Training a robot in the real world is slow, costly, and occasionally destructive, which is why Antioch would rather you barely touch the real world at all. Greylock just wrote a $32 million check to back that bet. Let's check them out!
โพ๏ธ The Elevator Pitch
Software teams ship fast because they can test an idea in seconds. Robotics teams cannot. Every tweak to a drone, a forklift, or a humanoid has to be proven on real hardware, on real test tracks, one expensive run at a time. As Antioch's founders like to point out, the vast majority of the industry doesn't use simulation at all, so progress crawls along at the speed of physical reality. That is a painful bottleneck right as Goldman Sachs is forecasting millions of humanoid robots within a decade.
Enter Antioch. Founded in 2025 by five Stanford engineers who cut their teeth on Tesla Autopilot, Google DeepMind, and Meta Reality Labs, Antioch spins up high fidelity digital twins of a robot and its environment in the cloud, then runs thousands of test scenarios in parallel before anything touches hardware. Think of it as a flight simulator for any machine that moves. CEO Harry Mellsop puts the goal plainly: make simulation feel just like the real world from the perspective of your autonomous system. The pitch is landing, with a $32 million Series A led by Greylock in September 2026, on top of an earlier $8.5 million seed.
๐ The Drop Down
๐ Website: antioch.com
๐๏ธ Founded: 2025, New York City
๐ฅ Founders: Harry Mellsop (CEO, ex Tesla Autopilot), Alex Langshur, Michael Calvey (the three earlier sold Transpose to Chainalysis), Colton Swingle (ex DeepMind), and Collin Schlager (ex Meta Reality Labs)
๐ฐ Stage: Series A, $32M led by Greylock, with A*, Category Ventures, BoxGroup, and Icehouse Ventures; angels include Palantir CTO Shyam Sankar
๐ Traction: Amazon's Ring and Launchpad Build AI are customers; partnerships with NVIDIA and Nebius
๐งช Product: cloud digital twins that run thousands of robot test scenarios in parallel
๐ฎ Tech Trend: physical AI and robotics simulation
๐ฏ Target Market: drone, autonomous vehicle, industrial, and humanoid robotics teams
๐ Why We Like It
๐ The robot boom finally needs a proving ground: Robotics startups pulled in $9.4 billion in 2025, up 41% in a year, and Goldman Sachs just multiplied its humanoid forecast fivefold. All those machines need somewhere to practice, and the real world is too slow to keep up. Antioch is selling the picks and shovels for that gold rush.
๐ซ A team that has shipped autonomy and sold a company: The five cofounders met at Stanford and worked on Tesla Autopilot, Google DeepMind, and Meta Reality Labs. Three of them already built the data startup Transpose and sold it to Chainalysis in 2023, so this is a crew that has both built hard systems and delivered a return.
๐ฏ Customers who tested the claim: This is the reassuring part. Amazon Ring's VP of software says Antioch's simulations closely matched their physical test results, even on scenarios they deliberately held out of calibration. The challenge will be that Antioch builds on top of NVIDIA's own simulation stack while NVIDIA keeps expanding its first party tools, and no simulation ever fully closes the gap with messy reality. Staying the indispensable layer above the platform it depends on is the whole game.
๐ค Get Involved with Antioch
๐ฑ Book a demo if you build robots, drones, or autonomous systems
๐ฐ Read Forbes on the bet that robots don't need a billion dollars of real world data
๐ Read Antioch's own Series A announcement
๐ผ Join the team, they are hiring simulation, ML, and infrastructure engineers
๐ฆ Follow Antioch on X
