Quesi raises $15M to build AI chips designed by AI agents
What's the deal? Quesi TechnologyDealroom has a profile for this one. Try Dealroom → has raised a new round of US$13.8M, the AI chip startup announced in September 2026. Yunqi PartnersDealroom has a profile for this one. Try Dealroom → joined the round, with existing backers Monolith (曦思资本)Dealroom has a profile for this one. Try Dealroom → and Fudan Kechuang continuing to invest.
What's the endgame? Quesi builds custom AI chips using what it calls Agentic EDA, which brings AI agents into the chip design process. It is working on what it describes as the world's first AI inference chip designed end-to-end by AI.
What's the money for? The company's next priority is to define, design, and tape out a 6nm dedicated inference chip alongside a customer. Target markets span phones, smart glasses, AI PCs, robots, smart driving, and cloud data centres.
Why now? Founder Pan Hongyang argues that AI models pass through three stages, and many have now matured past the exploratory phase into parameter tuning and low-cost deployment. General-purpose GPUs and NPUs suit early-stage models, he says, but later stages call for customer-specific chips built for the best power, performance, and area.
Pan, a Fudan University PhD, founded the company in 2025 while still pursuing his doctorate. His focus, he says, is on physical AI and edge inference — using AI to reshape not just the back end of chip design but the way chips are defined in the first place.
The signal: The past two years have seen a sharp rise in AI hardware startups, with wearables, AIoT, and security among the leading use cases. Quesi is betting that as those models mature, the market shifts from flexible general-purpose silicon toward custom inference chips — and that AI itself can design them faster.
Read more: m.sohu.com
Image credit: Generated with Gemini
