Antioch raises $32M to bring software speed to physical AI development. Antioch Team·08 September 2026 Antioch has raised $32 million to bring software speed to physical AI development. Together with the $8.5 million seed round announced earlier this year, this brings its total funding raised to $40.5 million. Antioch is also welcoming Greylock General Partner Saam Motamedi to Antioch's board. Saam and the Greylock team share its belief that the next leap in physical AI will come not only from better models, but from faster ways to develop, evaluate, and improve the complete systems in which those models operate. Antioch couldn't be more excited to be working with Saam and the wider Greylock team. "Software-speed innovation in the physical world will be the foundation of the next industrial revolution. We believe Antioch is building the core development platform for this next phase of physical AI." - Saam Motamedi, General Partner, Greylock Develop physical AI at the speed of software. Software improves quickly because it has a tight feedback loop. An engineer - or increasingly, an AI agent - can make a change, run tests, inspect failures, and try again hundreds or thousands of times before anything reaches a user. In aggregate, this results in superhuman development speeds. Physical AI does not yet have an equivalent. A robotics team cannot rerun every field test after every change, cheaply reproduce each failure seen in deployment, or test thousands of combinations of weather, lighting, terrain, hardware, and behavior. Instead, teams rely on physical test sites, recorded datasets, hardware-in-the-loop rigs, and manually operated fleets. These tools are necessary, but slow, expensive, difficult to parallelize, and limited in coverage. As a result, the pace of physical AI development is still constrained by the speed of the physical world. Antioch is changing this. Antioch is building the verifier. Antioch is building the verifier for physical AI: a development environment that predicts whether a proposed change will improve a physical system before it reaches hardware. A useful verifier must be inexpensive enough to run continuously, scalable to thousands of parallel scenarios, and accurate enough to guide real engineering decisions. That requires much more than a physics engine, or a world model. Antioch integrates a customer's hardware, sensors, software, models, environments, and operating constraints into a continuously calibrated simulation and evaluation environment. Teams can create scenarios, run evaluations at cloud scale, inspect failures and regressions, and use real-world outcomes to keep the simulator accurate. This enables continuous - and increasingly automated - physical AI development. An engineer or agent should be able to change a perception model, planner, controller, sensor configuration, or mechanical design; test it across thousands of conditions; understand what improved and regressed; and decide what to try next. And now, they can. What this makes possible. Teams use Antioch daily for three key purposes: Continuous evaluation. Every change can be thoroughly tested against every edge case before it reaches hardware. Faster development. Engineers and agents can modify systems, inspect results, diagnose failures, and verify improvements without waiting for the next physical test program. Synthetic data generation. Teams can generate large datasets representing rare failures, unusual environments, new hardware configurations, and other cases that would be expensive, impractical, or dangerous to collect in the real world. From classical simulation to learned world models. Today, most approaches to simulating the physical world sit toward one of two extremes. At one end of the spectrum is classical simulation: explicit models built with 3D engines, physics solvers, and hand-defined sensor and system behavior. Classical simulation is controllable, interpretable, and usable without large real-world datasets. Teams can vary environments, hardware, and operating conditions without first collecting every example in the real world. However, its limitation is fidelity. Hand-built models struggle with the parts of the physical world that are hardest to specify: contact-rich manipulation, deformable objects, and the small interactions that arise in uncontrolled environments. At the other end of the spectrum are end-to-end learned world models: systems that learn directly from data how a physical system and its environment will respond to actions. Given enough representative data, learned models can capture behavior that would be extremely difficult to program by hand. Antioch believe this is the approach that will win. As data and compute grow, more of the simulator - and eventually the entire simulator - will be learned end to end. The practical limitation today is data. Physical data is expensive and slow to collect. Most of it reflects normal operation, while failures and edge cases are scarce. Data from the latest system