Global electronic design automation (EDA) company Siemens EDA is aggressively pursuing an “AI-Native Design” strategy that embeds artificial intelligence (AI) into every stage of semiconductor design. Building on its collaboration with NVIDIA, the company aims to dramatically accelerate chip design speed and productivity by constructing a workflow where multiple AI agents autonomously make decisions and verify and correct their own results.

Anchor Gupta, Senior Vice President of Siemens EDA’s IC Products division, spoke at the “Siemens EDA Forum Seoul 2026” held at the Lotte Hotel in Jamsil, Seoul, on August 11. “An engineer can tell the AI system, ‘I’m going to sleep now. Finish this task. Verify it, debug it, and fix any problems. Don’t ask me questions—resolve them yourself,'” Gupta said. “It’s like delegating overnight work to a subordinate.”

Gupta identified “AI Everywhere” as a core transformation in the semiconductor industry. “By 2028, 70% of all chips produced are expected to have AI accelerators onboard,” he noted. “Unprecedented silicon-to-system complexity and the rapidly changing pace of the AI market are impossible to keep up with using traditional methodologies.”

As a solution, Siemens EDA presented an AI-native design strategy centered on speed, productivity, and reliability. The core idea is to embed AI not merely as a design assistance tool but directly into the design process itself, automating repetitive tasks and verification procedures.

The strategy’s centerpiece is the collaboration with NVIDIA. The two companies have strengthened a “self-verifying agentic AI” workflow for semiconductor and printed circuit board (PCB) design. Specifically, Siemens’ “Fuse EDA AI Agent” system is combined with NVIDIA’s accelerated computing infrastructure, Nemotron inference models, NeMo Gym, and Openshell secure runtime. This allows AI agents to connect with EDA engines, verify the decisions made during the design process, and autonomously correct errors.

Siemens EDA’s AI evolution is progressing rapidly. In June last year, the company first introduced Fuse, which allows users to interact with EDA tools using AI and natural language. In March of this year, it applied a single AI agent. Last month, it added the capability for multiple agents to collaborate on long-duration tasks and self-verify the results. “We’ve evolved from conversational AI to multi-agent AI,” Gupta explained.

When applied to “cell characterization”—the process of calculating the power and speed of basic semiconductor cells under varying temperature, voltage, and process conditions—the setup and execution time was reduced from weeks to just days. Characterization processing speed improved by more than 10 times, and token costs were reduced by 5 to 10 times. The AI agents locate the necessary files and tools, perform calculations and verification, and if problems arise, debug and iterate corrections. “There is no human intervention in between,” Gupta emphasized.

Results generated by AI are re-verified by EDA engines that incorporate mathematical algorithms and semiconductor physics laws. “In semiconductors, even a single small error can lead to chip failure or yield loss,” Gupta said. “The most important mission is to deliver trustworthy results.”

Siemens diagnosed that increasing semiconductor complexity—such as designs integrating over 200 billion transistors and thermal densities exceeding 100W/cm²—makes it difficult to keep up with development speed using conventional methods alone. The expansion of “heterogeneous integration,” which combines chips made from different types and processes into a single package, is another factor increasing design and verification complexity.

Siemens EDA provides an integrated environment connecting design to verification, based on its proven product portfolio including Calibre, Questa, and Solido. Calibre, a physical verification tool, is used by 98 of the world’s top 100 chip development companies. Using Questa for functional verification alongside Tessent for semiconductor testing can increase automatic test pattern generation (ATPG) simulation speed by 3 to 5 times compared to previous methods.

An “open ecosystem” that avoids locking AI models to a specific vendor is another pillar of Siemens EDA’s strategy. While utilizing NVIDIA’s Nemotron, Fuse is designed to connect with various large language models (LLMs) from OpenAI, Google, and Anthropic. This supports a “Bring Your Own Model” (BYOM) approach, allowing customers to choose their preferred AI model. “In a rapidly changing AI ecosystem, relying on a single model would slow down the pace of innovation,” Gupta said. “We designed it from the ground up so customers can bring and use the models they want.” Additionally, customer tool usage data is not used for AI model training, strengthening design data security.

Regarding differentiation from competitors Synopsys and Cadence, Gupta stated, “In the agent era, multiple agents can communicate with different tools, reducing dependency on a single platform. Our top priority is delivering high-quality EDA engines.”

German industrial giant Siemens has invested over 25 billion euros in its software business since 2007, building related capabilities. In 2017, it significantly expanded its business scope by acquiring U.S. EDA company Mentor Graphics.

Siemens’ third-quarter fiscal 2026 (April-June) revenue reached 20.794 billion euros, up 7% year-over-year. Industrial business profit rose 25% to 3.524 billion euros. Revenue for Digital Industries, which includes the EDA business, increased 12% to 4.932 billion euros, while its business profit surged 44% to 923 million euros.

Key financial metrics are as follows:

ItemQ3 Fiscal 2026 (Apr-Jun)YoY ChangeTotal Revenue20.794 billion euros+7%Industrial Business Profit3.524 billion euros+25%Digital Industries Revenue4.932 billion euros+12%Digital Industries Business Profit923 million euros+44%

Note: The Digital Industries segment includes the EDA business.

Gupta disclosed that the EDA business grew over 30% in the recent quarter. “South Korea is a leading market not only in memory but also in AI chip and accelerator design,” he said. “We will collaborate with South Korean customers to support faster development of products needed in the AI era.”

Siemens EDA’s AI-native strategy is noteworthy for its potential to fundamentally reshape the semiconductor design paradigm. By having AI agents autonomously perform repetitive verification tasks that once required constant engineer intervention, dramatically shortening design cycles, and securing reliability through an open ecosystem that avoids dependency on a single AI model, this approach is expected to set a new benchmark in the semiconductor industry’s race for development speed.