Why are so many startups developing chip design tools?

Part 1: Primitive Instruments, and hardware at the speed of software.

Silimate. Silology. InstaChip. Primitive Instruments. Astrus. Primis. There are a lot of companies developing tools for chip design all of a sudden. But why?

Chip design is a massive industry, but it’s been a massive industry for decades. For years, the world of Electronic Design Automation (EDA) tooling was the realm of a small number of huge companies like Cadence and Synopsys. Few startups dared to try to enter the space. Most who did, died, and the ones who didn’t were acquired early on by Cadence or Synopsys. Only recently have startups really started going after the EDA space in earnest.

There’s clearly something about the present moment that makes starting EDA startups more attractive. Personally, I think the new slate of EDA startups is representative of larger changes in the chip design world, and especially the effect that the AI boom has had on chip development. But as much as I love writing my own thoughts, I wanted to get a deeper insight – so I went and talked to the founders of two of my favorite chip tooling startups, Silimate and Primitive Instruments. This week, we’re focusing on Primitive Instruments, and how they are building tools to help make hardware at the speed of software.

More startups are making chips.

In 2010, if you were selling EDA tools, you only had a few potential customers, and they were mostly large chip companies like Intel, Nvidia, and AMD. These big companies are pretty set in their ways: they have a certain design process for their chips and that design process is fairly calcified. They’ve spent years building up bespoke internal tooling to help them design their chips. Overall, they’re really hard to sell to.

But now, with the AI chip boom, there are many new companies trying to design chips. These companies are smaller, and more willing to try new things. More importantly, though, these new companies have far fewer engineers than the incumbent behemoths like Intel. That means they need to rely more heavily on powerful tools to give them a leg up. It also means that they don’t have the resources to develop their own bespoke internal tools for every challenge they might face; instead, they need to look to the market to find tools that can help them.

Startups like Primitive Instruments are some of the tools these new chip companies might leverage to build chips faster and cheaper than their incumbent competitors. According to Dylan Stien, the cofounder and CTO of Primitive Instruments, this new batch of ASIC companies is a big part of why they think it’s the right time to be starting a company building silicon tooling: “There’s been an explosion of new silicon, especially with new ASIC and RISC-V startups [...] Silicon valley is booming with new chip companies. We should sell tools to help them make those chips better.”

Hardware development is taking inspiration from software development.

In the past couple of decades, software development has been made into a science. With CI/CD pipelines, Jira dashboards, automatic log parsing, and more, so many of the annoying and manual pieces of software development have been automated away.

That isn’t the case in the hardware industry. Every chip design company I’ve been at built their own CI/CD and regression testing process from scratch. Log files were often inspected by hand, with management hoping that the engineers wouldn’t miss a critical warning. Engineers had to jockey for simulation resources and for access to test hardware. It feels like software development in the early 2000s.

Primitive Instruments is trying to bring hardware development into the modern era. Chase Zimmerman, their CEO, describes Primitive as a “full-stack company for building physical products.” “Software for hardware hasn’t gotten a lot of love, and we want to change that. Everything from designing a chip to getting a physical product in your customers hands should all be integrated,” Chase said. Essentially, Primitive wants to be the backbone on which all hardware development runs, starting with chips.

For chip companies using Primitive’s tools, the problem of building out custom CI/CD pipelines, managing floating licenses, parsing log files, and tracking test progress will be automated away, just like it is in the world of software. This is even more important for chip startups, who don’t have the resources to staff their own dedicated, in-house teams to build custom internal infrastructure. At chip startups, talented chip engineers often get sidetracked building bespoke tools. Soon, they’ll be able to buy those tools from Primitive, and spend more time focused on actually designing chips.

Primitive is also taking a much more modern approach to software development than incumbent players in the EDA tool industry, like Cadence and Synopsys. Most tooling for hardware engineers is powerful, but incredibly difficult to use. Chase and Dylan want users to have a different experience: “You should log on to our tool and be happy.”

This is even more important as chips get larger and more complex, and fewer and fewer American students choose to go into chip design as a career. According to Chase, “[s]ilicon companies are feeling the bottleneck of a lack of engineering talent, and designs are getting more complex. [...] I went through the computer engineering program at USC, and there’s not an insignificant number of students who went through VLSI [classes] and went, ‘I don’t want to do this’. They don’t want to do this in their day-to-day lives because the tools suck. You have to make this job easier for people to choose.”

As somebody who has made that choice to be an electrical engineer, and has to fight against the legacy tools every day to get my work done, I hope Primitive keeps succeeding. If hardware engineers want to keep building more powerful and complex chips to support new AI and scientistic computing workloads, we need startups Primitive to disrupt the status quo by building powerful, easy-to-use software for silicon engineering teams.

Disclosure: Normal Computing, my employer, is a customer of Primitive Instruments.