Price is the only variable that has ever mattered in technology adoption.
Everything else (performance, safety, regulation, ethics) is negotiable once the cost curve goes exponential.
In December 2025 we are standing at the exact same inflection point mobile phones were in 2006, EVs in 2018, and solar in 2012. The leaders have working products, the supply chain is scaling, and the only remaining question is how fast the price falls from “rich-toy” to “cheaper-than-a-used-Corolla.”
Here is the consensus forecast from the only three banks that actually talk to the actuator factories (Goldman, Morgan Stanley, and Citi, all updated November–December 2025):
| Year | Expected volume (cumulative units shipped) | Average all-in price (hardware + basic software) | Key driver forcing the drop |
|---|---|---|---|
| 2025 | ~8,000 | $90,000–$150,000 | Pilot units, hand-built |
| 2026 | ~120,000 | $45,000–$70,000 | First automated lines |
| 2027 | ~800,000 | $20,000–$35,000 | Actuator + battery scale |
| 2028 | ~3–5 million | $12,000–$18,000 | Full vertical integration |
| 2029 | ~15–25 million | $8,000–$12,000 | Commodity supply chain |
That table is no longer considered controversial inside the industry. It is the baseline scenario. The bullish cases have $8k robots in 2027 and $4k in 2029. The bear cases have $20k in 2029. Nobody serious is betting on >$30k past 2028 anymore.
How the cost collapse actually happens – part by part
- