Astrus raises $12M to automate chip design with reinforcement learning

Astrus, a Kitchener-Waterloo- and Toronto-based startup, has closed a $12-million USD Series A round led by Caffeinated Capital with support from Khosla Ventures to automate physical chip design using reinforcement learning. The company plans to launch its solution in October. Astrus targets analog designers who convert logic into physical manufacturing designs, a process currently done manually due to a lack of public data. By partnering with major foundries, the startup generates synthetic data through reinforcement learning, claiming it will soon possess more physical design data than the rest of the world combined. The funding will support scaling the team and a US go-to-market presence, though Moon says the company will remain in Canada, citing access to AI talent. Potential customers include major semiconductor manufacturers, with possible partnerships with logic design tool providers.