version is scarcer still. The scenarios that matter most are often the hardest, most dangerous, or most expensive to gather. Classical simulation works with little data but cannot model everything. Fully learned models can capture far more complexity, but do not yet have the data required to work reliably across most real systems. Program what is known. Learn what is not. Antioch combines the strengths of both approaches. Antioch explicitly represent what is known: geometry, kinematics, hardware specifications, sensor layouts, and physical constraints. Antioch use real-world data to learn what is harder to describe: sensor behavior, physical effects that explicit models miss, complex interactions, and the gap between the nominal model and the real system. In practice, this produces high-fidelity simulations that its customers use to guide real engineering decisions today. As more data becomes available, the boundary moves. Learned components replace explicit approximations, and larger parts of the simulator can be trained together. Its platform integrates the real-to-sim-to-real loop, continuously improving and retraining the simulator - improving fidelity, and driving towards an end-to-end learned world model. The destination is fully learned. The practical path to it is hybrid, and Antioch is driving the frontier. Already accelerating physical AI teams. Antioch is already being used by many leaders in Physical AI to reduce cost and improve evaluation coverage, accelerate development, and train cutting edge models across automated manufacturing, intelligent perception systems, drones, autonomous vehicles, robots, and more. Antioch is proud to be partnering with Amazon to accelerate the development across many product lines: "Antioch's simulations have closely matched our physical test results, including in scenarios we deliberately held out of calibration. That confidence lets us move more testing and development into simulation, reducing reliance on costly physical test programs, accelerating engineering cycles, and allowing our teams to focus on delivering better products for customers." - Jason Mitura, VP of Software Development at Amazon and Chief Product Officer of Ring Launchpad Build AI is using Antioch to accelerate the development of AI-powered manufacturing systems: "With Antioch, we have been able to meaningfully accelerate our time to market. We can test all of the scenarios that our systems will see in production, and edge cases that are otherwise impractical or uneconomical. This is critical for a company like ours, with teams on each side of the Atlantic. Simulation isn't new, but simulation with this kind of impact is." - Jon Quick, CEO, Launchpad Build AI These are the sentiments Antioch has seen repeatedly this year, and the impact that Antioch is driving at scale across industries. Building with the physical AI ecosystem. Delivering this impact at scale requires deep integration with the rest of the physical AI stack. Simulation engines and cloud infrastructure provide important foundations. Antioch turns them into a customer-specific development system by integrating complete hardware and software stacks, managing scenarios and evaluations, analyzing failures and regressions, and continuously closing the sim-to-real gap. Antioch is partnering with NVIDIA and integrating with its open physical AI stack, including NVIDIA Omniverse libraries, NVIDIA Isaac Sim, and NVIDIA Isaac Lab. These technologies help developers build complex physical systems, test them across large sets of simulated scenarios, and scale robotics development in the cloud. Antioch is also working closely with Nebius to make high-performance simulation infrastructure available at the scale required for continuous physical AI development. "Simulation has the potential to be a massive unlock for physical AI, but it remains one of the hardest parts of the stack to get right. Antioch is the best team we've seen at solving the problems that matter most, including scene creation, eval analysis, and most importantly, closing the sim-to-real gap. Together with Nebius infrastructure, they're turning simulation from a highly specialized capability into something physical AI teams can use continuously to develop, test, and improve their systems." - Evan Helda, Head of Physical AI, Nebius Join Antioch. Antioch is building the development infrastructure needed to automate the physical world - the defining economic opportunity of its time. This funding will help Antioch expand the platform, bring Antioch to more physical AI teams, and grow the team building it. Antioch is hiring across simulation, machine learning, infrastructure, graphics, go-to-market, and operations. Antioch Team

