TerraFirma raises $115M to build robotic construction company — starting with Tupperware robots demolishing buildings in Texas

TerraFirma raises $115M to automate construction — starting with Tupperware robots in Texas

Noah Schochet and his co-founder left SpaceX after watching construction timelines drag while building Starship infrastructure in Boca Chica. The contradiction stuck: a rocket the size of a skyscraper built at roughly one per month, while a road or factory takes years. TerraFirma is their answer — a full-stack robotic construction company that retrofits existing heavy equipment with autonomous software and operates directly as a contractor.

The hardware approach

TerraFirma takes Caterpillar, John Deere, and Komatsu machines and retrofits them. Everything above the iron — all software and manufacturing engineering systems — is built in-house, modeled on the factory operating software Schochet and his co-founder developed at SpaceX.

Zero to one

Twelve months ago the company was four people. The first robots were built from Tupperware containers, Raspberry Pi units, and Arduino kits from college. Schochet and his co-founder found their first customer in their own landlord, who offered them a building to demolish in Texas. They knocked it down, got paid, and validated that the model worked.

“TerraFirma was founded by my cofounder and I after we were working at SpaceX, and we were trying to build Starship rockets down in Boca Chica, and the construction was just too slow... We actually built a bunch of robots using Tupperware containers that our lunchbox came in and, like, Raspberry Pi and Arduino from our college kits. We got robots working and our landlord's like, hey, I got a building for you to demolish. My cofounder and I literally operated the robots ourselves — knocked down a couple buildings and got paid our first check.”

The autonomy target

Schochet frames the economics around two curves: the cost of achieving full autonomy rises exponentially as you approach 100%, while the labor productivity gains from each incremental machine added per operator diminish sharply. Going from one operator controlling one machine to two machines cuts labor in half; going from four machines to five only shifts the ratio from 25% to 20% of labor cost. He puts the crossover point at roughly one operator controlling three or four machines — about 75% autonomy — as the commercial sweet spot for construction.

Behind Schochet is a teleoperation center with 20 desks capable of running 80 machines simultaneously. The interaction model is closer to Starcraft than traditional crane operation: an operator assigns a task to an excavator, it runs autonomously for 20 minutes, then the operator queues the next machine.

On permitting

Schochet pushes back on permitting as the binding constraint. His argument is that construction currently operates at roughly 25% utilization — nine to five, five days a week — and autonomous machines could run 168 hours a week at full utilization. That alone would make construction four times faster before permits become the bottleneck. His approach mirrors how SpaceX treated regulatory approval: get everything else ready, make the bureaucrat's job easy, and approvals follow.

Materials and design

Schochet expects the shift to robotics to force redesigns well beyond software. Drawing on SpaceX's experience with Starlink — where the piece parts had to be redesigned across versions to enable robotic assembly at scale — he expects changes to pipe connections, concrete chemical formulations, welding methods, and eventually building codes themselves.

The round

TerraFirma has raised $115M in total, with a $100M Series A led by Kleiner Perkins. Bain Capital and Magnetar Capital are among the co-investors.

The company's near-term target is 75% machine autonomy across earthworks. The five-to-ten-year goal, in Schochet's framing, is a SimCity-style interface where a user specifies what they want built and a robot fleet executes it — a timeline he says coincides with credible Mars construction ambitions.