EDA: The Hidden Software Behind the US-China Chip Race

Most discussions about the chip industry focus on factories and lithography machines. Having worked on security IP for the semiconductor industry, I find the software side just as interesting: no modern chip can be designed without it.

What EDA actually does

Electronic design automation (EDA) is to chip designers what CAD is to architects, at a vastly larger scale. A modern processor contains billions of transistors, far beyond what anyone could lay out by hand. EDA tools take a design from a high-level description through logic synthesis, placement, routing, timing and power optimisation, and finally verification. Each step relies on a different specialised tool, and foundries also use EDA to confirm that a design can actually be manufactured before committing to an expensive production run.

A small market with enormous leverage

Measured in revenue, EDA is a niche, worth only a small fraction of the overall semiconductor market. Yet it is concentrated in very few hands: Synopsys, Cadence and Siemens EDA (formerly Mentor Graphics) together hold roughly three quarters of the global market. That combination of small size and high concentration is exactly what makes EDA a pressure point. Restricting the software of a few companies can slow down chip design in an entire country.

From targeted controls to a full stop, and back

  • 2019: Huawei loses access to US EDA tools after being added to the US Entity List.
  • August 2022: the US introduces export controls on EDA tools needed to design chips with gate-all-around (GAAFET) transistors, the most advanced transistor architecture.
  • May 2025: the US requires licences for EDA exports to China, and the three leading vendors pause sales and support there.
  • July 2025: the requirement is lifted again as part of a broader trade arrangement involving rare earth exports.

The episode showed both ends of the lever. For China, it exposed a deep strategic dependency. For the EDA vendors, for whom China is a significant share of revenue, it showed how quickly a major market can disappear overnight.

Why replacing EDA is so hard

China has made domestic EDA a strategic priority and now has more than a hundred local EDA companies. Yet the largest of them, Empyrean, still holds only around six percent of its home market. The reasons are familiar to anyone who has built engineering tools:

  • A complete flow is needed: chip design depends on a tightly integrated tool chain, not on a single good tool.
  • Foundry trust takes years: tools must be qualified for each manufacturing process in close cooperation with the fabs.
  • Decades of know-how: algorithms, corner cases and customer feedback have been accumulating since the 1980s.
  • Scarce talent: EDA development requires a rare mix of software engineering, mathematics and semiconductor physics.

Software is also harder to control than hardware. A lithography machine is large and traceable; software licences can be retained, copied or obtained through intermediaries, which makes the real effect of software export controls harder to predict.

Lessons for engineers outside the chip industry

Even if you never design a chip, this story is worth thinking about. Our own work depends on tools we rarely question: compilers, static analysis tools, simulation environments and cloud services. In automotive projects, for example, a qualified toolchain is often used for years and cannot simply be swapped. Knowing which tools are critical, where they come from and what the alternatives are is part of building resilient software, not just a procurement question.

Sources

  • MIT Technology Review, explainer on EDA and US-China export controls, August 2022. Link
  • TechCrunch, “US imposes new rules to curb semiconductor design software sales to China”, May 2025. Link
  • Network World, “US lets China buy semiconductor design software again”, July 2025. Link
  • TrendForce, EDA market share data, 2024. Link