EDA: The Philosophical Perspective

The Free Silicon Foundation, or F-Si, sits on the policy and philosophy side of this cluster. Among its co-founders is Thomas Kramer, an electrical engineer who studied at the Eidgenössische Technische Hochschule Zürich (ETH Zurich, the Swiss Federal Institute of Technology Zurich).

Thomas’s path into open silicon started in a university lecture in the 2010s. As an electrical engineering undergraduate at ETH Zurich, he took a Very Large Scale Integration (VLSI) design class to learn how to arrange the millions or billions of transistors on a modern chip: The homework software required a signed NDA. That did not feel right to him, so he asked his lecturer if there were open source alternatives he could use at home — the lecturer did not know.

After Luca Alloatti completed his PhD in electrical engineering in Germany and a postdoctoral position at MIT — where he helped build the first microprocessor to use light rather than electricity for some of its internal connections — he found himself inside the “intellectual force field” of the Free Software Foundation. Richard Stallman, the FSF’s founder, has spent more than 40 years arguing that software users deserve the freedom to run, study, modify, and share the code they depend on. “That contaminated me,” Luca says with a smile. He went back to Europe to apply the same idea to silicon.

In 2018, when Luca brought the idea back to European academia, his colleagues called him quixotic, from Cervantes’ deluded knight. But the windmills were not imaginary. Today, hundreds of chips have been taped out with open tools, the European Commission is funding open EDA development, and many of the same professors who shook their heads at Luca now describe themselves as long-standing supporters of open silicon. The knight turned out to have been looking at giants all along.

F-Si, part of the Next Generation Internet Zero (NGI0) Consortium, runs the annual Free Silicon Conference (FSiC) and publishes formal policy recommendations to the European Commission. They’re also developing the Hardware GPL (HGPL), a copyleft license for silicon chip designs that requires derivative works to carry the same terms. The HGPL will address a real, documented gap. The strongly reciprocal CERN Open Hardware Licence, CERN-OHL-S, is incompatible with the GPL, a clash of two strong copyleft licenses that both CERN and F-Si acknowledge. Because a complex silicon design is likely to incorporate or link GPL-licensed code, F-Si argues a GPL-compatible hardware license is needed to keep the free-hardware and free-software communities from fragmenting.

The foundation also maps where the field’s knowledge actually lives: F-Si ran a language model over a September 2025 dump of the OpenAIRE thesis database and pulled out 13,272 related to chip design and EDA, each scored across 53 categories. The result is a public, searchable database of where the discipline has already done its thinking, useful both to a researcher entering a niche and to F-Si when they look for university groups to visit.

Underneath the policy and licensing work is a framing Luca returns to often. “Digital infrastructure is public infrastructure,” just like roads, hospitals, and the power grid. “We all need it. We all rely on it.” When public infrastructure goes private, the public pays twice: once with taxes to build it, again with fees to use it.

EDA: The Industrial Perspective

Naja, an open source EDA project that has received funding from NGI0 programs, sits on the industrial-adoption side of the EDA cluster. The company that built it, Keplertech, was founded by Christophe Alexandre (CEO) and Noam Cohen (CTO).

Christophe has spent his entire career in EDA. He did a PhD, co-founded a startup in 2005 and watched it get acquired by Mentor Graphics (originator of the Calibre sign-off tool), watched Mentor get acquired by Siemens, and decided after 15 years that he had had enough of corporate politics defending market position rather than improving products. Noam arrived from a parallel direction: electronic engineering studies in Israel, an internship at the US chip company Marvell, and a job at Synopsys in Tel Aviv that brought him to Paris through an internal transfer. He and Christophe met at Siemens. Both of them smile when they describe realizing they wanted to leave and build something else.

They had a specific frustration with the standard startup playbook in the EDA space. That playbook, Christophe explains, is: build a small plugin that solves one narrow problem inside one of the big three’s toolchains, generate some traction, get acquired. “If you do just plugins,” he says, “you’re going to maybe replace one box in the stack, but you need to replace 50 boxes” to make the stack open. And that was not on offer.

Naja’s choice of Apache 2.0 follows directly. A permissive license, it lets industrial users incorporate the code into their own products without being required to release their work in turn. F-Si pushes copyleft for a different reason, to preserve the public-good character of the work downstream. The contrast between the two licenses is tactical rather than philosophical. Both want adoption and openness: They are aiming at different doorways.

Naja’s foundation is a set of EDA tool-building libraries written in C++. These building blocks are accessible through Naja EDA, a thin Python wrapper that can be installed with a single command (pip install najaeda). On top of Naja EDA sits Kepler Formal, an equivalence checker — a tool that confirms two versions of a chip design are functionally identical even after someone has rewritten or optimized one of them. OpenROAD has adopted Kepler Formal to verify that its design transformations don’t change chip functionality between iterations.

Naja started near the bottom of the human-relevant part of the stack, in a step called post-synthesis, and is now climbing upward toward the engineers who write the original code. Christophe describes the trajectory as moving “closer to the humans.” A SystemVerilog front end is integrated; logic equivalence checking at the human-readable level is the next step.

Kepler Formal is also about to ship as a GitHub Action, an automation that runs whenever someone proposes a change to an open chip design project hosted on GitHub. If a contributor submits a pull request, a GitHub runner can invoke Kepler Formal and confirm that the change has not silently broken the design’s logic. Continuous integration like this is unremarkable in software development, but it does not exist anywhere else in EDA today.

A note in defense of the industry Naja is trying to “chip” away at, which Noam and Christophe both volunteered without prompting: Chips cannot be patched. A bad chip is a recall, a lawsuit, or a generation of products written off. The cost of an EDA license, even at the proprietary list price, is dwarfed by just the electricity bill for the servers that run the tools — never mind the cost of materials and the other consequences of failure. The industry’s conservatism is rational, even when the lock-in it creates is not.

Different Doors, Same Building

F-Si and Naja can look like opposites: a philosophical, copyleft-pushing, policy-focused European foundation on one side and a permissive-licensed, industrial-minded startup on the other. The people doing the work do not see it that way.

Luca puts the point plainly: “I would say there are not different approaches; it’s one big one.” Within the open ecosystem, between F-Si and Naja, OpenROAD, and dozens of others, there is no real conflict. People share tools, attend each other’s conferences, and contribute to each other’s projects. Christophe has presented Naja at F-Si’s own Free Silicon Conference every year from 2022 to 2025. The choice of license is a tactical decision about which doors a project wants to walk through, not a wall between rival camps.

The cooperation runs deeper than conference invitations. Naja’s front end relies on Verilator and the Slang parser, both of which originated in the US open source ecosystem. OpenROAD, which began life at the University of California, San Diego, is integrating Kepler Formal into its own workflow. The users span continents, and the dependencies cross borders; the ecosystem is international by design.

In this industry, “international by design” cuts to the heart of what digital autonomy can and cannot mean. A modern chip is the product of a global supply chain that nobody, anywhere, fully controls. The foundries that build leading-edge silicon are concentrated in Taiwan, with smaller capacity in Korea, the US, and (increasingly) Germany. The machines those foundries use to print transistors at the nanometer scale come from a single Dutch company, ASML, with no real competitors at the leading edge. End markets are everywhere. No region is anywhere close to being able to do the whole thing on its own.