Drill Down Insider #24: Vinci CEO Hardik Kabaria

Cory Johnson (0:05)

Welcome to The Drill Down Insider. This is the Business Stories Behind Stocks and Move. I'm Cory Johnson, Chief Market Strategist for Epistrophe Capital Research. We're looking at a really interesting company called Vinci. It says AI gets physics right without ever having to train on customer designs. More than a dozen chips and electronic companies are using this company's software, but will it replace Cadence and Synopsys? The CEO is going to answer those questions. But first, it's sponsor time. The Drill Down brought to you by Aera. Never miss another critical insight or event ever. With Aera, customize your company's watch lists and track key events, mentions, filings and more. All with an easy to use customizable interface. That's Aera, AI, eera.com.

Well, I'm Cory Johnson. Welcome to The Drill Down. Inside of we look at the businesses that are running technology here.

AMD, no less than AMD, put money into a startup whose software predicts how its chip will heat up and bend before anyone builds it. Hardik Kabaria joins us right now, the CEO of that startup, Vinci, down there in Palo Alto.

Hardik, glad to have you. Thanks for your time.

Hardik Kabaria (1:03)

Well, thanks for having me, Cory.

Cory Johnson (1:05)

Now, you've got a Stanford doctorate, I guess, in computational geometry. You've been at, you led software at Carbon and founding this company, Vinci, in 2023 with your CTO, Sarah Ostentowski, I guess is how we say her name. Talk to me about the problem you were trying to solve. Because the issues around designing chips have been well addressed, I would think, by the likes of Synopsys and Cadence and that other company that does so well in that industry and in electronic design.

What is it that was missing in the market?

Hardik Kabaria (1:36)

Yeah, I mean, let's start with something basic. Like physics is everywhere. Everything around us is governed by laws of physics. Energy in will be energy out. The laws of physics would be the same inside the firewall of TSMC, inside the firewall of AMD, or outside. And everything we design will be performed by it. In a different way of saying, you can say physics is the compilation layer of hardware.

And we want to explore more and more ideas so that we can build the next-gen best ship, satellite, rocket ship, all of these parts which are going to be governed by laws of physics. More ideas we generate, we need a mechanism to grade it. Like which is a better idea? Which idea will lead to more efficient compute performance? Which idea will allow us to have less weight, better performing cooling? And that grading happens with physics. Now, to answer your question, aren't the design problems solved? Or isn't the design process completely optimized? Today, to perform a physics analysis, the first thing you need is a person who generally has a background like me. They would have a Ph.D. or a graduate school in electrical, mechanical, chemical engineering.

And they would use these software, and it would take them a long time amount of human work and a compute time to perform an analysis.

And that still doesn't allow them to perform analysis at manufacturing resolution. So on one hand...

Cory Johnson (3:08)

So explain that. How so?

Hardik Kabaria (3:11)

So just if you think about semiconductor, the feature size keeps getting smaller. It's a single-digit nanometer. And then the systems keep getting complex.

In other words, if you think about memory, memory we are stacking on top of each other.

Cory Johnson (3:24)

Sure, sure, sure. I bet the memory is crazy.

Hardik Kabaria (3:27)

Or now we are not just stacking memory. We also want to stack up semiconductors. So you have a 3D IC chip on top of chip with hybrid bonding layer.

The feature size gets small, the systems get complex. As it is, we couldn't analyze it perfectly without human in the loop.

So now the gap is bigger.

You cannot do it with the existing human talent, and the tooling is not allowing you to go to the manufacturing resolution. So that's the gap why we started the company. That hey, if you have a physics question, the universe gives you answers. Like you now drop my phone, it just goes there. There's a perfect traction. There is energy will be balanced. So if that's the case, why can't we build that so that anybody who has an idea, they can get that idea evaluated, how this part will be here in the real world? How is it going to get manufactured? Is it during the manufacture, is it going to bend so much that it will deform? That's what semiconductor industry called warpage problems. And that's going to causing so much delays and reliability issues. So that's why we started the company. If humans can manufacture it, if humans can design it, we should be able to evaluate it at compute capacity.