Dulo Raises $130M Seed for Hardware-Design Foundation Models

Choosing how to manufacture a part can change the design itself. Dulo, the San Francisco startup co-founded by Sebastian Thrun and Adrian Aoun, is taking hardware-design foundation models into that demanding territory with a $130M Seed round confirmed by Aoun on October 29, 2026.

The company is developing models for hardware design, with its public site still carrying a coming-soon notice. Its financing brings substantial backing to an effort whose commercial importance will depend on the relationship between generated designs and the engineers expected to use them.

For engineering teams, the useful question is how a model helps them explore possibilities while respecting the requirements of an actual manufacturing process. Dulo has disclosed the ambition and the financing; the specific product, customer workflow and public performance evidence are still taking shape.

Dulo's seed financing and founding team

Adrian Aoun confirmed the $130M seed round in a public post that also identified Sebastian Thrun as his co-founder. Aoun said they started Dulo earlier in 2026 and described its focus as hardware-design foundation models. The announcement date is October 29; an exact legal closing date has not been established by the reviewed sources.

TokenPost's account names Eric Schmidt, John Doerr, Marc Benioff, Yuri Milner and Ram Shriram among the individual backers. It also identifies Construct Capital, Lowercarbon Capital, Bling Capital, BoxGroup, SV Angel, Haun Ventures and Scribble Ventures. The Information separately reported DST Global's participation. The reviewed sources identify neither a lead investor nor a valuation, and an individual's investment should not be silently attributed to an institution associated with that person.

Thrun's Stanford-hosted biography traces his work through the self-driving project that became Waymo, Google X and Google Brain. Those experiences give the founding story a particular direction: a researcher associated with machines acting in the world is now working on models concerned with designing hardware. That history explains the relevance of his move without supplying a performance result for the new company.

Hardware design begins with constraints

Dulo describes its team as drawing on leaders from Waymo, Google Brain and Stanford's AI laboratory, SAIL. Its public landing page states a hardware-design focus and a manufacturing-speed ambition, while leaving the implementation largely undisclosed. There is no verified basis here to say which CAD systems Dulo supports, what data trains its models, or which hardware categories it will address first.

The surrounding engineering discipline offers a concrete way to understand the assignment. Autodesk's generative-design documentation describes a process in which users specify geometric, performance and manufacturing requirements, then explore candidate designs. The person setting up that study has already made consequential choices about the problem the software should solve.

Autodesk's manufacturing-method guidance adds a practical complication: applying manufacturing constraints limits the shapes a system can generate and changes the resulting tradeoffs. An additive process, a milling operation and a casting process bring different requirements. These are examples of the category's existing demands, rather than claims about Dulo's architecture or any relationship with Autodesk.

That context gives engineering buyers a more useful frame than the size of the round alone. A model's contribution has to make sense inside the assignment the engineer actually has, including the method by which a design might become a manufactured object. Dulo's eventual product explanation will need to show where it enters that work and what it hands back to the person responsible for it.

The commercial choice inside the model ambition

For Dulo, a broad technical ambition leaves an important commercial choice open: which engineering job should a customer entrust to the company first? A buyer considering design exploration could require a different interface, evidence package and review process from a buyer evaluating a tool for an established production workflow. Those are possible buying situations, not announced Dulo product lines.

The distinction matters because the customer's time continues after software generates an output. Engineers have to understand what assumptions went into a candidate, judge whether it fits the intended task and decide what additional work it needs. The commercial opportunity lies partly in how much useful exploration the model enables relative to the evaluation it leaves behind.

DevCuration's coverage of Foundry Robotics' autonomous assembly business examines a different place in the hardware chain. That company concerns assembly, while Dulo's disclosed focus concerns design. Reading those assignments separately helps keep an expansive physical-AI narrative attached to the actual work each company says it is pursuing.

The financing gives Thrun and Aoun resources to pursue Dulo's stated direction. A detailed allocation to compute, hiring, expansion or product development has not been verified, so the round cannot be converted into a specific operating plan. The founders still have room to explain how their company expects to enter an engineer's existing routine.

What engineering buyers can follow from here

Dulo's current public materials provide a company identity, a founding team, a funding event and a development focus. In the sources reviewed, they leave customers, pricing, supported integrations, independent benchmarks and a release timetable undisclosed. A coming-soon page gives a reader a reason to follow the company while keeping the present state of the product clear.

Useful future disclosures would identify a particular design task, the requirements supplied to the model and the way an engineer can assess the result. Customer evidence could then explain how the software fits a working process. That information would make the founding ambition legible in the language of the person who has to choose a design and move it forward.

Thrun and Aoun have attracted substantial backing to this early stage of the work. As Dulo introduces its product, the consequential conversation will become more specific: which constraints belong in the model's assignment, which judgments stay with the engineer, and what a manufacturer receives when the design leaves the screen.