From Simulation to Shop Floor: How Factories Learn to Trust AI Robots
As artificial intelligence moves from training data centers to manufacturing facilities, the challenge shifts from building capable machines to deploying systems that can adapt and learn alongside human workers.
Artificial intelligence is transitioning out of the controlled environment of data centers and into manufacturing plants where robots must contend with unpredictable conditions and shifting requirements. Yet conventional industrial robots remain confined to cages, locked into single repetitive tasks, and lack the flexibility needed for modern production demands. Today's manufacturers seek adaptable, general-purpose systems capable of handling multiple jobs and improving as factory operations change. The real difficulty lies in actually implementing these systems, according to Adrien Gaidon, co-founder and chief strategy officer at Walden Robotics Inc.
If you're changing things and you have machines that cannot change and continuously learn in deployment, then you become a burden. What we do is we do machines that continuously learn from deployment and adapt so customers can change their mind and improve the process and give them the superpower to change even more and faster.
Adrien Gaidon, Walden Robotics
Gaidon and John Mancuso, vice president of field engineering at CoreWeave Inc., shared insights during a conversation at Fully Connected, broadcast through theCUBE, SiliconANGLE Media's livestreaming platform. Their discussion centered on how cloud resources, virtual simulation and human support enable robots to develop capabilities and respond to changing conditions in actual factory settings.
Physical AI moves from data centers to the factory floor
Walden emerged from Toyota Research Institute in January 2026 and secured $300 million in funding by July to deploy general-purpose robotic systems in manufacturing environments. The company's robots have completed full production shifts working alongside human employees since May. They perform machine tending, parts kitting and subassembly—work that manufacturers either struggled to automate or never attempted, Gaidon noted.
There's half a million open positions for machinists in the U.S. because it's a hard job and it requires skills. But part of that job does not require the craft or the human creativity.
Adrien Gaidon, Walden Robotics
Preparing a robot for factory deployment demands substantial computational resources long before the machine arrives on the production line. CoreWeave, which has positioned itself as a supporter of physical AI, brings mechanical engineers and robotics specialists into customer projects alongside its computing infrastructure, Mancuso explained.
In the end, before a model is ever going to do something in the physical world, it's going to have a tremendous amount of simulation that happens in the virtual world. We're providing the infrastructure.
John Mancuso, CoreWeave
Although simulation within data centers can make a robot functional, it cannot guarantee flawless performance. Walden combines autonomous robots with remote human operators standing by to assist. This arrangement mirrors the setup of automated textile looms from a century ago, when a single operator managed 40 machines, Gaidon observed.
https://www.youtube.com/embed/y6VIAuF7H08?feature=oembed
Our robots are autonomous, but they know how to call for help. They know when they need some assistance – and that assistance feeds the most valuable data to improve over time.
Adrien Gaidon, Walden Robotics