Sebastian Thrun, the computer scientist behind Google's self-driving car project, has raised $130 million in seed funding for Dulo, a stealth startup building artificial intelligence models for engineering and hardware design. The Information first reported the financing on October 29, citing a person familiar with the transaction, in what ranks among the largest seed rounds ever raised by a newly founded company.
Dulo was founded in San Francisco earlier in 2026 by Thrun and co-founder Adrian Aoun. The company is developing foundation models that can automate engineering design and verification, allowing a user to enter parameters and ask the model to produce a feasible design for a physical product such as an airplane or a car, while other companies handle manufacturing. Dulo describes its mission as enabling manufacturing at lightspeed.
Backers bet on the founding team
The round drew former Google chief executive Eric Schmidt and Kleiner Perkins chairman John Doerr, alongside venture firms DST Global, Construct Capital and Lowercarbon Capital. The Information's report names Schmidt, Doerr and Kleiner Perkins among the investors, while other coverage of the round also lists DST Global, Construct Capital and Lowercarbon Capital as participants. No lead investor or valuation has been disclosed.
The size of the round is the story in itself. At $130 million, the seed financing is unusually large for a company founded this year, and it reflects substantial investor conviction in Thrun's track record rather than in a shipped product. Dulo has not publicly disclosed detailed product specifications, commercial customers, revenue or independent performance benchmarks, and its technical capabilities are not yet fully visible outside the company.
A veteran founder returns to frontier work
Thrun founded Google X, the company's moonshot laboratory, co-founded Google Brain, and launched the self-driving project that later became Waymo. He is also a Stanford professor whose earlier work helped define modern autonomous vehicles and online education. Dulo surfaced in August with a roster of Waymo, Google Brain and Stanford robotics alumni, but with no disclosed funding, product or timeline until this week's report supplied the first dollar figure attached to the project.
The bet is that the same foundation-model techniques that transformed software and language can be extended to the physical world, where designing and verifying hardware remains slow, expensive and expert-driven. If Dulo's models can reliably generate designs that pass engineering checks, the payoff could be shorter development cycles for everything from vehicles to industrial equipment. The open question is whether the models can improve real engineering workflows enough to earn manufacturing adoption, a bar that demands far more precision than generating text or images.
A crowded race for engineering AI
Dulo is not alone in chasing the idea. Prometheus is developing AI for engineering and manufacturing, and Dynatomics is building a large language model for hardware engineering, among others. Dealroom noted that Dulo's $130 million seed ranks above the 99th percentile for all-time seed rounds in United States robotics by funding amount, and one analysis found it larger than five other AI seed rounds recorded between October 21 and 29 combined.
For now, Dulo remains an early-stage investment with more questions than answers: no product launch date, no named customers and no public demonstration of its models. But with $130 million in the bank and some of Silicon Valley's best-known names on its cap table, Thrun's latest venture has bought itself the time and talent to find out whether AI can design the machines of the future as fluently as it now writes their software.
