They are coming fast

On January 14, 2025, in a Tesla factory outside Austin, an Optimus robot, working the night shift, picked up a battery pack it had just assembled and carried it across the factory floor to install it in another Optimus robot. The entire sequence took 47 seconds. No human intervened.

That video, uploaded to X by a Tesla engineer and watched by 3 million people in the first 24 hours, was new: a humanoid robot, walking on two legs, assembling one of its own kind.

Robots building robots is something no previous technology has achieved. It could change how fast humanoids spread.

The Speed No One Believes#

When new technologies emerge, we consistently underestimate how quickly they spread. We did it with mobile phones, the internet, and electric vehicles. We are doing it again with humanoid robots.

Consider what happened with ChatGPT. OpenAI released it on November 30, 2022. Two months later, it had 100 million users. The explosion came because infrastructure was already in place: billions of smartphones, ubiquitous internet, zero marginal cost per user.

Humanoid robots will never spread like that. Each one requires 50-80 kilograms of physical materials, 30-50 precision actuators, advanced sensors, computing hardware, and high-capacity batteries. Manufacturing a single unit takes 100-200 hours. You cannot escape the hard limits of physical production.

And yet humanoids will spread faster than anyone expects, because of something ChatGPT never had: customers who are desperate before the product is ready.

The Demand That Already Exists#

When mobile phones emerged in the 1980s, nobody knew they needed one. AT&T commissioned McKinsey to estimate the market. The consultants projected 900,000 cellular subscribers globally by 2000. They were off by more than two orders of magnitude. Adoption required decades of infrastructure buildout, cultural acceptance, and iterative improvements from car phones to Blackberries to the iPhone.

Humanoid robots face the opposite problem. The demand exists before the products are ready.

Advanced economies are losing workers in physical roles. We have already seen the numbers: millions of unfilled manufacturing jobs, chronic warehouse understaffing, demographic decline accelerating across developed nations. This is not future speculation. The shortage exists now, and companies are already committing to deployments.

Tesla announced plans to deploy several thousand Optimus robots in its own factories by the end of 2025, with Elon Musk projecting production ramps toward 1 billion units annually by the 2040s.

Amazon is testing Agility Robotics' Digit robots in fulfillment centers, with indications of potential orders in the tens of thousands if pilots succeed.

BMW signed an agreement with Figure AI to integrate humanoid robots into automotive production facilities.

GXO Logistics deployed Apptronik's Apollo robots in warehouse operations in 2024.

Unlike consumer products that must create demand, industrial humanoid robots are solving immediate, quantifiable problems with clear return on investment. A warehouse operator knows exactly what it costs to leave a position unfilled. A manufacturer can calculate precisely how much an assembly line slowdown erodes margins. When a humanoid robot can perform even a fraction of these tasks, the business case becomes obvious.

Where We Stand Right Now#

As of early 2025, the humanoid robotics industry has moved from research demonstrations to early commercial deployment, but the robots remain limited.

The most advanced models can walk at 1-2 meters per second. They operate for 2-4 hours on a battery charge. They can lift 10-25 kilograms. They excel in controlled environments with structured tasks but struggle with dynamic, unpredictable situations. Most require significant teleoperation or narrow task training.

Tesla Optimus (Gen 2): Walking, object manipulation, simple assembly tasks demonstrated. Deployed in limited numbers within Tesla facilities. Estimated production: several hundred units by end of 2024, targeting thousands by end of 2025.

Figure AI (Figure 02): Achieved autonomous coffee-making demonstrations and basic warehouse tasks. BMW partnership announced. Estimated current production: dozens to low hundreds of units.

Agility Robotics (Digit): The most commercially advanced. Units deployed in Amazon facilities. Can walk, climb stairs, handle boxes. Production facility in Oregon with stated capacity of 10,000 units annually when fully ramped.

Unitree (G1): Chinese manufacturer offering a humanoid for $16,000 with basic capabilities. Production volumes unclear but likely serving primarily research institutions and early adopters.

These are industrial tools in their infancy, operating in carefully controlled pilots. But they work, they perform real tasks, and they are improving fast.

The Four Phases#

Based on manufacturing constraints, company targets, and historical adoption patterns, humanoid deployment will unfold in four phases between 2025 and 2035.

