Sanctuary AI Is Moving Around the Humanoid Bottleneck
Sanctuary’s latest proof point is a 2.54-second wire-insertion task on existing industrial robot hardware

Sanctuary AI, the Vancouver-based robotics company founded in 2018, says its Physical AI software achieved a 99.5%+ success rate at a 2.54-second cycle time on a plug-insertion task for a global Tier 1 automotive supplier, manipulating a flexible wire that shifts dynamically while moving on a conveyor.
The system matched the throughput of the customer's live production line although the customer has not been named. Sanctuary's release shows the task running on two industrial robotic arms, not on Phoenix, its own humanoid platform.
Sanctuary frames this as a deliberate strategy shift, rather than waiting for humanoid hardware to reach mass commercialisation, the company is deploying its Physical AI software on existing commercial robotic platforms now, while building toward the same software running on industrial humanoids later. Sanctuary's updated materials reflects that framing, describing the company around deploying production-ready Physical AI across existing and future robotic hardware, broader than the Phoenix-centred positioning that defined the company through 2023 and 2024, when Phoenix reached its seventh and then eighth generation.
The company's industrial product line integrates its Physical AI with FANUC and Universal Robots arms, paired with custom or off-the-shelf end effectors. Sanctuary positions this as the faster commercial path: traditional fixed automation is too rigid for dynamic, dexterous work, while industrial humanoids are still years from enterprise reliability by the company's own estimate. For this specific wire plug-insertion task, Sanctuary says the policy reached its 99.5%+ performance level using 5.5 hours of teleoperation data.
The software groundwork for this predates the announcement. Sanctuary's Carbon control system and its Microsoft collaboration on AI models for general-purpose robots gave the company a model layer trained on behavioral data rather than tied to one robot body. That work is what is now being pointed at third-party industrial arms instead of, or alongside, Phoenix.
Sanctuary's other major technical investment is in its own hydraulic hands, a distinct path compared to most humanoid hand approaches, which tend to favour electromechanical actuation since hydraulics typically add weight, complexity, and leak risk. Sanctuary has argued the tradeoff is worth it for power density and force control in fine manipulation. The company has also shown the hands mounted on industrial arms rather than only on Phoenix. Co-founder and CTO Olivia Norton has described the hands, with up to 21 degrees of freedom in the latest generation, as achieving in-hand manipulation through reinforcement learning and sim-to-real transfer, with tactile sensors added for richer touch feedback.
Sanctuary had existing enterprise relationships ahead of this release, including ties to Magna, Accenture, and Microsoft, and had raised more than $140M in total investment by mid-2024 from backers including BDC Capital and InBC.
Sanctuary already had the humanoid platform, the manipulation history and the industrial relationships; the new piece is where it is choosing to prove the work first.
Have a robotics update Korthos should review? Send news, deployments, product releases, funding rounds, research, or media to tips@korthos.xyz or reach out on X at @agkorthos.
Regular Korthos briefings on robotics, drones, physical AI, supply chains, funding, product launches, and the companies shaping the stack.