A Silicon Valley startup is making a bold move into the custom computer chip industry, announcing Thursday that it has secured $24 million in seed funding to develop AI-powered tools that could dramatically change how chips are designed.
The company, Architect Labs, is setting its sights on competing with chip industry giants Broadcom and Marvell, both of which currently help major cloud computing companies — including Amazon and Alphabet’s Google — design specialized chips for artificial intelligence and general computing. That custom chip business generates tens of billions of dollars in revenue and serves as an alternative to the powerful hardware made by Nvidia.
Right now, designing a custom chip is an expensive and time-consuming process, typically taking around two years and costing hundreds of millions of dollars in labor and research and development. Architect Labs wants to make that process faster and far less costly.
Co-founder Ebrahim Hussain told Reuters the company plans to work with both chip manufacturers looking to speed up their design workflows and software companies that might benefit from custom chips to make their applications run more efficiently.
“Their biggest problem today is not necessarily the backend execution or the layout,” Hussain said. “Their biggest thing is how can I take this workload that I want to deliver to the world, whether it be AI or robotics or anything like that, and how can I build the (chip) architecture.”
Hussain co-founded the company alongside Aaditya Subedi. The Palo Alto, California-based firm currently employs about 18 people, with staff divided between machine learning and hardware disciplines.
Subedi described the company’s broader vision as making chip design as widely accessible as Taiwan’s TSMC has made chip manufacturing.
The funding round was led by Kindred Ventures, with participation from TQ Ventures, Race Capital, and Together Fund. Google DeepMind Chief Scientist Jeff Dean, as well as executives from OpenAI and Nvidia, also contributed to the investment.
