Antioch is a cloud simulation platform for teams building physical AI systems, including robotics, drones, industrial automation, and perception hardware. The product calibrates high-fidelity simulations to a customer's specific hardware and runs thousands of parallel test scenarios in the cloud, so autonomy teams can validate behavior in software before committing to expensive physical test programs. This launch marks a $32 million Series A led by Greylock, with participation from A*, Category Ventures, Box Group, and Icehouse Ventures, following an $8.5 million round earlier in 2026.
The company was founded in early 2025 by Harry Mellsop, Alex Langshur, Michael Calvey, and Collin Schlager, whose backgrounds include Tesla Autopilot, Google DeepMind, and Meta Reality Labs. Three of the co-founders previously built Transpose, a security and intelligence platform that was acquired. Antioch says it is already deployed with Amazon's Ring and has infrastructure partnerships with NVIDIA (integrating Omniverse, Isaac Sim, and Isaac Lab) and Nebius, positioning it inside the mainstream physical AI tooling stack.
The pitch to founders and operators is straightforward. AI has compressed software development cycles from weeks to days, but robotics teams still bottleneck on physical validation, and existing simulators tend to be fragmented or too low fidelity to trust for deployment decisions. Antioch is arguing that a single high-fidelity environment covering design, validation, and deployment can shift most of that testing burden into software, which is why the round is landing now as capital and enterprise interest flood into physical AI.
Comments (15)
Sign in to join the discussion.
N
Nikhil RaoSep 9, 2026
The jump cut from sim to real at 0:42 did more for me than the entire pitch deck probably does. Whoever edited this understood assignment.
M
Marta KowalczykSep 9, 2026
Tagline rewrite, on the house: 'Ship robots like you ship code.' You can Venmo me later.
D
Deji AkindeleSep 9, 2026
Every physical AI launch video has the same warehouse b-roll. At this point I feel like I've toured the same building fourteen times.
Y
Yuki T.Sep 9, 2026
'Speed of software' is doing a lot of heavy lifting when the bill of materials still has a 14-week lead time.
B
Bogdan PetrescuSep 9, 2026
Right, and 'developer velocity' means nothing when your compiler is a machine shop in Shenzhen waiting on tariffs.
C
Chen-Wei LiaoSep 9, 2026
Cutting the tweet mid-sentence with a t.co link is a bold engagement hack. Rude, but bold.
P
Priya M.Sep 9, 2026
Genuinely curious what the training loop looks like here. Are we talking teleop data, sim rollouts, or some frankenstein of both?
T
Tomas HolmSep 9, 2026
Any roles open for a robotics perception eng with 4 years at a humanoid startup? Got someone quietly looking, DMs open.
K
kaiaSep 9, 2026
Physical AI is the new phrase we all agreed to use so we don't have to say 'robots' and scare the LPs.
F
Fenna de VriesSep 9, 2026
'Robots' also implies they work. 'Physical AI' implies you'll get a demo in Q3.
O
Omar B.Sep 9, 2026
Reminds me a lot of the early thesis behind a couple of my portfolio bets in the manipulation space. Curious how the moat holds when the foundation model layer commoditizes.
L
Lior BentalSep 9, 2026
Every physical AI company has the exact same moat slide and it's always 'proprietary data flywheel'. At some point one of you has to actually mean it.
G
Grigor P.Sep 9, 2026
Would love a quick call before you go dark on press. Working on a piece about Series A patterns in embodied AI and you'd be a great voice.
S
Sanjay IyerSep 9, 2026
Everyone's celebrating the raise but nobody's talking about how you actually get these units into a customer facility, installed, and supported. That's where the wheels fall off.
A
Ada MensahSep 9, 2026
Building something adjacent in the sim-to-real space and honestly happy to see this get funded. Rising tide, though I hope your data pipeline is better than ours.
Work With Us
Book an intro call to explore our suite of services. We might have a waitlist.
Antioch
Develop physical AI at the speed of software
Develop physical AI at the speed of software
Tags
TLVC Rating
Community Rating
No ratings yet
to rate this launch.
Antioch is a cloud simulation platform for teams building physical AI systems, including robotics, drones, industrial automation, and perception hardware. The product calibrates high-fidelity simulations to a customer's specific hardware and runs thousands of parallel test scenarios in the cloud, so autonomy teams can validate behavior in software before committing to expensive physical test programs. This launch marks a $32 million Series A led by Greylock, with participation from A*, Category Ventures, Box Group, and Icehouse Ventures, following an $8.5 million round earlier in 2026. The company was founded in early 2025 by Harry Mellsop, Alex Langshur, Michael Calvey, and Collin Schlager, whose backgrounds include Tesla Autopilot, Google DeepMind, and Meta Reality Labs. Three of the co-founders previously built Transpose, a security and intelligence platform that was acquired. Antioch says it is already deployed with Amazon's Ring and has infrastructure partnerships with NVIDIA (integrating Omniverse, Isaac Sim, and Isaac Lab) and Nebius, positioning it inside the mainstream physical AI tooling stack. The pitch to founders and operators is straightforward. AI has compressed software development cycles from weeks to days, but robotics teams still bottleneck on physical validation, and existing simulators tend to be fragmented or too low fidelity to trust for deployment decisions. Antioch is arguing that a single high-fidelity environment covering design, validation, and deployment can shift most of that testing burden into software, which is why the round is landing now as capital and enterprise interest flood into physical AI.
Comments (15)
The jump cut from sim to real at 0:42 did more for me than the entire pitch deck probably does. Whoever edited this understood assignment.
Tagline rewrite, on the house: 'Ship robots like you ship code.' You can Venmo me later.
Every physical AI launch video has the same warehouse b-roll. At this point I feel like I've toured the same building fourteen times.
'Speed of software' is doing a lot of heavy lifting when the bill of materials still has a 14-week lead time.
Right, and 'developer velocity' means nothing when your compiler is a machine shop in Shenzhen waiting on tariffs.
Cutting the tweet mid-sentence with a t.co link is a bold engagement hack. Rude, but bold.
Genuinely curious what the training loop looks like here. Are we talking teleop data, sim rollouts, or some frankenstein of both?
Any roles open for a robotics perception eng with 4 years at a humanoid startup? Got someone quietly looking, DMs open.
Physical AI is the new phrase we all agreed to use so we don't have to say 'robots' and scare the LPs.
'Robots' also implies they work. 'Physical AI' implies you'll get a demo in Q3.
Reminds me a lot of the early thesis behind a couple of my portfolio bets in the manipulation space. Curious how the moat holds when the foundation model layer commoditizes.
Every physical AI company has the exact same moat slide and it's always 'proprietary data flywheel'. At some point one of you has to actually mean it.
Would love a quick call before you go dark on press. Working on a piece about Series A patterns in embodied AI and you'd be a great voice.
Everyone's celebrating the raise but nobody's talking about how you actually get these units into a customer facility, installed, and supported. That's where the wheels fall off.
Building something adjacent in the sim-to-real space and honestly happy to see this get funded. Rising tide, though I hope your data pipeline is better than ours.