metrics that directly drive production decisions.
Summary
Read full transcript →Vinci raises $250M to build a physics foundation model for semiconductor design
Hardik Kabaria, co-founder and CEO of Vinci, is building what he describes as a foundation model for physics, trained to evaluate hardware designs against the laws of thermodynamics, mechanics, and electromagnetism. The $250M round includes AMD and Applied Materials as corporate venture investors alongside Coastal Ventures.
“What we are building is a foundation model for physics... Physics is the compilation layer for hardware. It's been manual. With our product, we are making it automatic and scalable. This applies to semiconductor, high bandwidth memory. It also applies to satellite... We are today training on terabyte of data. And we will go to 10x, 100x over the next couple of quarters.”
The core thesis
Physics, Kabaria argues, is the compilation layer for hardware. When engineers generate thousands of chip design variations, they need a way to grade each one against physical reality before committing to silicon. Current tooling can't handle the thermal complexity of leading-edge products like high-bandwidth memory or GPUs at the fidelity and speed customers need. Vinci automates that evaluation.
The model is trained on the laws of physics rather than proprietary design files, which Kabaria says is the key to deploying inside the firewalls of TSMC, NVIDIA, and Micron without needing access to their actual designs. Physics inside a semiconductor fab is the same as physics outside it. That framing allows Vinci to offer on-premises deployment, zero forward-deployed engineers, and strict data privacy, which he treats as a sales requirement, not a feature.
Go-to-market
Sales are strictly top-down. Kabaria is direct that winning over a single engineer at a large chipmaker gets you nowhere. The entry point is a VP conversation anchored on two metrics that consistently drive decisions: time to market and performance. Customers building for a CES launch don't have room to miss either.
The industry's opacity, where semiconductor companies guard vendor relationships as IP, cuts both ways. It slows initial deals but also means successful deployments spread by reputation through a small, tightly networked industry rather than public case studies.
Expansion path
Vinci started with thermal problems in semiconductors and electronics. The roadmap extends to thermo-mechanical and electromagnetic modeling, covering every physics modality with established equations. Training data is currently in the terabytes, with a 10x to 100x expansion planned over the coming quarters.
Kabaria's background is in geometry and physics software, including eight years at Carbon, the Sequoia-backed 3D printing company, where he rose to VP of Software and shipped software across manufacturing floors in the US, Asia, Eastern Europe, and Mexico. He started Vinci three years ago.
Every deal, every interview. 5 minutes.
TBPN Digest delivers summaries of the latest fundraises, interviews and tech news from TBPN, every weekday.