Synopsys Plans to Plug Chinese AI Models Into Chip Design in China

Synopsys, a California-based supplier of electronic design automation (EDA) tools, is exploring the possibility of working with Chinese AI model developers. The goal is to give customers in China a package of its own software and local LLMs that would speed up chip development. The reason is practical: OpenAI services are unavailable in China, and without a language model a modern EDA suite with AI assistants loses a noticeable share of its features. The company stated its main condition up front: everything must comply with US export restrictions.

Why EDA companies need language models

Synopsys and its main competitors have long been building AI into the design flow. That includes optimizing block placement on the die, generating tests, and assistants that write Verilog code, parse verification reports, and answer questions about documentation. As far as is known, some of these features at Synopsys were built on OpenAI models via the Microsoft cloud. In China, that route is closed. Customers there get a cut-down set of capabilities, even though demand among local companies for designing their own chips is high precisely because of restrictions on importing ready-made accelerators.

Chinese models are an obvious candidate in this situation. Among local LLMs there are large open families that handle code reasonably well. The source names no specific models, so for now this is an intention, not a signed contract.

What is known so far

• Initiator: Synopsys, a California company and EDA tool supplier.
• Idea: combine its own products with Chinese AI models for local customers.
• Reason: OpenAI services are unavailable in China.
• Constraint: the collaboration must comply with US export rules.
• Status: the company is "considering" the option; partners, timelines, and terms have not been disclosed.

Where the legal line runs

US export controls on EDA have already shown how fragile this business is. In spring 2025, Washington restricted shipments of such software to China, and in the summer, after trade agreements, lifted the restrictions. The episode reminded the whole industry that access to design tools is used as a pressure lever, and that rules can change fast.

So the wording about compliance with restrictions is not a formality here. Any integration runs into several questions. Where does customer data go: to a Chinese provider's cloud or to the customer's own local server? Does design data end up in model training or fine-tuning, and, in the other direction, does Synopsys technology leak out? And finally, which chips are being developed with the help of such a package? Advanced process nodes and high-performance AI accelerators fall under strict control. A language model that suggests a block architecture to an engineer could, by this logic, also become part of a regulated technology chain.

There is also reputational risk. An American company that openly cooperates with the Chinese AI sector will inevitably draw the attention of Congress and the relevant agencies. For Synopsys, this is a balancing act between the Chinese market, which remains a substantial part of its revenue, and political pressure at home.

What to watch next

For investors and analysts, the main indicators are these: whether Synopsys names a specific partner, whether the solution will work only inside China, and how US regulators react. If the scheme is approved, or at least not blocked, Cadence and other tool suppliers may follow Synopsys. Chinese models would then gain a new enterprise segment: engineering tasks in the semiconductor industry.

For engineering teams, the practical question is different. If you work with EDA tools in China, or on projects with Chinese contractors, check how design data transfer is set up in your environment and who is responsible for compliance. Today, companies have only a statement of intent. The actual terms will become clear once partner names and the technical details of the integration emerge.