Sanctuary AI’s humanoid robot, Phoenix Gen 8.
John Koetsier
Earlier this summer, Sanctuary AI announced a 99.5%-plus success rate achieved at a 2.54-second cycle time on a complex wire-plugging task for a global Tier 1 automotive supplier. The task required inserting flexible wired plugs into a moving target on a live conveyor belt, executed by two disembodied hands reaching for, rotating, and inserting the wires into moving automotive targets—with no humanoid robot in sight.
Essentially, Sanctuary, which developed both a body and a brain, has discovered that its Physical AI—paired with robotic hands—functions as a highly effective industrial worker.
This complex task perfectly aligns with the insights of former CEO Geordie Rose, who stated that “half of the complexity of the robot is in the hands.” It also represents a strategic expansion rather than a mere pivot: rather than waiting for humanoid hardware to reach mass commercialization, Sanctuary is deploying its Physical AI within existing commercial industrial processes and robots. This hardware-agnostic approach accelerates current industrial adoption while laying the foundation for the next generation of intelligent robotic systems, including industrial humanoids. The timing is particularly relevant, as Meta is actively exploring similar robotic applications for its data centers.
“META IS TESTING robots that can plug in cables, reset servers, and handle other tasks inside its data centers, according to several current and former workers familiar with the projects,” a recent Wired story reported.
This mirrors the work at Sanctuary, prompting a deeper look into the company’s strategy under new CEO Daniel Friedmann, formerly of space robotics firm MDA and clean energy company Carbon Engineering.
Although Phoenix, Sanctuary’s full humanoid robot, remains on the company website—now featuring a Generation 8 platform with enhanced manufacturability and sensor data—the homepage has shifted its primary focus to “physical AI for industrial automation.” Visuals prominently highlight arms, hands, and grippers integrated into production lines.
Is this a shift away from humanoids? A side quest? Or an introductory offering for a broader suite of automation tools?
I put several questions to Friedmann.
John Koetsier: How was the 99.5% success rate measured?
Daniel Friedmann: Success was calculated by tallying successful insertions against unsuccessful attempts.
John Koetsier: Over how many production cycles was that performance achieved?
Daniel Friedmann: The test spanned 40 minutes and included 313 plug insertion trials.
John Koetsier: How does the 2.54-second cycle time compare to human workers and existing automation?
Daniel Friedmann: This task cannot be automated with traditional methods. The 2.54-second cycle time was benchmarked against the customer’s existing line performance.
John Koetsier: Is this a strategic shift away from humanoids?
Daniel Friedmann: This represents an expanded approach to Physical AI to address critical labor challenges. Our hardware-agnostic capabilities expedite industrial adoption, while building the foundation that will support the next generation of intelligent robotic systems, including industrial humanoids.
John Koetsier: Why deploy your AI on industrial robots instead of the Phoenix humanoid?
Daniel Friedmann: Rather than waiting for humanoid hardware to reach mass commercialization, we are deploying Physical AI on existing commercial platforms today. This delivers production-ready performance to customers now, while advancing toward the next generation of intelligent robotic systems, including industrial humanoids.
John Koetsier: What did building humanoids teach you that enabled this deployment?
Daniel Friedmann: Global industrial leaders face unprecedented labor shortages and rising operational costs. When solving these challenges, function matters more than form.
John Koetsier: Do you consider Sanctuary primarily an AI company or a robotics company?
Daniel Friedmann: Sanctuary AI is a full-stack company developing both Physical AI and robotic hardware. We need access to top-tier hardware to train our Physical AI with high-quality data, and we require highly capable Physical AI to utilize that hardware to its fullest extent. The relationship between hardware and Physical AI is symbiotic.
John Koetsier: Is this deployment already in production?
Daniel Friedmann: No, this is currently at the proof-of-concept stage.
John Koetsier: How broadly can this system be deployed across manufacturing tasks?
Daniel Friedmann: Our Physical AI is designed to handle contact-rich, dexterity-intensive tasks that have historically exceeded the capabilities of traditional automation. For companies in manufacturing, logistics, and other labor-constrained industries, it is deployable on existing robotic hardware today, offering production-ready performance and a clear path to future intelligent industrial systems.
John Koetsier: How long does it take a customer to deploy the technology?
Daniel Friedmann: Our goal is deployment within a few weeks, though timelines currently vary based on task complexity.
John Koetsier: What is the typical ROI for customers?
Daniel Friedmann: We are focused on maintaining customer throughput at a lower cost, though it is too early to share specific metrics.
John Koetsier: Does the system work with existing robot hardware?
Daniel Friedmann: Yes. Our Physical AI operates on currently available industrial robots using both off-the-shelf end effectors and custom ones designed by our team. We also intend to support next-generation intelligent robotic systems as they enter the market, including industrial humanoids.
John Koetsier: Thank you for your time.