Fundraise receipts - september 8th: Antioch robotics, Tazapay, Poseidon Aerospace, herdr. tl;dr: frontrun flagged Antioch on april 17 2026 at 124 followers, 144 days before the company announced a $32M series A led by Greylock, with A*, Category Ventures, BoxGroup and Icehouse Ventures. founders Harry Mellsop, Michael Calvey, Colton Swingle and Collin Schlager. frontrun flagged Tazapay 212 days before Circle announced its acquisition of the cross-border payments platform. frontrun flagged Poseidon Aerospace on april 25 2026, 136 days before the company announced a $60M series A led by TQ Ventures at a $350M valuation, with JAWS, Hanwha, G Squared, Starship Ventures, Drover Ventures and Draper. frontrun flagged herdr on may 27 2026, 104 days before the company announced a $6M seed led by Bessemer Venture Partners, with Y Combinator, e2vc and angels Tobi Lütke, Dane Knecht and Görkem Yurtseven. founder Can Celik. Antioch robotics: $32M series A, flagged 144 days early. FUNDRAISE |: @frontrunvc flagged @antiochrobotics on April 17th, and was #1 on its "12 robotics companies under 1,000 followers" list on June 11th. 144 days later: $32M Series A led by @GreylockVC, with @A_StarVC, @Category_VC, @BoxGroup and @IcehouseVenture. 12 robotics companies you've never heard of under 1,000 followers. 1. Antioch (@antiochrobotics) - 486 followers frontrun flagged Antioch on april 17 2026, at 124 followers. on june 11 it was #1 on its list of 12 robotics companies under 1,000 followers, at 486 followers. on september 8 2026, 144 days after the flag, Antioch announced a $32M series A led by Greylock, with A*, Category Ventures, BoxGroup, Icehouse Ventures and angels. by then 17 tracked accounts were following it. Antioch builds cloud simulation for physical AI. teams onboard a robot, build high fidelity scenes and test autonomy in hours instead of weeks in the field. TechCrunch called it "the Cursor for physical AI." customers include Amazon Ring, NVIDIA Robotics and Nebius. the series A follows an $8.5M seed at a $60M valuation from A* and Category Ventures in april 2026, with Palantir CTO Shyam Sankar as an angel. the founders are Harry Mellsop, CEO (@HarryMellsop, ex Tesla Autopilot), Michael Calvey (@michaeljcalvey), Colton Swingle (@colton_swingle, ex Google DeepMind) and Collin Schlager (@CollinSchlager, ex Meta Reality Labs). they are in the follow up post on X. follow from the front. frontrun.vc Tazapay: acquired by Circle, flagged 212 days early. ACQUIRED |: @frontrunvc flagged @tazapay on February 8th. 212 days later: acquired by @circle frontrun flagged Tazapay on february 8 2026, at around 800 followers. on september 8 2026, 212 days later, Circle announced it is acquiring Tazapay. no terms were disclosed. 4 tracked accounts followed Tazapay before the news. Tazapay is cross border payments infrastructure: collect, hold and pay out across 100+ markets, with $25B in annualized volume and about 60% of it already moving in stablecoins. that is why Circle wanted it. follow from the front. frontrun.vc Poseidon Aerospace: $60M series A, flagged 136 days early. FUNDRAISE |: @frontrunvc flagged @poseidonaero on April 25th. 136 days later: $60M Series A led by @TQVentures. Autonomous cargo planes, designed and built from the ground up. frontrun flagged Poseidon Aerospace on april 25 2026. on september 8 2026, 136 days later, the company announced a $60M series A led by TQ Ventures, with JAWS, Hanwha, G Squared, Starship Ventures, Drover Ventures and Draper, at a $350M valuation per the company's announcement video. 8 tracked accounts were following it at announcement. Poseidon Aerospace builds autonomous cargo aircraft from the ground up: the Egret, an unmanned fixed wing, and the Heron, an unmanned seaplane. the goal is lowering cost per flight ton mile so goods move faster and cheaper, anywhere they are needed. the round lands ahead of the company's first pilotless test flight and ten months after its $11M raise in november 2025. founder and CEO: David Zagaynov (@DavidZagaynov). follow from the front. frontrun.vc herdr: $6M seed, flagged 104 days early. frontrun flagged herdr on may 27 2026. on september 8 2026, 104 days later, herdr announced a $6M seed led by Bessemer Venture Partners, with Y Combinator and e2vc, and angels Tobi Lütke (CEO, Shopify), Dane Knecht (CTO, Cloudflare) and Görkem Yurtseven (co-founder, fal). 20 tracked accounts were following it at announcement, and Bessemer's own X account followed herdr on september 1, seven days before it led the round. herdr (@herdrdev) is the runtime your coding agents live on: an open, terminal native runtime for AI agents, written in Rust, built by Y Combinator founder Can Celik. the seed goes to hiring people who live in terminals. source: herdr raised a $6M seed. follow from the front. frontrun.vc these already happened. the useful list is the one where the round or the launch has not happened yet. frontrun watches the follow graphs of 2,000+ tracked venture investors on X and flags the companies they converge on, with a date, the morning it happens. see the signal this article is built on frontrun tracks the follow graphs of 2,000+ venture investors and flags companies months before rounds are announced. companies and founders surfaced on frontrun. curated, every week.

