Why did Cadence acquire ChipStack?

And what does it mean for other AI EDA startups?

This week, Cadence Design Systems, one of the big 3 electronic design automation (EDA) companies that dominate the market for the software that engineers use to design microchips, acquired a small startup called ChipStack. ChipStack uses AI to speed up silicon verification, which is the process of testing chip designs in a simulation before sending them off to be manufactured. Cadence didn’t announce how much they spent buying ChipStack, which usually means it was a somewhat small acquisition — in the range of tens to hundreds of millions of dollars. But the implications for other startups developing AI-powered EDA software is massive.

For those of us who know the EDA industry well, this isn’t a surprise. As a matter of fact, this is the standard modus operandi of the EDA industry. Startups innovate, get acquired for $50-500M by the big players, and then get folded into their product portfolio. It’s much easier to sell EDA software when it’s bundled with existing Cadence products rather than packaged as a standalone product. That just leaves us to wonder: why did Cadence acquire ChipStack specifically? And what does this mean for all the other startups in the space?

Why ChipStack?

From the outside, ChipStack doesn’t look all that different from their competitors, like ChipAgents and Bronco. However, if you dig in a bit deeper, the differences get clearer. ChipAgents just raised a $21M Series A, while ChipStack has only raised about $7M to date — that means that ChipStack’s valuation is lower, so they could be acquired for a much lower price. Bronco, on the other hand, is run by college dropouts, and probably not a great acquisition target for a large, corporate company like Cadence.

A lot of the other AI-powered verification startups have similar stories; Cognichip raised $33M, so they probably would have been a lot more expensive to acquire than ChipStack. Normal Computing¹ offers an AI-powered verification product, but also develops thermodynamic computing chips, which are totally outside of Cadence’s wheelhouse. Ultimately, ChipStack was probably the most reasonably priced acquisition target that would fit well within Cadence as a parent company.

What about other AI chip design companies?

For AI-powered chip verification companies building direct competitors to ChipStack, like ChipAgents and Bronco, this acquisition is both a blessing and a curse. With Cadence clearly planting its flag in AI-powered verification software, the pressure is on for Cadence’s two biggest competitors, Synopsys and Siemens EDA, to make a move. This could mean that ChipStack’s competitors acquire their own AI-powered verification startups over the next few months.

However, there are a bunch of different AI-powered verification startups out there, and the EDA industry is an oligopoly. After Cadence, Synopsys, and Siemens all acquire AI-powered verification capabilities, there’s no other large player who could acquire any remaining competitors to ChipStack. That means that all those other startups either need to make it to IPO or get acqui-hired for far less than they’re worth — or they’ll go out of business.

Not all of the AI-powered chip design startups are competitors to ChipStack, though. Silimate, for example, is building an AI-powered tool to predict the performance of chip designs earlier on in the design process. Silogy is building an AI agent to debug test failures, rather than just writing tests using LLMs like ChipStack. These companies are in a much better position following this acquisition, because it signals that potential exit opportunities are opening up, without closing the door to a potential acquisition by Cadence.

Ultimately, the ChipStack acquisition is likely the first in a series of M&A announcements we’ll hear over the next couple years, as the EDA oligopoly gobbles up the best deals on AI-powered EDA startups. And unfortunately, the companies that don’t get picked by the end of it may end up dying.

1. Disclosure: I used to work at Normal Computing as their silicon engineering lead. I am still an advisor there.