Verification is repetitive, exacting, and unforgiving of small mistakes. A single missed logic flaw can delay a product launch or cost millions in rework. ChipAgents, a Santa Barbara startup led by UC Santa Barbara professor William Wang, wants to make that process faster and less error-prone using agentic AI systems trained to generate and verify chip designs.
ChipAgents, founded in 2024 and based in Santa Barbara, has raised $21 million in a Series A round led by Bessemer Venture Partners, with participation from Micron Ventures, MediaTek, Ericsson, and ScOp Venture Capital. The new round brings the company’s total funding to $24 million.
“Electronic design automation (EDA) has powered decades of semiconductor progress, but today’s engineers still struggle with fragmented toolchains, steep costs, and workflows that haven’t kept pace with rapidly growing chip complexity,” Wang said in a statement announcing the funding. “As designs scale, productivity gains have slowed and verification remains highly manual. Our vision is to change that.”
Reinventing Verification
ChipAgents builds generative and reasoning-based AI systems that automate parts of chip design and verification—tasks that often consume more than half of a chip’s total development time. Modern chips can contain billions of transistors, requiring engineers to manually validate every interaction to ensure functional correctness before fabrication. A single verification error can cost millions in rework or failed silicon.
The company’s flagship product combines large language models, formal reasoning, and deep reinforcement learning to generate, test, and debug register-transfer level (RTL) code. Engineers can use natural language prompts to define or validate logic, while the AI agents handle tasks such as testbench generation and waveform analysis. The result, according to ChipAgents, is an 80% increase in verification productivity and a compression of multi-week workflows into days or minutes.
ChipAgents operates at the front end of the design process, before physical layout or manufacturing, where speed and correctness have the highest leverage. Its system is designed to integrate with existing EDA tools rather than replace them, a strategy that allows the company to embed itself in established semiconductor workflows.
In January, the company reported that it achieved a 97.4% pass rate on NVIDIA’s VerilogEval-v2 benchmark, outperforming models developed by UC San Diego and NVIDIA’s own VerilogCoder. The benchmark measures the ability of AI systems to generate functionally correct hardware code from textual specifications.
A Signal of Confidence in AI-Driven EDA
The new investment comes with strong endorsements from both venture and semiconductor veterans. Bessemer partner Lance Co Ting Keh called ChipAgents “the best product in the market that does AI-powered RTL design, debugging, and verification for chip developers,” adding that the platform “brings together disparate EDA tools from spec ingestion to waveform analysis.”
Micron Ventures investment director Henry Huang said, “ChipAgents’ mission to reinvent chip design through AI agents aligns closely with our focus, especially as chipmakers require faster, more intelligent design solutions to meet the accelerating demands of AI hardware.”
The company has also assembled a high-profile advisory board that includes Wally Rhines, former CEO of Mentor Graphics; Raúl Camposano, former CTO of Synopsys; Jack Harding, former CEO of Cadence; and Erez Tsur, former CEO of Cadence Israel. All four helped build the companies that defined the traditional EDA landscape, and their participation lends credibility to a startup aiming to reshape it.
According to Wang, ChipAgents has deployments with 50 semiconductor companies and has seen sales grow 50-fold year over year, alongside a 6,377% increase in usage in the first half of 2025. The company is establishing a new headquarters in Santa Clara to be closer to its customers, while maintaining its Santa Barbara R&D operations under its parent entity, Alpha Design AI.
From Research to Workflow
Wang’s academic background continues to inform the company’s direction. At the University of California, Santa Barbara, he holds the Duncan and Suzanne Mellichamp Endowed Chair in Artificial Intelligence and Design, where his group developed DeepPat, a reinforcement learning framework for reasoning models. He sees the company’s technology as an extension of that work: applying explainable, formal reasoning to one of the most error-sensitive domains in computing.
In the long term, Wang wants to create a full agentic workflow that spans from RTL design to GDS layout, connecting pre-silicon and post-silicon validation. The aim is to build self-improving AI systems that can reason about hardware intent and optimize future iterations based on real-world performance data.
Wally Rhines, who has advised multiple generations of EDA companies, said, “ChipAgents demonstrates the major impact that AI can have on the full range of integrated circuit design tasks. I’ve met with three major semiconductor companies that have done competitive assessments of AI-based design solutions, and ChipAgents is the number one choice at all three.”
For now, ChipAgents is positioning itself as a complement to traditional EDA workflows rather than a replacement. But its approach, embedding AI reasoning into the verification stack, suggests a shift in where engineering value is created. As Wang put it, the competitive edge in chip design may no longer come from who builds the most powerful chips, but from who can best collaborate with AI to design them.
