Robots Will Take Factory Jobs Before AI Takes Office Jobs

Will humanoid robots perform autonomous, unsupervised production tasks in 500+ global manufacturing sites by end of 2028?

If you work in manufacturing, logistics, or warehousing — humanoid robots are closer than you think. If you invest in automation — this is the next wave.

Target: Dec 2028(995 days until resolution)

Assessed Probability

40%

Roughly even odds

Based on 2 expert predictions, 4 evidence items

Community Forecast

Cast your vote

Be the first to weigh in below

Your Prediction

Where do you think this lands?

Join others who've weighed in

5%95%

50% — More likely than not

Share on LinkedIn Share on X

Everyone assumes AI kills white-collar jobs first. The humanoid robot data tells a different story. Figure AI's Figure 03 loaded 90,000+ parts into BMW vehicles during an 11-month deployment at their Spartanburg plant — but at ~25% of human speed. Boston Dynamics Atlas began autonomous factory operations at Hyundai's Georgia plant in January 2026, with production capacity planned at 30,000 units per year. Tesla commenced Optimus Gen 3 mass production targeting 50,000 units in 2026 at $20-30K each. China holds 85-90% of global humanoid robot shipments with 140+ manufacturers and 330+ models. NVIDIA's GR00T N1.6 open reasoning VLA model was built specifically for humanoid full-body control. But the 'reality gap' is the key barrier: VLA models achieve 95% accuracy in simulation, dropping to 60% in real-world conditions. That 35-point gap is the difference between demo and deployment. UBTECH has 500+ Walker S2 robots in use including border patrol — but China's lead means most Western manufacturers are 2-3 years behind.

Scenarios

Current value: ~50-100 sites with some form of humanoid presence (Boston Dynamics, Figure AI, UBTECH). Most still supervised or limited scope.

S-curve position: Very early — proof of concept established, mass deployment not yet started

Bear Case

Under 100 sites (reality gap persists, safety regulations slow deployment, ROI unclear at current capability levels)

Base Case

200-500 sites (concentrated in automotive, electronics, warehousing; mostly in China and select Western factories)

Bull Case

1,000+ sites (Tesla Optimus at $20K achieves cost parity with human workers, Chinese manufacturers flood the market)

How We'll Know

What we measure: Number of global manufacturing sites where humanoid robots perform autonomous, unsupervised production tasks (not demos, not cobots, not teleoperated)
Confirmed if: Verified autonomous humanoid robot deployments in 500+ distinct manufacturing/warehouse sites globally by end 2028
Refuted if: Fewer than 100 sites have autonomous humanoid deployments, or most remain in demo/pilot stage
Data sources: IFR World Robotics Report
Company deployment announcements (Boston Dynamics, Tesla, Figure AI)
Gartner warehouse robotics tracker
NVIDIA Isaac/GR00T deployment data

Evidence Trail

Evidence For

Mar 9, 2026
Figure 03 loaded 90K+ parts at BMW Spartanburg over 11 months. Boston Dynamics Atlas in commercial production at Hyundai (30K units/year capacity). Tesla Optimus Gen 3 mass production started Jan 2026 (targeting 50K units at $20-30K). China: 85-90% of global humanoid shipments, 140+ manufacturers, 330+ models. UBTECH: 500+ Walker S2 in use including border patrol. Gartner: 80% of global warehouses will use AI-driven robotics by 2028. NVIDIA GR00T N1.6 VLA model for humanoid control.→ Probability: 35%
Mar 9, 2026
Jensen Huang (CES 2026): 'the ChatGPT moment for robotics.' Sim-to-real transfer improving — VLA models at 95% simulation accuracy. BotQ factory ramping to 12K units/year. Multiple platforms achieving 10-hour autonomous shifts. Cost trajectory ($20-30K per unit) approaching ROI breakeven for 24/7 manufacturing shifts.→ Probability: 40%
Apr 10, 2026
Tesla deployed 1,000+ Optimus Gen 3 robots across Gigafactory Texas and Fremont — the largest-scale humanoid deployment in manufacturing history. Tesla targeting 1M units annually by late 2026, with a dedicated 10M/year factory under construction. Figure AI's Figure 03 launched October 2025 with Hyundai (targeting 30,000-unit/year scale from 2028) and Google DeepMind as 2026 anchor customers. BotQ facility scaling from 12K toward 100K robots/year. Real production deployment is now measured in thousands, not hundreds. The unsupervised-operation question remains open (current Tesla deployments appear task-limited) but the scale trajectory is ahead of March baseline.→ Probability: 40%

Evidence Against

Mar 9, 2026
Reality gap: 95% simulation accuracy drops to 60% in real-world conditions — a 35-point gap. Figure 03 at ~25% of human speed. Most deployments remain supervised or teleoperated. Safety certification for unsupervised humanoid operation doesn't exist yet. ROI uncertain at current capability levels. 500 sites is a high threshold for 2.5 years.

How Our View Evolved

Apr 10, 202635%↑40%
Tesla Optimus Gen 3 deployed 1,000+ robots across Gigafactory Texas and Fremont — largest humanoid manufacturing deployment yet. Figure 03 with Hyundai (30K-unit/year target) and Google DeepMind as 2026 anchors. Scale trajectory is ahead of baseline, restoring +0.05 from the March 15 audit correction. Unsupervised-at-500-sites threshold still aggressive; conservative partial restoration.
Mar 15, 202640%↓35%
Audit recalibration: fixed prediction mapping (P-034 safety→P-006,P-014 robotics). 500+ sites by 2028 is ambitious — Figure 02 at BMW is 1 site. Reality gap (95%→60%) persists. Slight downward correction.
Mar 9, 2026Initial assessment: 40%
Baseline — initial published assessment. Based on Figure AI/BMW, Boston Dynamics/Hyundai, Tesla Optimus, NVIDIA VLA data.

What Experts Say

Sam Altman

CEO, OpenAI

Track record: 7/10

“Robots will be executing tasks in the real world by 2027”

Dec 2025 | blog

We assess this claim as 75% likely

Gary Marcus

AI Researcher, NYU Professor Emeritus, AI critic