Physical AI Simulation Startup Antioch Raises $32M Series A Led by Greylock

AI has compressed software development cycles from weeks to days. That same acceleration hasn’t meaningfully reached the physical world — every improvement to a robot or autonomous system still has to be validated on real hardware, and each test burns time, equipment and engineering resources. That’s a real bottleneck for an economy where physical activity accounts for an estimated $50 trillion a year. Simulation has long promised a way out, but existing platforms have lacked the fidelity and scale to reliably predict real-world performance.

Antioch, which is trying to close that gap, just raised a major follow-on round. The company’s simulation platform lets robotics and autonomy teams develop, test and validate AI systems in high-fidelity environments, scaling from a single experiment to thousands of parallel evaluations before real-world deployment. Founded in New York in May 2025 by second-time founders and alumni of Tesla and Meta Reality Labs, Antioch already counts Amazon’s Ring among its physical AI customers and has forged infrastructure partnerships with NVIDIA and Nebius.

The company previously raised a $4.25 million pre-seed in December 2025 and an $8.5 million seed in April 2026. Combined with the new round, Antioch has raised $44.75 million in just over a year.

Antioch announced a $32 million Series A led by Greylock, with existing investors A*, Category Ventures, Box Group and Icehouse Ventures returning. The company’s angel investors include Palantir CTO Shyam Sankar, Foxglove CEO Adrian Macneil (an angel since the seed round), and Ian Andrews, an NVIDIA executive and former Groq executive.

Antioch co-founder and CEO Harry Mellsop said AI has transformed the pace and possibilities of software development “in the blink of an eye,” and that physical AI is now extending that same transformation into the real world, which requires a new approach to testing powered by high-fidelity simulation for faster, safer and more reliable deployment. Co-founder Alex Langshur said simulation isn’t new to the industry, but low fidelity, lack of scale and tooling fragmentation have limited its role in development — calling Antioch “the first platform that takes autonomous systems from design to validation to mass deployment in a single simulation environment.”

Greylock general partner Saam Motamedi said it’s increasingly clear that physical AI and robotics are hitting a key inflection moment, and that as AI accelerates every industry, the advantage goes to teams that can harness it to speed up innovation cycles. He said software-speed innovation in the physical world will be the foundation of the next industrial revolution, and that Antioch is building the core development platform for that next phase.

Jason Mitura, VP of Software Development at Amazon and Ring CPO, said Antioch is changing how the company develops and validates complex systems by enabling it to rely on simulation with a level of confidence that wasn’t previously possible. He said Antioch’s simulations have closely matched Ring’s physical test results, including in scenarios deliberately held out of calibration — confidence that lets the team move more testing into simulation, reduce reliance on costly physical test programs, and accelerate engineering cycles.

Antioch is also integrating NVIDIA’s open physical AI stack, including NVIDIA Omniverse libraries and open robotics simulation and learning frameworks like Isaac Sim and Isaac Lab, to help developers scale testing across thousands of simulated scenarios. Evan Helda, Head of Physical AI at Nebius, said Antioch is the best team he’s seen at solving the problems that matter most in simulation — scene creation, eval analysis, and closing the sim-to-real gap — and that combined with Nebius’s infrastructure, they’re turning simulation into something physical AI teams can use continuously. The new funding will accelerate product development, expand Antioch’s engineering team, and deepen its simulation capabilities as demand grows across robotics, drones, industrial automation and fixed perception.

Where Antioch Sits Competitively

Antioch doesn’t build its own world models — it layers domain-specific tooling on top of general-purpose physics simulation from NVIDIA Cosmos and World Labs, making it more complementary than competitive with NVIDIA’s own infrastructure. Its more direct rivals are Israel’s Cognata, which focuses on autonomous-vehicle and ADAS validation and has raised $23.5 million through a Series B, and San Francisco-based Parallel Domain, which focuses on synthetic data for computer vision and has raised $44 million. Both are largely autonomous-vehicle-focused, unlike Antioch, which spans warehouse robots, agricultural machinery, drones and other domains at once.