AI-Powered Chip Design: How ChipAgents is Redefining the Semiconductor Landscape

The semiconductor industry is undergoing a transformation driven by artificial intelligence, a shift that promises to fundamentally alter how chips are designed, verified, and brought to market. At the forefront of this revolution is ChipAgents, a Santa Barbara-based AI startup that is making waves with its advanced AI-driven platform for Register Transfer Level (RTL) design and verification.

Chip design has long been one of the most intricate and time-consuming tasks in the technology sector, requiring engineers to navigate increasingly complex architectures with billions of transistors. Verification, which ensures a chip’s functionality aligns with its design specifications, often consumes more time and resources than the initial design process itself. ChipAgents aims to disrupt this paradigm by automating significant portions of the workflow through AI agents that can analyze specifications, generate RTL code, and validate functional correctness in ways that were previously impossible.

At the heart of this innovation is William Wang, CEO of ChipAgents and an AI professor at the University of California, Santa Barbara. Wang has spent more than 15 years researching artificial intelligence and its applications in various domains, and he believes that AI’s impact on semiconductor design is only just beginning.

“We are at a pivotal moment in the evolution of EDA (Electronic Design Automation) tools,” Wang said. “AI is already transforming software development, and now it’s time for hardware engineers to experience that same acceleration in productivity. ChipAgents is designed to integrate seamlessly into existing chip design workflows, allowing engineers to iterate faster, catch errors earlier, and ultimately bring products to market with greater efficiency.”

AI That Understands Hardware Design

One of ChipAgents’ biggest achievements has been its performance on NVIDIA’s VerilogEval-v2 benchmark, a widely recognized test that measures how well AI models can generate RTL code from written specifications. Achieving a 97.4 percent pass rate, ChipAgents outperformed leading competitors, including NVIDIA’s own VerilogCoder, which scored 94.2 percent, and UC San Diego’s MAGE, which achieved 95.7 percent.

The VerilogEval benchmark evaluates AI-driven chip design tools based on their ability to take a written hardware specification and generate syntactically and functionally correct Verilog code. Since modern semiconductor designs contain millions of lines of RTL code, the ability to automate and accelerate this process is highly valuable.

“The challenge in chip design is not just writing code—it’s writing correct code that adheres to complex functional requirements,” Wang explained. “Large language models alone struggle with this task. Even GPT-based models typically achieve only around 50 percent accuracy on VerilogEval. The key to our success has been integrating AI agents with advanced search algorithms and iterative refinement, allowing our tool to explore multiple solutions and arrive at an optimal outcome.”

ChipAgents does more than just generate RTL code. It actively assists engineers with functional verification, helping them create test vectors, optimize coverage, and debug designs more efficiently. By embedding AI into the verification process, the tool significantly reduces the time required to achieve a functionally complete and validated chip design.

Bringing AI to the Heart of EDA

The EDA industry has long been dominated by major players such as Synopsys, Cadence, and Siemens, companies that have spent decades refining the software tools used in chip design. ChipAgents, which was founded in 2024, is betting that AI-driven automation will disrupt this traditional landscape and create new efficiencies for semiconductor companies.

One of the company’s biggest advantages is that its AI-powered approach is tool-agnostic, meaning it integrates with any existing chip design environment. The platform supports industry-standard simulation and debugging tools, including Synopsys VCS, Cadence Xcelium, and Siemens Questa, ensuring that companies can adopt AI-driven automation without overhauling their existing workflows.

“Our approach allows engineers to use whatever tools they’re comfortable with,” Wang said. “ChipAgents doesn’t replace existing EDA tools; instead, it enhances them by acting as an intelligent co-pilot that helps engineers make better decisions, faster.”

The company has also demonstrated its AI capabilities beyond the VerilogEval benchmark. In December 2024, ChipAgents ranked number one on SWE-bench, an industry benchmark that evaluates AI-driven tools for software engineering. This result highlights the company’s deep expertise in large language models, search algorithms, and AI-driven automation, technologies that are now being applied to semiconductor design.

Solving the Scale Problem in Chip Verification

One of the greatest challenges in modern semiconductor development is scale. As chips grow in complexity, verification becomes exponentially more difficult. Engineers must contend with vast amounts of RTL code, multiple interconnected IP blocks, and a wide variety of test cases designed to validate every possible function of a chip.

ChipAgents addresses this challenge by using AI to understand the engineer’s workspace and dynamically prioritize verification efforts. Instead of requiring engineers to manually sift through thousands of lines of code, the AI intelligently identifies key areas of concern and suggests tests that maximize functional coverage.

Wang explained that this approach is inspired by AI techniques used in fields such as natural language processing and deep learning, where large datasets are analyzed efficiently by breaking them into smaller, more manageable components.

“In software engineering, AI has already proven its ability to write, analyze, and debug complex code,” Wang said. “We’re bringing those same principles to chip design, but with a focus on the unique challenges of hardware verification. By leveraging AI, we can help engineers reduce verification cycles from weeks to days.”

A Startup Betting on the Future of AI in Semiconductors

ChipAgents remains a small company, with a team of 10 engineers working out of its headquarters in Santa Barbara. Despite its size, the company is making a significant impact, drawing interest from semiconductor firms that see AI as the key to shorter design cycles and reduced verification bottlenecks.

As a pre-seed startup founded in 2024, ChipAgents is still in its early stages, but Wang is optimistic about its future. The company is exploring new funding opportunities and expanding its capabilities to support a broader range of design languages, including SystemC and high-level synthesis (HLS) languages.

Looking further ahead, Wang envisions AI playing a role not just in RTL design, but also in physical design, place-and-route, and even manufacturing optimization. “AI will touch every aspect of semiconductor development,” he said. “The tools engineers use today will look completely different five years from now.”

Beyond the Lab: A CEO with a Passion for AI and the Outdoors

Despite his intense focus on AI and chip design, Wang maintains a passion for the outdoors. Originally from Pittsburgh, where he completed his PhD at Carnegie Mellon University, he moved to Santa Barbara in 2016 and quickly took up fishing as a hobby.

“There’s something very calming about going out to the pier and spending a few hours fishing,” he said. “It’s a great way to clear my mind after a long day of work.”

And when asked the age-old question—cats or dogs?—Wang laughed and responded without hesitation. “I’m a dog person,” he said. “I have an eight-year-old golden retriever, and he’s an essential part of my daily routine.”

The Road Ahead

As AI continues to reshape industries, ChipAgents is positioning itself at the forefront of AI-powered semiconductor design. With record-setting performance on industry benchmarks, deep expertise in machine learning, and an ambitious roadmap for the future, the company is demonstrating that AI is not just an enhancement to chip design—it’s a necessity.

For those interested in seeing AI in action, the full interview with William Wang is available on YouTube.

Tags: AI, Automation, Chip Design, EDA, Machine Learning, RTL, Semiconductors, Verification

Categories: ChipAgents, Video Interview

This entry was posted on Saturday, March 8th, 2025 at 6:19 pm. You can follow any responses to this entry through the RSS 2.0 feed. You can skip to the end and leave a response. Pinging is currently not allowed.