Cadence Design Systems expects India to develop a semiconductor-startup ecosystem comparable in scale and activity to that of the United States within the next decade, as artificial-intelligence agents begin reducing the time required for complex chip-design and verification work. Paul Cunningham, Cadence’s senior vice-president and general manager for its System Verification Group, said India already occupies a central position across the company’s workforce, product development and customer base.
Approximately one-third of Cadence’s global workforce is based in India across several locations, Cunningham estimated. The country’s engineering teams contribute to product execution and road maps, while many of the company’s largest semiconductor customers also maintain substantial engineering operations in India. Cadence plans to continue growing its net workforce in the country, although hiring and selective restructuring may occur simultaneously as investment is redirected between functions.
Cadence has assigned hundreds of employees across its portfolio to build AI agents for chip design and verification. Its AgentStack strategy includes ChipStack for digital front-end design and verification, InnerStack for physical implementation and sign-off, Verisium for verification, and AuraStack for advanced packaging and printed-circuit boards.
ChipStack has reached its second major release. Cadence is engaged in about 25 substantial customer deployments involving groups ranging from tens to hundreds of users, with some customers using the technology in production programmes rather than limiting it to pilot projects. Adoption has not yet reached the thousands of users Cadence would classify as deployment at broad scale.
The company has recorded cases in which agents reduced five-week assignments to one week, five-day tasks to a day and day-long work to several hours. Applications include design creation, verification planning, test-bench generation and the production of formal-verification checkers. Cadence is retaining human supervision and its established non-agentic tools for critical approvals such as final chip sign-off. In advanced deployments, agents can operate autonomously for 12 to 24 hours before returning to an engineer, although many current use cases involve tasks lasting between five and 20 minutes.
Cadence’s latest financial performance provides the commercial backdrop to the expansion. Second-quarter revenue reached $1.584 billion, compared with $1.275 billion a year earlier. Core electronic-design-automation revenue grew 18%, system design and analysis revenue increased 37%, and the hardware business added 12 customers. The company ended the quarter with a record $8.1 billion backlog and raised its 2026 revenue forecast to between $6.26 billion and $6.34 billion.
Cunningham expects engineers eventually to supervise teams of tens or potentially hundreds of agents. The current deployment phase remains focused on making those systems reliable and predictable across different designs, with the quality of human instructions and certified verification tools remaining integral to the development process.
