Jabil Outlines Humanoid Robot Path from Prototypes to Mass Production

While humanoid robots often attract attention through demo videos and fundraising, the quieter work of development and finding a value proposition continues. Jabil Inc., a design, supply chain, and manufacturing services provider, shared its perspective on how humanoids progress from prototypes to mass production.

Companies preparing to scale humanoid production include Tesla, 1X Technologies, XPeng, and UBTECH. Jabil works with robotics developers to bring products to market at scale.

Jabil, based in St. Petersburg, Florida, has worked with Apptronik to bring its Apollo humanoid into active manufacturing environments. Thomas Brown, Jabil's senior director for engineering and technology, including digital commerce and robotics, answered questions from The Robot Report.

The discussion began with investor and customer expectations for mobile manipulators and humanoids.

Brown said many customers have deployed humanoids in low volumes, but the industry is entering a phase of tens of thousands. At that scale, hand assembly on site no longer works; companies need cost structures, manufacturability, and better quality. Jabil can add value there, he said.

When robots are expensive, production is limited for AI training data and customer validation. Jabil helps cost down designs to reach scale.

Compute is now sufficient for many demo experiences, Brown said, but more edge computing capacity is needed. NVIDIA Jetson is popular among Jabil customers developing robotics.

Compute and memory are increasing, but prices remain a pain point for scaling. Foundation model makers are developing intelligence beyond current capacity; expanding it is key.

Automotive partners such as Hyundai, Tesla, and BMW bring design-for-manufacturing advantages. Humanoid wiring harnesses need automotive robustness in a very small form factor, like high-grade consumer electronics. Developers need partners for rapid iteration.

Besides sensing and compute, a range of power supplies allows swapping. Getting humanoids to market involves challenges.

Before smartphones, products were simpler: capabilities could be built to a cost point and scale. Now companies are resetting and diversifying supply chains. Brown said Jabil is working with major humanoid developers in the U.S. and abroad but cannot name them yet.

In factories and warehouses, humanoids make sense in many spaces, Brown said. Most automation is built for specific tasks, but humanoids' applicability to a wide range of tasks changes the math from a purpose-built line. Many humanoid companies have partnered with end customers to gather clear requirements, a solid approach.

In consumer space, use cases remain vague, such as folding laundry. On capabilities, Brown said the goal is a superset of abilities that can be applied in bits and pieces to other markets. Some humanoids use wheeled bases instead of bipedal legs, an easier problem.

A company that conquers the superset can conquer other challenges; one that conquers just one may struggle to move up the stack. The superset can be decomposed into cost-optimized solutions by use case.

Mobile manipulators have a place. Traditional automation like machine tending at 1.5 m/s is fast; a humanoid at that speed would be tremendously fast, especially carrying a load. Needs differ by facility and line.

Humanoids will be supremely flexible but not always the right device. Traditional high-speed automation will remain important. Consumer robots may arrive in about five years, Brown said, though it could be 20.

Progress is rapid. If model development and computing continue at current pace, AI and electronics hardware will be ready. There will be a market for multiple humanoid companies because needs are diverse: manufacturing, consumer, hospitality, food prep.

Policy development in one area does not guarantee wins in others. Actuator performance levels create many possibilities for fine-tuned solutions.

As humanoid fleets become ubiquitous, they will change line layout assumptions. If layout can be defined in software, work instructions sent, and a "Go" button hit, with humanoids pulling out workstations and changing end effectors, manufacturing changes fundamentally. Lines are no longer the key limiting factor; they become a programmable substrate for manufacturing.

Safety and certification remain concerns. Many robotic arms have safety certifications, standoff distances, and cages. Infrastructure built around end applications means the industry needs time to understand humanoid operation, failure modes, and certification in new environments.

Companies are tracking legislation and developing safety standards. Brown said he is optimistic, sees progress, and enjoys working with customers to lower price points and improve functionality.