Antioch announces $32M Series A to bring software speed to physical AI development. Enables physical AI teams to move from simulation to real-world deployment with confidence. By Robotics 24/7 Staff September 8, 2026 Antioch co-founders (L-R) Harry Mellsop, Alex Langshur, Michael Calvey and Collin Schlager, as the team secured a $32M Series A round. Stay up-to-date with news and resources you need to do your job. Research industry trends, compare companies and get weekly market intelligence with Robotics 24/7. Antioch, a simulation platform for physical AI, announced a $32 million Series A led by Greylock, with participation from A*, Category Ventures, Box Group, and Icehouse Ventures. The company said that it will use the funding to accelerate its product development, expand its engineering team and deepen its simulation capabilities as demand grows across robotics, drones, industrial automation, fixed perception and other physical AI applications. Antioch adds second funding round in 2026. Antioch said that it enables robotics and autonomy teams to develop, test and validate AI systems in high-fidelity simulation, scaling from a single experiment to thousands of parallel evaluations before real-world deployment. "AI has transformed the pace and possibilities of software development in the blink of an eye," said Harry Mellsop, co-founder of Antioch. "Now, physical AI is extending that same transformation into the real-world. Bringing these systems into reality requires a new approach to testing, powered by high-fidelity simulations that enable faster, safer, and more reliable deployment." Antioch said that it is already working with leading physical AI teams - including Amazon's Ring - helping customers accelerate development cycles, while dramatically increasing the volume of testing they can run before deploying systems in the real world. The company added that it has established partnerships with physical AI infrastructure players, including NVIDIA and Nebius. "Antioch is changing how we develop and validate complex systems by enabling us to rely on simulation with a level of confidence that wasn't previously possible", said Jason Mitura, VP of software development at Amazon and CPO at Ring. "Antioch's simulations have closely matched our physical test results, including in scenarios we deliberately held out of calibration. That confidence lets us move more testing and development into simulation, reducing reliance on costly physical test programs, accelerating engineering cycles, and allowing our teams to focus on delivering better products for customers". As AI changes, simulation platforms need to keep up. Antioch said that AI has fundamentally changed how software is built, compressing development cycles from weeks to days. But the company said that same acceleration has yet to meaningfully reach the physical world, despite the physical economy accounting for an estimated $50 trillion in annual economic activity. As investment in physical AI accelerates, Antioch said that physical validation has emerged as one of the industry's biggest bottlenecks. The company said that while simulation promises to accelerate development, today's platforms lack both the fidelity and scale needed to accurately predict real-world performance. "Simulation isn't new to the industry, but low fidelity, lack of scale, and tooling fragmentation have limited its role in the development process," said Alex Langshur, co-founder of Antioch. "Antioch is the first platform that takes autonomous systems from design to validation to mass deployment in a single simulation environment." Antioch said that it gives physical AI teams the trusted development loop to build, test and validate autonomous systems before deploying them in the real world. By calibrating high-fidelity simulations to each customer's hardware and running them at cloud scale, the company said that its platform enables enterprises to accelerate development, increase test coverage and bring new capabilities to production faster. "It's increasingly clear that physical AI and robotics are hitting a key inflection moment," said Saam Motamedi, general partner at Greylock. "As AI accelerates every industry, advantage comes from how teams can harness it to accelerate innovation cycles, and the teams that do this best will win. Software-speed innovation in the physical world will be the foundation of the next industrial revolution. We believe Antioch is building the core development platform for this next phase of physical AI." Latest in test. Latest in artificial intelligence. Latest robotics news.