Agentrys Raises $24.5M for Self-Improving AI Chip Design Platform

Agentrys secures funding for autonomous chip design platform

Agentrys has closed $24.5 million in funding to advance what the company calls Agentic Design Automation (ADA), a platform that enables semiconductor teams to deploy AI agents across chip design workflows. The financing includes a $19.1 million seed round led by Etna Labs and a $5.4 million pre-seed from MediaTek, according to details first reported by Business Wire.

Founded by Mark Ren, who led ChipNeMo—the first industrial large language model for chip design—at NVIDIA Research, Agentrys positions itself as an open platform rather than a fixed set of vendor tools. Engineering teams can build custom agent workforces that integrate with commercial EDA tools, internal systems, and existing infrastructure.

The company's design intelligence layer learns continuously from customer data and evaluation signals. Agent-native tooling and custom models extend capabilities beyond what general-purpose AI can accomplish in specialized chip design tasks.

Technical demonstrations and customer traction

Agentrys has demonstrated an autonomous multi-agent workflow that took a 32-bit CPU from specification through to sign-off-clean GDS layout without human intervention. On NVIDIA's CVDP verification benchmark, the platform achieved over 90% accuracy, with performance improving across successive iterations through what the company describes as agent evolution techniques.

The startup has partnered with multiple top fabless semiconductor companies, a leading global foundry, and several chip startups. Current customer engagements span full digital and analog design flows, with system design planned for future releases.

Why it matters

Chip design remains bottlenecked by manual engineering work and scarce expertise, even as semiconductor demand accelerates. Traditional EDA tools automate individual tasks but don't learn or improve from repeated use. Agentrys's approach—building systems that compound knowledge within a customer's own environment rather than offering generic rental capability—addresses a fundamental scalability challenge in an $800+ billion industry where time-to-market and performance targets determine commercial success. The recursive self-improvement model becomes practical in chip design because results can be objectively evaluated through rigorous engineering metrics.

Investment rationale and roadmap

Etna Labs, a research-driven firm backing over 20 AI companies focused on foundation models and frontier applications, led the seed round. The investment team cited chip design as one of the first high-value domains where recursive self-improvement can become practical due to objective evaluation frameworks.

MediaTek's Brian Hsu, Managing Director of the MediaTek Innovation Fund, emphasized the domain-specific nature of agentic AI development in semiconductor R&D. The strategic backer sees potential for the platform to capture and reuse domain knowledge, scaling agent development systematically.

Agentrys plans to allocate the new capital toward recruiting, developing agent-native tooling, and expanding customer engagements across verification and physical design workflows.

The Agentrys team includes researchers and engineers from NVIDIA, Meta, AMD, Samsung, Google, and Siemens, with backgrounds spanning AI for chip design, EDA, LLM systems, and silicon infrastructure. The group previously contributed to systems including ChipNeMo, NVCell, DREAMPlace, VerilogEval, and CVDP.

These details were first reported by Business Wire.