ChipAgents is partnering with Nvidia to create an agentic AI platform to accelerate chip design

ChipAgents

A few months ago I wrote an article on Forbes about Cadence Design SuperAgents that front-end the company’s chip design tools to speed elements of the design workflow by up to 10X, according to the company. Now, several EDA AI startups have entered the game with similar claims, and each is quite different. Let’s take a look. (Cadence Design is a client of Cambrian-AI Research, the author’s firm.)

Specifically, three startups disclosed details about funding and their software at the Design Automation Conference (DAC) event: Agentrys, ChipAgents, and Cognichip. Many design teams use a combination of tools from the Electronic Design Automation (EDA) leaders, Cadence, Synopsis, and Siemens, and consequently an AI agent platform that allows the team to orchestrate design efforts across tool vendors could have some appeal.

EDA AI Startups Race To Automate Chip Design

The semiconductor industry may be entering a new software cycle, one in which value shifts from point-tool optimization toward autonomous workflow execution. Agentic AI for chip design is emerging as a distinct layer on top of traditional EDA, aimed at automating multi-step engineering tasks across RTL generation, verification, debugging, and design orchestration. And two of the three startups come to market with custom AI models that could help differentiate them from the EDA big boys.

Verification has become one of the biggest bottlenecks in modern chip programs, especially as 3D ICs, chiplet-based architectures, and software-defined systems increase the combinatorial complexity of validation. Siemens has framed the problem as a widening verification productivity gap, while industry discussions increasingly describe agentic verification as an orchestration layer across existing flows rather than a single AI feature.

MORE FOR YOU

The New EDA AI Startups

The new startup cohort is pursuing the opportunity from different angles.

1. Agentrys is building what it calls Agentic Design Automation, a platform centered on self-improving agents and agent-native toolchains for chip-design workflows.
2. Cognichip is taking a model-first approach with its physics-informed Artificial Chip Intelligence platform, arguing that semiconductor design requires reasoning grounded in both logic and physical behavior rather than an agentic AI wrapper around a generic large language model.
3. ChipAgents, meanwhile, has evolved beyond a verification copilot story into a vertically integrated platform with workflow automation and its own fine-tuned foundation model, Renoir. The company claims to have over 120 customers on board, including some household names.

The segmentation matters because the long-term economics of the market may not resemble classic EDA seat licensing. A recent investor-oriented market analysis argued that agentic EDA may create a new revenue layer tied to verified design throughput rather than just tool access. If that view proves right, the companies with the strongest claim on workflow completion, measurable cycle-time reduction, and customer-specific feedback loops could capture disproportionate value.

So, which EDA AI Startup Will Win? “It Depends…”

The near-term winners may be the companies that attack the highest-friction but most measurable tasks first. Regression triage, waveform search, assertion synthesis, testbench scaffolding, root-cause analysis, and design-spec interpretation all offer clearer ROI than a full autonomous tapeout narrative. That practical wedge explains why verification remains the most commercially attractive beachhead even for companies with much larger long-term ambitions.

The debate, then, is no longer whether AI belongs in chip design. It is where durable value will settle. If model quality dominates, Cognichip’s physics-informed approach could prove especially powerful. It is critical to realize that public RTL and verification data remain limited; much of the best training material is locked inside private design organizations, and customers increasingly want secure deployment models that keep sensitive IP behind their own firewalls. But if Cognichip can break this stranglehold, it could win big, or possibly become a takeover target. Note that Cadence purchased ChipStack last Fall to build their SuperAgents.

Although, if orchestration and enterprise control matter most, Agentrys may be well positioned. And, if vertically integrated companies that own both the workflow and the model stack deliver the best cost, accuracy, and deployment profile, ChipAgents’ positioning with Renoir becomes significantly more compelling.

And if risk management is the higher prioity, then the current EDA vendors should do very well. The incumbents are not standing still. Cadence is already positioning its agentic SuperAgents as a way to enable autonomous chip and system design workflows, and Synopsys has formalized “EDA Agentic AI” as a multi-agent approach to design automation. That incumbent push validates the market, but it also narrows the room for startups that offer only workflow layers without proprietary data, differentiated models, or deep integration into production toolchains. If the EDA leaders are already working on their own foundational models, which is likely, they still must deal with agentic AI access to other vendors’ tooling.

The Takeaway

The broader takeaway is that the next phase of competition will be defined less by demo quality and more by who can prove trusted automation inside production design teams, with measurable gains in verification throughput, engineering efficiency, and schedule compression.

Agentrys is pitching a broader agent-native design automation platform, Cognichip is positioning around a physics-informed foundation model for end-to-end chip design, and ChipAgents is more explicitly focused on RTL, verification, and root-cause analysis productivity. And the major EDA vendors have agentic platforms themselves, with the resources to help clients ensure successful adoption.

I was talking with the CEO of an AI semiconductor company the other day, and I asked him about his company’s view of agents in the design flow. I came away wondering if design teams would risk schedule by becoming distracted by a new shiny agentic product. But if that is the case, the hesitancy is unlikely to last long and the EDA AI startups have a shot. There is too much at stake to ignore the potential benefits of agentic AI and domain-specific models.

Disclosures: This article expresses the opinions of the author and is not to be taken as advice to purchase from or invest in the companies mentioned. My firm, Cambrian-AI Research, is fortunate to have many semiconductor firms as our clients, including Baya Systems BrainChip, Cadence, Cerebras Systems, D-Matrix, Flex, Groq, IBM, Infleqtion, Intel, Micron, NVIDIA, Qualcomm, SImA.ai, Synopsys, Taalas, Tenstorrent, Ventana Microsystems, and scores of investors. I have no investment positions in any of the companies mentioned in this article. For more information, please visit our website at https://cambrian-AI.com.