Reuters reports ChipAgents raised $60 million after Nvidia tie-up expanded

Why it matters

ChipAgents is financing an independent AI layer for chip verification while entrenched EDA vendors deploy their own Nvidia-backed agents. Trusted results and distribution will determine whether Wang can break into production workflows.

William Wang (@WilliamWangNLP), the artificial intelligence professor behind ChipAgents, has raised $60 million to accelerate its use of AI agents in semiconductor design, Reuters reported on July 29th.

The reported financing comes less than six months after ChipAgents closed a $50 million Series A1 led by Matter Venture Partners, with Bessemer Venture Partners, Micron, MediaTek and Ericsson participating. ChipAgents said that February 17th round brought its total capital raised to $74 million. Reuters' new $60 million figure follows that financing, though ChipAgents has not attached a valuation to its recent rounds.

Wang founded Alpha Design AI, the corporate entity behind ChipAgents, in 2024 after building a career around language models and machine reasoning. He is a computer science professor and the Mellichamp Endowed Chair in Mind and Machine Intelligence at the University of California, Santa Barbara, where he directs the university's natural language processing group and founded its Center for Responsible Machine Learning.

His path into chip verification began with theorem proving rather than semiconductor software. In a 2025 interview with Bessemer, Wang traced the idea to his Carnegie Mellon University doctoral research in 2011, when he worked on formal methods and scalable inference. By 2024, he concluded that advances in agentic systems were ready to be applied to electronic design automation, the software layer engineers use to design and verify chips.

Wang also worked on Amazon Q at AWS Bedrock between 2022 and 2024, according to his ChipAgents biography. That combination of language-model research and formal reasoning underpins ChipAgents' pitch: semiconductor engineering requires agents that understand hardware intent, invoke established design tools and verify their own output rather than behave like general-purpose coding assistants.

Funding a domain model for chip engineering

ChipAgents is using the capital to build around Renoir, a domain-specialized large language model and agent system introduced on June 17th. Renoir is designed to read specifications and reason across register-transfer-level code, verification plans, assertions, logs, waveforms and outputs from electronic design automation tools.

The system can be deployed within customer-controlled infrastructure, an important requirement for semiconductor companies whose design files and debugging records contain closely guarded intellectual property. ChipAgents says Renoir was fine-tuned from an open-weight mixture-of-experts model using public semiconductor material and proprietary data generated through its internal training system.

On July 26th, ChipAgents expanded its collaboration with Nvidia to train and optimize Renoir. ChipAgents is using Nvidia's Megatron Core for distributed model training, NeMo Megatron Bridge for checkpoint conversion and Model Optimizer for quantization and inference efficiency.

Nvidia's role centers on AI software and accelerated computing. The relationship gives ChipAgents access to an established training and deployment stack as Wang attempts to move the product from agents that automate individual engineering tasks toward systems that can complete longer, closed-loop design and verification workflows.

That distinction is central to the amount of capital flowing into the category. Chip design agents need model training, semiconductor-specific data, integration with existing EDA products and extensive validation. A generated software function can be revised after deployment. A design error that survives verification can force an expensive chip respin and delay production.

ChipAgents said in February that its platform had been deployed at 80 semiconductor companies and that annual recurring revenue had grown 140-fold year over year. Its current LinkedIn profile says the product has been used at more than 120 semiconductor companies. Those adoption and growth figures are supplied by ChipAgents, which has not published corresponding revenue dollars.

The performance numbers are also company measurements. ChipAgents has claimed its agents can accelerate specification comprehension by 15 times, formal assertion generation by 240 times and Universal Verification Methodology environment generation by 400 times. The commercial test will be whether those task-level gains persist across full production programs, where engineers must manage proprietary toolchains, large codebases and signoff requirements.

Nvidia is working across the field

ChipAgents is entering a market controlled by EDA vendors with deep customer relationships and products embedded throughout semiconductor development. Those incumbents are adding their own agent layers, often with Nvidia's technology.

Cadence launched ChipStack AI Super Agent in February to automate front-end chip design and verification. Cadence says the system can generate designs and testbenches, create verification plans, run regression tests and debug failures by orchestrating its existing EDA products.

Synopsys is building AgentEngineer with Nvidia as part of an open, hardware-accelerated agentic EDA stack. Siemens introduced Fuse EDA AI Agent in March and expanded its Nvidia collaboration on July 26th to support self-verifying workflows across semiconductor and printed circuit board design.

Nvidia therefore provides ChipAgents with infrastructure and industry validation without giving Wang an exclusive technical channel. ChipAgents must compete through its domain model, deployment options and ability to work across tool environments owned by larger vendors.

Startups are pressing the same opening. Cognichip raised a $60 million Series A in April for a physics-informed foundation model for chip design. The competing approaches vary, but the financing thesis is consistent: growing chip complexity and a limited supply of experienced engineers have made design productivity an economic constraint for semiconductor companies.

Wang has surrounded ChipAgents with executives from the industry it is trying to change. Its advisers include former Mentor Graphics CEO Wally Rhines, former Synopsys chief technology officer Raul Camposano and former Cadence CEO Jack Harding. Matter founding managing partner Wen Hsieh joined the board after the February financing, while Marvell Data Center Group President Sandeep Bharathi joined the advisory board.

The reported $60 million raise gives Wang additional resources for a contest that will turn on verification and distribution. ChipAgents has to prove that an independent AI layer can earn the trust required to operate inside semiconductor programs while Cadence, Synopsys and Siemens bundle agents directly into the tools engineering teams already use.