Jabil、ヒューマノイドロボットのプロトタイプから数万台への増産計画を明示、エッジコンピューティングにNVIDIA Jetsonを挙げる

Reported by The Robot Report on October 9, 2026. Summarized and analyzed by XAIPART Editorial.

What happened

Jabil Inc., a St. Petersburg, Fla.-based design, supply chain, and manufacturing services provider, is working with humanoid robot developers as the category moves from low-volume pilots to tens of thousands of units per program.^S1 In a Q&A with The Robot Report, Thomas Brown, senior director for engineering and technology (including digital commerce and robotics) at Jabil, said customers that previously hand-assembled small quantities now need a cost structure, manufacturability and quality systems that only a contract manufacturer can supply.^S1

On compute, Brown named the NVIDIA Jetson series as the popular edge-AI platform for Jabil's robotics customers, while flagging price as an ongoing pain point at scale.^S1 He said foundation-model intelligence is currently "outside that capacity" and that expanding edge memory and compute is a key gating factor.^S1 For mechanical design, Brown described humanoid wiring harnesses as needing "a hybrid of automotive-type robustness with an extremely small form factor like high-grade consumer electronics," and noted that designers need partner support for rapid iteration across sensing, compute and swappable power supplies.^S1

Jabil confirmed it has worked with Apptronik on bringing the Apollo humanoid into active manufacturing environments, and said it is working with additional humanoid developers in the U.S. and abroad that it could not name at this time.^S1 Other companies preparing to ramp humanoid production named in the source include Tesla, 1X Technologies, XPeng, and UBTECH.^S1 Brown framed mobile manipulators as a parallel form factor rather than a stepping stone, noting that a traditional machine tender running at 1.5 m/s (3.3 mph) is "pretty fast," which sets the bar for any bipedal alternative.^S1

On commercial outlook, Brown said manufacturing has "a ton of spaces" where humanoids make sense because they decouple the production line from purpose-built automation, while consumer humanoids remain "the science fiction route."^S1 He estimated consumer humanoids could arrive in roughly five years, "maybe in five-ish years. It could be 20 years," contingent on model and compute progress.^S1 He also flagged that humanoid fleets could turn line layout into a software-defined, programmable substrate, and that safety certifications, standoff-distance norms and failure-mode analysis still need to be extended to cover humanoids.^S1

Corrections: none at publication. This brief will be updated with a timestamp if the source corrects its report.