Sebastian Thrun spent years teaching machines to fly and drive. His newest venture asks them to design the hardware that does both. Dulo, co-founded with Forward founder Adrian Aoun, has collected $130M in seed money to build what the pair call an Engineering Superintelligence.
The thesis is narrow on purpose. Language models learned to write software first because code is cheap to test and instantly reversible. Physical parts are neither. A model proposing a bracket or a cooling channel has to respect tolerances, materials, manufacturability and failure modes that no amount of text pretraining conveys.
Dulo wants foundation models trained specifically for that constraint set, aimed at engineers working on complex hardware programmes rather than general knowledge workers.
The seed itself is a statement. Aoun disclosed the financing on LinkedIn, naming backers including Marc Benioff, John Doerr, Eric Schmidt, Ram Shriram and Yuri Milner, alongside Construct Capital, Lowercarbon Capital, BoxGroup, SV Angel, Haun Ventures and Scribble Ventures. Few startups at this stage have ever amassed that roster.
Thrun’s record explains the appetite. The Stanford professor built the winning car in the 2005 DARPA Grand Challenge, ran Google X, and took Udacity public. Aoun has shipped consumer robotics and healthcare hardware.
The size of the round also sets expectations. A $130M seed leaves little room for a research project that stalls, and hardware engineering is a field where AI pilots have a long history of looking impressive in a demo and unhelpful on a real bill of materials.
Still, capital has been moving steadily from digital productivity tools into physical design. If Dulo can shorten a hardware development cycle at all, the payback is measured in months of engineering payroll rather than keystrokes saved.
