AI ROBOTICS

Emerging

Who Puts AI to Work on the Factory Floor?

Turning AI-powered machines into reliable, affordable work.

Strategic Brief

September 11 – October 11, 2026 · 30-day retrospective · Evidence checked through October 11, 2026

The bigger trend

Humanoid robots are moving from demonstrations into factory training, but proof that they do useful work at a sensible cost is still thin.

Hyundai opened a training center for its Atlas humanoid robots inside a working car plant, with plans for tens of thousands of units. Tesla is reported to be ramping production of its Optimus robot, and NVIDIA is reported to be weighing another large investment in Figure. These are plans and production numbers, not yet proof of economic work.

Outside → In

What changed — Sense (verified)

September 21: Boston Dynamics opened a center to train Atlas humanoid robots at Hyundai's Metaplant near Savannah, Georgia. Hyundai Motor Group plans to deploy 25,000 Atlas robots across Hyundai and Kia plants and to build a U.S. plant able to make 30,000 robots a year.

September 24: Amazon announced a $100M+ plant in Greenwood, Indiana, to make products for its fulfillment and robotics network.

September 25 (REPORTED): Tesla is now making several hundred Optimus robots a week, up from several dozen earlier in the year, and is targeting more than 1,000 a week by year-end, according to The Information. Tesla has not confirmed the figures.

October 7 (REPORTED): NVIDIA has discussed investing another $1 billion in Figure AI, The Information reported; no deal has been announced.

What it means — Interpret (RAW PEC view)

Four stages matter: robots shown, robots built, robots deployed, and robots doing useful work at a cost that beats the alternative. This month's news moved the first three; none of it shows the fourth. Suppliers of robot "brains" and simulation, such as NVIDIA and Siemens, can sell to many robot makers at once, but whether that pays off is still unproven.

What to consider — Position

Which repetitive, well-defined tasks would you hand to a robot first, and what evidence of safety, reliability and cost per task would you need before deploying it?

Inside → Out — suggested enterprise actions

Suggested actions for readers, not a report of actions taken.

Who needs to align — Align

Operations, manufacturing engineering, safety, IT, finance and workforce leadership.

What the enterprise could execute — Execute

Pick one safe, measurable task and run a time-boxed pilot with a specialist or an established industrial supplier.

How to measure and learn — Learn

Task completion rate, human interventions, uptime, safety incidents, integration cost and cost per completed task.

Supporting competitive moves (this period)

  • NVIDIA / Figure AI · Partner ·

    Talks on a further $1B investment (reported)

    Primary source · www.humanoidsdaily.com

  • Amazon · Build ·

    $100M+ Indiana plant supporting its fulfillment and robotics network

    Primary source · press.aboutamazon.com

  • Hyundai / Boston Dynamics · Build ·

    Atlas training center at the Georgia Metaplant

    Primary source · bostondynamics.com

  • Siemens / NVIDIA · Partner ·

    Industrial AI Operating System, starting at Erlangen

    Primary source · news.siemens.com

  • SoftBank / ABB · Buy ·

    Agreement to acquire ABB Robotics for $5.375B (closing pending)

    Primary source · group.softbank

Related: AI Driving (Who Controls the Future of Driving?) · AI Economics (How AI Cost, Value and Pricing Connect)

The management question

Which physical tasks in your operations are ready for AI robots, and should you build that capability, partner with a specialist, or rely on established industrial suppliers?

Stack layers

Contested layers: 5 Domain platforms; 6 Workflow, agents and orchestration

Adjacent layers: 1 Physical and energy foundations; 4 Models and intelligence

Contest dimensions

  • Robot intelligence
  • Manufacturing
  • Deployment
  • Cost of work

Participants

Robot makers

  • Hyundai / Boston Dynamics (Challenger): Atlas training center at Hyundai's Georgia plant; 25,000 Atlas robots planned (September 21, 2026)
  • Tesla (Challenger): reported Optimus output of several hundred a week, aiming for more than 1,000 a week by year-end; not confirmed by Tesla (reported September 2026)
  • Figure AI (Challenger): last valued at $39 billion (September 2025); NVIDIA reported to have discussed a further $1 billion investment (October 7, 2026)
  • SoftBank (ABB Robotics) (Incumbent): agreed to buy ABB's robotics business for $5.375 billion (October 2025); closing expected mid-to-late 2026, pending regulatory approvals

Robot intelligence and industrial software

  • NVIDIA (Incumbent): robot models and simulation tools introduced at GTC (March 2026), used by ABB, FANUC, KUKA and Yaskawa
  • Siemens (Incumbent): building an Industrial AI Operating System with NVIDIA, starting at its Erlangen electronics factory (January 2026)
  • Physical Intelligence (Challenger): general-purpose robot AI; in talks to raise at an $11 billion valuation (March 2026)

Operators

  • Amazon (Incumbent): $100M+ Indiana plant making products for its fulfillment and robotics network (September 24, 2026); has built more than one million robots in the U.S.

Ecosystem Moves

How the players are building, buying and partnering to compete in this battleground. Newest first.

2 Build · 1 Buy · 2 Partner · 0 Compete

Independent and unsponsored: no company pays to appear in these battlegrounds.

Sources
Talk with Ron