← NewsAll
Cadence unveils AI agent to speed chip design
Summary
Cadence introduced ChipStack AI Super Agent to help automate parts of chip design, and the article says it can build a model of a design and speed some tasks by about ten times.
Content
Cadence Design Systems has launched an AI "agent" called ChipStack AI Super Agent to assist with designing complex computer chips. The announcement responds to a long-standing bottleneck: engineers can spend a large share of their time writing and testing the code-like descriptions that define a chip before it becomes physical silicon. The agent analyzes a chip's design, forms what Cadence describes as a "mental model" of how the chip should work, and can use Cadence tools to test designs and fix bugs. Cadence said the tool speeds some tasks by about ten times and is in early use by several companies.
Key details:
- The ChipStack AI Super Agent inspects a chip design at the circuit-description stage and interacts with existing Cadence tools to test and debug designs.
- The article reports engineering teams can spend up to 70% of their time on writing and testing code for chip designs, creating a productivity bottleneck.
- Cadence states the agent speeds some tasks by roughly tenfold and the article mentions Nvidia, Altera and startup Tenstorrent as early users.
- Paul Cunningham, Cadence's vice president and general manager of R&D, is quoted saying the company expects to shift toward renting virtual engineers by the end of the decade.
- An industry analyst quoted in the article said such AI productivity tools could be important in the broader technology competition between the U.S. and China, noting China is developing its own chip design tools.
Summary:
The agent is presented as a way to reduce manual coding and testing time in chip development and to change how engineering work is organised. The article reports early adoption by several firms and a stated company plan to move toward offering virtual engineering services; the timeline for broader rollout is undetermined at this time.
