Analysis
The Information reported Wednesday that Nvidia and ChipAgents are launching an evaluation for AI-powered chip design, according to The Information. The outlet's full reporting is paywalled, so the scope and participants of the evaluation have not been detailed publicly.
Pulse previously covered ChipAgents' $60 million Series A, led by B Capital, which the company used to expand its Nvidia partnership to train and optimize Renoir, its chip-design model, using Nvidia's Megatron Core for distributed training, the NeMo Megatron Bridge for checkpoint conversion, and Model Optimizer for quantization. ChipAgents CEO William Wang has declined to say whether Nvidia is an investor in the company, even as the technical collaboration has deepened.
“ChipAgents CEO William Wang has declined to say whether Nvidia is an investor in the company, even as the technical collaboration has deepened.”
What's changed since that July raise: the relationship has moved from model training infrastructure to a joint evaluation program, suggesting Nvidia is now testing ChipAgents' output against its own chip-design workflows rather than simply providing compute. ChipAgents says its agents can read a design specification, write verification tests, run them, and iterate without an engineer intervening at each step, and claims this cuts design cycles by more than 50% -- a company-reported figure that hasn't been independently verified. Reported funding totals for ChipAgents have varied across outlets, from $60 million to figures above $130 million depending on how prior rounds are counted, so any cumulative number should be treated as approximate until confirmed directly.
