Theses›Robotics & physical AI›Robot autonomy

Robot autonomy

General-purpose robot autonomy (open-world dexterous manipulation, not geofenced driving) remains data-bottlenecked and unreliable through 2030.

  • Bear
  • Early
  • $47B by 2035

The callGeneral-purpose robots are coming and the bottleneck is touch, but Figure and Neura build their own skin and power, so the merchant parts window has not opened.

D1D2R1R2C1foragainst D1D2R1R2C1foragainst
Fig. 1The balance. Drivers and catalysts against restraints, each block sized by its weight; catalysts hang dashed because they have not happened yet, sized by their odds. It tips toward the restraints.
  1. D1Humanoid volume is arriving, with China's 2026 forecast doubled to about 50,000 units, and components follow units even if autonomy stalls.
  2. D2Robot hands have almost none of the 17,000 touch receptors in a human one and joint motors overheat, so touch skin, readout and joint power are what every maker needs.
  3. R1Figure builds full-stack and Neura ships its own skin, while TI, Infineon and EPC have reference designs, so the merchant component window is contested.
  4. R2If a robot scaling law shows up early, value flips back to the models and the component bet looks wrong; humanoid prices have already fallen from $85k to $25k.
  5. C1A top-5 humanoid maker discloses a merchant supplier for joint power or touch skin in production, turning those parts into catalogue components.2031
Rests onthe beliefs underneath
Supported by · confidence
What would prove itdated tests, open and resolved
20272028202920302031today123456predictionswatch 20272028202920302031today123456predictionswatch
Fig. 2Each ring is a dated test: dashed while open, struck through when overdue, ticked or crossed once resolved.
  1. 1AUTONOMY PART (the real claim).Dec 2027 · open
  2. 2DEPLOYMENT PART.Dec 2027 · open
  3. 3A robot data-scaling law, or a model at 90% first-try success on new objects in unseen places without teleoperation.Dec 2027 · open
  4. 4Any humanoid maker reports more than $100M of recurring revenue, and whether it is teleop-assisted.Dec 2027 · open
  5. 5A robot maker standardises on a merchant power part, or a whole-body skin plus readout ships; a down-round wave is the other signal.Dec 2028 · open
  6. 6Through 2030 the defensible value in the robotics-AI stack accrues to the data/eval/world-model flywheel + the component layer, NOT to robot OEMs (margin-compressed) or to standalone foundation models.Dec 2030 · open
Expressionslisted, private, technologies
Private · 11 private companies, 3 of them in the US.

1X Technologies · Unitree Robotics · Figure AI · Physical Intelligence · Skild AI · Apptronik · Agility Robotics · Neura Robotics · Wayve · Waymo · AgiBot

1 more private company tracked, not shown

Technologies
Who has already bet on itprivate rounds, from public sources

$9.2B raised by 9 companies, $9B of it in the last three years. Wave under way: Large rounds in the last three years; the category is funded and getting crowded.

Backed by a top-tier venture firm: Physical Intelligence, Agility Robotics. Backed by a strategic investor: Figure AI, NEURA Robotics, Skild AI, Wayve, Physical Intelligence, Apptronik, Agility Robotics, AgiBot.

1 more funded company in the market.

What changedthe record, newest first
  • 6 Oct 2026One former chip executive's view: NVIDIA has historically been weak at the edge, and robotics is the one area that might force a genuinely new edge chip architecture. It is a contrarian read against NVIDIA's current push into robot compute, and still unproven.
  • 22 Jun 2026Robot autonomy is gated on touch data, not models or robots; the durable value is the data flywheel and the sensing/power components every robot has to buy.
  • 13 Jun 2026Robots are having their GPT moment in the demos, but general-purpose autonomy is still bottlenecked by data and by hands that cannot feel. The money is being made selling picks to the gold rush (sensors, power electronics, data infrastructure), not on the robots themselves, whose price is already collapsing.