WASHINGTON, Feb 10, 2026, 09:37 EST
- Cadence unveiled its ChipStack AI Super Agent, aiming to automate sections of chip design and verification, the company announced.
- Cadence claims certain tasks now run as much as 10 times faster, with early adopters including Nvidia, Altera, and Tenstorrent.
- Analysts noted that AI-powered productivity tools are emerging as a fresh battleground in the U.S.-China chip rivalry.
On Tuesday, Cadence Design Systems rolled out an AI “agent” aimed at accelerating chip design for clients like Nvidia. This move comes as companies push for quicker product cycles while facing a growing shortage of skilled workers. Reuters
The pitch is straightforward: modern chips have grown so complex that engineers dedicate a large chunk of their time to writing and testing code-like descriptions that define a circuit before it turns into physical silicon.
This is crucial now as the AI surge drives more companies to tackle advanced chip projects. Yet, design teams face heavier verification workloads—ensuring designs function correctly—without a boost in staffing.
Cadence unveiled its ChipStack AI Super Agent, featuring what it calls “agentic AI”—software that acts autonomously within workflows instead of waiting on human commands. This system can generate and verify designs from specs and high-level descriptions. Cadence claims productivity gains of up to 10x across tasks like writing designs and testbenches, developing test plans, running regression tests, debugging, and even auto-fixing issues. Stocktitan
“ChipStack marks a significant advance” in how Cadence integrates AI into chip design workflows, CEO Anirudh Devgan said, emphasizing that it allows scarce engineers to concentrate on more valuable tasks.
Paul Cunningham, head of research and development at Cadence, explained that the agent creates a “mental model” of the chip’s intended function before using Cadence tools to test the design and track down bugs.
The company stated the agent supports both cloud-based and on-premises “frontier” AI models — meaning the biggest and most advanced systems — and can interface with Nvidia’s Nemotron models as well as cloud-hosted options like OpenAI’s GPT, tailored to each customer’s setup.
Early feedback from customers and partners is in. Altera’s Arvind Vidyarthi reported the tool slashed verification effort by roughly tenfold in certain areas. Nvidia’s Timothy Costa added that the method mixes “mental models” with automated test plan creation, boosting productivity.
Qualcomm’s Paul Penzes mentioned the company is testing the agent with a wider audience and described the initial findings as “encouraging.” Meanwhile, Tenstorrent’s Daniel Cummings reported that a three-month trial cut verification time by as much as four times on certain blocks.
Cadence goes head-to-head with Synopsys, Siemens EDA, and Ansys in the electronic design automation software arena. The company also flagged rising competitors from China in its latest annual report. Cadence
Analyst Dave Altavilla from HotTech Vision and Analysis emphasized that productivity tools like these might become crucial as Washington clamps down on advanced chip tech and China pushes to develop its own design ecosystem. “You need that capability to compete,” he said.
Trust remains the biggest hurdle. AI-generated code and fixes risk slipping in subtle bugs, and chip teams often hesitate to delegate key verification tasks without solid guardrails, clear explainability, and strict “engineer-in-the-loop” oversight—particularly when projects are nearing tapeout, the final stage before manufacturing.
Cadence announced that the ChipStack AI Super Agent is now in early access. Cunningham added that the company aims to shift from just selling tools to essentially “renting” virtual engineers by later this decade.
