Anna Goldie and Azalia Mirhoseini [pictured] are building software to automate the creation of cutting-edge silicon layouts. This move could let any outfit craft its own bespoke chips in weeks rather than years.

The pair have turned the top floor of a quiet suburban home into Ricursive Intelligence’s HQ and talked venture capitalists into handing over $35 million to get things moving. The Sequoia Capital and Striker Venture Partners cash dump has already pushed the startup’s valuation to $750 million, which is a decent haul for a company operating above someone’s living room.

Big tech giants already roll their own silicon for specific jobs. Amazon and Google built custom processors for data-centre and AI work, while the Fruity Cargo Cult Apple saved billions by shifting its devices to in-house chips like the M-series that propped up its MacBook sales. Tailored silicon tends to be cheaper, smaller, and less power-hungry because it is designed for a narrow task.

Traditional chip development remains a slow and pricey slog involving architecture design, verification and manufacturing prep, which usually drags on for two to three years. A tiny blunder late in the process can wipe out timelines and budgets.

Mirhoseini, now a Stanford computer science professor, and Goldie reckon that if they crack full automation, any tech firm could spin up a fresh chip design in a few days.

Mirhoseini said: “We think there will be a proliferation of custom silicon."

Former AI lab talent continues to lure serious investor attention. Fresh ventures include Periodic Labs, from former OpenAI researcher Liam Fedus, and Reflection, from ex-Google DeepMind staff. Goldie said Ricursive saw interest from more than 50 venture capitalists.

Sequoia Capital partner Stephanie Zhan, who is joining the board, said: “They’re really the ones who created this entire category around AI chip design,” adding that the mission is to “unlock this whole new universe of people who can create new chips specific to their own applications,” she said.