Agentrys is attempting to shift semiconductor R&D from manual, expert-dependent processes to a model of autonomous, self-improving engineering. The San Jose-based startup announced it has raised $24.5 million to scale its platform, which aims to replace traditional, task-specific automation with an "agentic workforce" capable of managing entire chip-design workflows. This funding includes an oversubscribed $19.1 million seed round led by Etna Labs, following a $5.4 million pre-seed round led by MediaTek. By targeting the high-stakes workflows of the $800 billion semiconductor industry, Agentrys is positioning its technology to address the critical scarcity of specialized engineering talent and the increasing complexity of modern silicon design.
Agentrys Funding and Strategic Expansion
The $24.5 million in total capital is earmarked for aggressive talent acquisition, the development of agent-native tooling, and broadening customer engagement within the verification and physical design sectors. Agentrys is led by founder Mark Ren, a veteran of NVIDIA Research and IBM Research who previously led the development of ChipNeMo, the first industrial large language model specifically for chip design. The company is attempting to define a new category termed Agentic Design Automation (ADA). Unlike traditional Electronic Design Automation (EDA) tools that focus on isolated tasks, ADA is designed to function as an intelligent system that learns from engineering observations to automate and improve entire workflows.
The company has already demonstrated the technical viability of this approach by running an autonomous multi-agent workflow that moved a 32-bit CPU from initial specification to a sign-off-clean GDS layout without human intervention. Furthermore, Agentrys has reported exceeding 90% accuracy on NVIDIA’s public CVDP verification benchmark. To support this scaling, the company is leveraging a team with backgrounds from NVIDIA, Meta, AMD, Samsung, Google, and Siemens. Current engagements include partnerships with several top fabless semiconductor companies, a leading global foundry, and various chip startups, with plans to eventually expand into system design.
Building an Autonomous Agentic Engineering Workforce
Agentrys is differentiating its platform from existing agentic EDA solutions by offering an open architecture rather than a closed set of vendor-controlled agents. This allows semiconductor engineering teams to build and own an agentic workforce that is customized to their specific internal tools, commercial EDA software, and existing infrastructure. The company argues that this approach allows firms to build a compounding intellectual asset rather than merely renting third-party capabilities. This "builder" model is intended to help companies capture and reuse domain-specific knowledge systematically across their R&D cycles.
Central to the Agentrys platform is a design intelligence layer that utilizes customer data, usage patterns, and evaluation signals to drive continuous improvement. The company is betting on the concept of recursive self-improvement (RSI), suggesting that because chip design results can be objectively measured using rigorous engineering metrics, AI agents can be trained to improve their own performance with every execution. By combining agent-native tools with custom models, Agentrys aims to extend the capabilities of general-purpose AI agents to meet the highly specific requirements of semiconductor R&D, specifically targeting the workflows that dictate whether chips meet functional and performance targets on schedule.
Key Takeaways
- Agentrys raised a total of $24.5 million, including a $19.1 million seed round led by Etna Labs and a $5.4 million pre-seed round led by MediaTek.
- The company's platform successfully completed an autonomous workflow for a 32-bit CPU from specification to sign-off-clean GDS layout without human involvement.
- Agentrys aims to establish the "Agentic Design Automation" (ADA) category, focusing on self-improving AI systems rather than traditional task-specific EDA tools.
TechInsyte's Take
In our view, Agentrys is making a high-stakes bet that the next frontier of semiconductor competitiveness will be defined by "agentic ownership" rather than tool proficiency. By providing an open platform that allows companies to build their own proprietary agentic workforce, Agentrys is directly addressing a major pain point for enterprise silicon teams: the risk of losing domain expertise to third-party AI vendors. This signals a shift in the EDA landscape where the value moves from the software tool itself to the intelligence layer that orchestrates those tools. If Agentrys can successfully implement recursive self-improvement in a production environment, they won't just be selling software; they will be selling a method for companies to systematically automate the most expensive and scarce resource in the industry—expert engineering intuition.
Questions & Answers
How does Agentrys differ from traditional Electronic Design Automation (EDA) providers?
Traditional EDA tools are generally designed to automate specific, isolated tasks within the design process. In contrast, Agentrys is building Agentic Design Automation (ADA), which aims to automate and improve entire engineering workflows through intelligent systems that learn from data and experience.
What is the strategic advantage of the Agentrys "open platform" model for semiconductor firms?
The open platform allows engineering teams to build and own a customized agentic workforce that integrates with their specific commercial EDA tools, internal infrastructure, and proprietary workflows. This enables companies to create a compounding intellectual asset rather than relying on fixed, vendor-controlled agents.
Can Agentrys' technology actually perform end-to-end chip design tasks?
The company has demonstrated an autonomous multi-agent workflow that successfully moved a 32-bit CPU from specification to a sign-off-clean GDS layout with no human in the loop. Additionally, the company has achieved over 90% accuracy on NVIDIA’s CVDP verification benchmark.
What role does "recursive self-improvement" play in the Agentrys roadmap?
Agentrys intends to leverage the fact that chip design results can be objectively evaluated using rigorous engineering metrics. This allows their design intelligence layer to use every workflow execution and evaluation as a signal to systematically improve the performance and capability of their AI agents over time.
Source: Businesswire
