A Waterloo, Ont.-based startup that trains artificial intelligence to design computer chips has raised US$12 million in a Series A financing round, the company confirmed this week, as it tries to crack one of the most stubborn bottlenecks in global semiconductor manufacturing.
Astrus, which operates out of Kitchener-Waterloo and Toronto, builds AI systems that use reinforcement learning to automate what is known as the physical design phase of chipmaking, the process of turning a chip’s logical requirements into a manufacturable, three-dimensional layout, according to BetaKit, the Canadian technology publication that first reported the round Sept. 22.
The round was led by San Francisco-based Caffeinated Capital, with Khosla Ventures, Garage Capital, MVP Ventures and RiSC Capital also participating, BetaKit reported. It follows an earlier US$8-million financing Astrus raised in 2025, also backed by Khosla Ventures, bringing the company’s total funding to roughly $23 million, according to Dealroom, a venture capital data platform that tracks the deal.
Chief executive Brad Moon, a former satellite-sensor chip designer who co-founded the company with reinforcement-learning researchers Zeyi Wang and Kenny Young, said most AI chip-design tools on the market focus on logic design, leaving physical design, the most labour-intensive part of the process, largely untouched. “The physical design side is incredibly secretive. There is no public data,” Moon said, according to BetaKit, adding that firms such as Apple and Taiwan Semiconductor Manufacturing Co. closely guard that information. To get around the problem, he said Astrus partnered with chip foundries to obtain manufacturing rules, which it uses to generate synthetic data to train its AI models rather than relying on proprietary industry data.
The global integrated circuit market is worth more than US$800 billion a year, and demand for new chip capacity has surged alongside the buildout of AI data centres worldwide. Moon said Astrus expects to have “more physical design data than the rest of the world combined” within a year, and the company is targeting an October product launch.
Unlike many Canadian AI hardware startups that relocate to the United States to be closer to chip clients and capital, Astrus intends to stay headquartered in Canada. Moon pointed to the country’s academic research base as a reason, citing the University of Alberta’s Rich Sutton, a pioneer of reinforcement learning, and the University of Toronto’s Geoffrey Hinton, a foundational figure in deep learning. “We have pipelines to two of the most important AI foundational technologies in the world,” Moon said.
The financing lands amid a broader run of AI-related capital flowing into Canadian technology companies this year, including a $195-million federal investment in Toronto-based quantum computing firm Xanadu announced in August, and comes as other Canadian firms, such as Toronto’s Tailscale, report growing traction with the same Silicon Valley investors and customers driving the AI infrastructure boom. For Waterloo’s tech sector, long known for producing engineering talent that often ends up at U.S. firms, Astrus is a test case for whether a homegrown AI hardware company can scale globally without leaving home.
