Agentrys secures $24.5 million for agentic design automation in chipmaking

Agentrys, a startup focused on automating and continuously improving chip design, announced that it has raised $24.5 million in funding. The total includes an oversubscribed $19.1 million seed round led by Etna Labs, following an earlier $5.4 million pre-seed round led by MediaTek, the company's first strategic backer. The new capital will support hiring, agent-native tooling development, and expanded customer engagements across verification and physical design.

Chip design remains economically critical yet still relies heavily on scarce expertise and thousands of manual engineering hours per chip. Agentrys was founded by Mark Ren, who brings nearly three decades of experience in Electronic Design Automation (EDA) and AI R&D from NVIDIA Research and IBM Research, including leading ChipNeMo, the first industrial large language model built for chip design. The company was created to close this gap by helping engineering teams embed an agentic workforce directly into their chip design workflows.

Agentrys is establishing a new category it calls Agentic Design Automation (ADA), which moves chip-design automation beyond traditional EDA tools focused on individual tasks toward intelligent systems that automate, learn from, and continuously improve entire engineering workflows. The company's long-term goal is to advance self-improving AI for chip design, allowing engineering systems to learn from experience, expand their capabilities, and handle increasingly complex design challenges. Through ADA, Agentrys targets the workflows that determine whether chips in the $800+ billion and growing semiconductor industry ship on time while meeting functional and performance targets.

Agentrys differentiates itself from other agentic EDA solutions in two main ways. First, it offers an open platform that customers can build on, rather than a fixed set of vendor-controlled agents, allowing them to develop and own an agentic engineering workforce tailored to their own workflows across commercial EDA tools, internal tools, and infrastructure. Second, its design intelligence layer continuously learns from customer data, usage patterns, and evaluation signals, while agent-native tools and custom models expand what general-purpose AI agents can achieve in chip design.

Mark Ren, Founder and CEO of Agentrys, said that building production-grade agents capable of reliably automating real engineering work is a difficult problem, and that solving it is the company's core purpose. He explained that Agentrys helps engineering teams convert manual workflows into an agentic workforce that designs chips faster, produces better results, and systematically improves over time, distinguishing this from simply renting capability that competitors can also access.

Members of Etna Labs' investment team noted that chip design is among the first high-value domains where recursive self-improvement can become practical, since outcomes can be objectively measured using rigorous engineering tools and metrics. They pointed to Ren's history of influential AI-for-chip-design work and said Agentrys is building a self-improving agent layer across the EDA stack that improves with every design. Etna Labs is a research-driven investment firm that has backed more than 20 AI companies, with a particular focus on foundation models and their application to science, robotics, and other frontier domains.

Brian Hsu, Managing Director of the MediaTek Innovation Fund, said semiconductor R&D has its own distinct domain specificity within agentic AI development. He described Agentrys' approach as a builder-oriented solution that lets semiconductor engineering teams customize and improve AI agents for chipset design workflows, which could help capture and reuse domain knowledge and scale agent development systematically for semiconductor R&D.

Agentrys has already demonstrated its platform end to end, running an autonomous multi-agent workflow that carried a 32-bit CPU from specification through sign-off-clean GDS layout without human involvement. On NVIDIA's public CVDP verification benchmark, Agentrys is among the first to surpass 90% accuracy, with its agent evolution techniques consistently improving performance across successive iterations. The company has partnered with several of the world's top fabless semiconductor firms, a leading global foundry, and multiple chip startups, and is currently engaging customers across full digital and analog design flows, with system-level design planned for the future.

Ren added that every workflow execution generates engineering observations, each of which is evaluated, and each evaluation feeds back into improving the system. He said the company's vision is an engineering workforce that continuously learns and improves, and that this new funding will help extend that capability across the full design flow for more teams.