Deep|LLM: RSI and Chip Design: AI Lowers the Barrier to Custom ASICs, Redistributing Profit Across Accelerators, EDA and Design Services - FUNDA
NVDAAVGOGOOGMRVLSNPSCDNSSIEGY
Deep|LLM: RSI and Chip Design: AI Lowers the Barrier to Custom ASICs, Redistributing Profit Across Accelerators, EDA and Design Services
Bei, Andy·
Executive Summary
AI-assisted chip design is moving from an internal tool at frontier labs to an increasingly visible point of competition. According to a November 3 report in The Information, Anthropic told investors at pre-IPO meetings that its newly formed 15-person chip design team is already using Claude for custom-chip engineering and design. OpenAI’s Jalapeño inference chip, unveiled in June, went from initial design to tape-out in nine months, with the company’s own models accelerating parts of the design and optimization process. On the October 29 Cheeky Pint podcast, OpenAI CEO Sam Altman also said that taking a chip from design through to tape-out files is “a problem AI is very good at.”
Our earlier reports in the recursive self-improvement (RSI) series focused mainly on software: in September, Claude wrote more than 90% of Anthropic’s new code, and frontier labs had largely handed kernel and performance optimization to models. The market’s questions about AI chip design center on two issues: how far AI can go in chip design, given that EDA tools are slow and tape-out cycles are long, making the task very different from software development; and what this means for the cost of compute.
Continue reading with FUNDA
This report is available to subscribers. Sign in or subscribe to read the full analysis.