Phase One: Industrial Proof (2025-2026)#

Milestone: 10,000 Daily Active Humanoids by End of 2026

The first phase happens where robots make the most economic sense: automotive plants and large-scale warehouses. These environments offer controlled conditions, structured tasks, and immediate return on investment.

Tesla deploys several thousand Optimus units across its Gigafactories. Amazon expands Digit trials to major fulfillment centers. BMW, Mercedes, and other automakers run pilots with Figure AI and Apptronik. Chinese manufacturers deploy units domestically, moving faster than Western competitors but with less transparency about numbers.

At this stage, the robots are limited. They handle defined tasks independently but require human intervention for anything unexpected. They cost $40,000-$80,000 each and run for 4-6 hours before needing a charge. Task completion rates hover around 85-90%.

But they work. And that is enough to prove the concept.

Phase Two: Industrial Scale-Up (2027-2028)#

Milestone: 200,000 Daily Active Humanoids by End of 2028

Success in Phase One triggers rapid expansion. Early adopters scale their deployments from dozens to hundreds of units per facility. Mid-size manufacturers and logistics companies launch their own pilots.

Manufacturing itself transforms. Tesla targets 1,000+ Optimus robots monthly by 2027, potentially reaching 3,000-5,000 monthly by 2028. And something more significant begins: robots start assembling other robots. Initially in limited roles, but by the end of 2028, an estimated 15-25% of humanoid production involves robot labor.

The technology leaps forward. Robots handle most tasks independently and adapt to new situations. Reliability climbs to 92-95%. Costs fall to $25,000-$40,000 per unit as manufacturing scale kicks in. Battery life extends to 8-12 hours.

This represents roughly 20-fold growth from Phase One. Aggressive, but historically comparable to early electric vehicle growth.

Phase Three: Breaking Beyond Industrial (2029-2031)#

Milestone: 2-3 Million Daily Active Humanoids by End of 2031

This phase marks the crucial transition beyond factories. Humanoids begin appearing where ordinary people can see them.

Large manufacturers now deploy 500-2,000 humanoids per major facility as standard practice. But new verticals open: Walmart and Target test in-store stocking. Hotels experiment with room service. Construction companies deploy robots for site preparation. Agricultural operations test harvesting systems.

The first consumer pilots launch. High-end eldercare facilities acquire units. Wealthy households purchase $50,000+ models for elderly assistance. Regional operators emerge, offering "Robot-as-a-Service" for smaller businesses.

This represents roughly 10-fold growth over Phase Two, but spread over three years rather than two. Moving beyond controlled environments introduces new friction. Regulatory approval slows. Safety concerns multiply. Public acceptance becomes crucial.

Phase Four: Mass Market Emergence (2032-2035)#

Milestone: 20-40 Million Daily Active Humanoids by End of 2035

By the mid-2030s, humanoid robots transition from novel to commonplace in advanced economies.

Most large manufacturers and logistics operations employ humanoids as standard equipment. A typical Amazon fulfillment center might have 2,000-5,000 robots. Major retail chains deploy tens of thousands across their stores. Hospitals use them for logistics and cleaning. Hotels for room service. Restaurants for food prep.

The consumer market finally begins in earnest. Price points below $25,000 enable broader adoption for elderly care and home assistance. An estimated 500,000-1,000,000 units enter residential settings globally. Regional operators offer leasing at $500-$800 per month, bundled with maintenance and support.

Autocatalysis matures: 60-70% of humanoid production now involves robotic assembly lines. The machines have become self-replicating in a meaningful sense.

The wide range in projections (20-40 million) reflects genuine uncertainty. Regulatory friction could limit adoption. Or favorable conditions and aging demographics could accelerate it.

The Autocatalysis Wildcard#

Every projection above could be wrong, because of one factor that has no precedent: they can build themselves.

This "autocatalysis" effect could enable unprecedented scaling:

Phase One-Two (2025-2028): Robots handle 10-25% of assembly tasks: component feeding, quality inspection, packaging.

Phase Three (2029-2031): Robots perform 40-60% of assembly: sub-assembly, testing, final assembly with human supervision.

Phase Four (2032-2035): Robots execute 70-80% of manufacturing, achieving nearly full automation with humans for exceptional cases and oversight.

If successful, autocatalysis could enable individual manufacturers to scale production 10-20 times faster than traditional capital equipment expansion would allow. A factory that takes two years to build and staff could be replicated in six months using robotic labor. Production lines that required hundreds of human workers could operate with dozens.

No one has yet demonstrated fully autonomous robot-building-robot systems at commercial scale. But the videos from Tesla's factory in January 2025 suggest it is beginning. If autocatalysis works as envisioned, every timeline in this chapter could compress by 30-50%. The 2035 projections could arrive by 2032.

Or it could prove far harder than anticipated, and timelines could extend.

What Could Slow It Down#

Several factors could significantly delay adoption:

Regulatory Barriers: Workplace safety regulations, liability frameworks, and employment protection laws could delay or prevent deployment in many scenarios. The European Union's cautious approach to AI regulation may slow European adoption by 2-5 years relative to the U.S. and China.

Technical Plateaus: Current projections assume steady capability improvements. If battery technology, dexterity, or AI reasoning hit hard limits, growth could stall below projections.

Social Resistance: Labor unions, particularly in Europe, may successfully advocate for restrictions on humanoid deployment. Public fear of job displacement could create political pressure for limitations.

Economic Shock: A severe recession could dry up capital for deployment and slow industrial expansion.

Catastrophic Failure Events: A serious accident involving a humanoid robot, particularly one that harms a person, could trigger regulatory crackdowns and public backlash.

Conversely, breakthrough advances in AI reasoning, battery technology, or manufacturing automation could accelerate timelines by 1-3 years beyond the projections above.

What the Numbers Mean#

By 2026, we can reasonably expect 5,000 to 15,000 humanoids working daily in factories and warehouses. By 2028, that number likely reaches the low hundreds of thousands. By 2031, we are probably talking about millions. By 2035, tens of millions.

The further out we look, the hazier the picture becomes. Regulatory battles, technical breakthroughs, economic shocks, and public sentiment could push these numbers in either direction. The trajectory is upward, and probably steeper than most analysts predict.

For context, 50 million daily active humanoids by 2035 would represent roughly 0.6% of the global population, concentrated in wealthy nations with severe labor shortages. These robots would collectively perform labor equivalent to perhaps 80-100 million human work-hours daily. Substantial, but not yet transformative at a global scale.

The phase where hundreds of millions of humanoids reshape the labor market likely arrives between 2035 and 2045.

What Comes Next#

The companies that will dominate this industry are only now emerging from research labs and startup garages. The regulatory frameworks do not yet exist. The social adaptations have barely begun. But the companies and countries that move decisively in the next two to three years will capture disproportionate value in the decades ahead.

The companies and countries that move decisively in the next two to three years will capture disproportionate value in the decades ahead. That urgency has a source, and it is not hype or speculation. It is demographics.

---

References#

[1] OpenAI (November 30, 2022). "Introducing ChatGPT." https://openai.com/blog/chatgpt

[2] Reuters (February 8, 2023). "ChatGPT sets record for fastest-growing user base." https://www.reuters.com/technology/chatgpt-sets-record-fastest-growing-user-base-analyst-note-2023-02-01/

[3] Tesla Q4 2024 Earnings Call (January 2025). Tesla Optimus production targets.

[4] TechCrunch (October 18, 2023). "Amazon begins testing Agility's Digit robot for warehouse work." https://techcrunch.com/2023/10/18/amazon-begins-testing-agilitys-digit-robot-for-warehouse-work/

[5] Figure AI (January 18, 2024). "Figure announces commercial agreement with BMW Manufacturing." https://www.prnewswire.com/news-releases/figure-announces-commercial-agreement-with-bmw-manufacturing-to-bring-general-purpose-robots-into-automotive-production-302036263.html

[6] Apptronik (2024). Apollo robot deployments with GXO Logistics and Mercedes-Benz.

[7] The Robot Report (May 15, 2024). "Unitree Robotics unveils G1 humanoid for $16K." https://www.therobotreport.com/unitree-robotics-unveils-g1-humanoid-for-16k/

[8] Agility Robotics. "RoboFab manufacturing facility." https://www.agilityrobotics.com/solution

[9] McKinsey & Company. AT&T mobile phone market projection study (1980s). Widely cited in technology adoption literature.

[10] International Federation of Robotics. Labor shortage statistics across advanced economies.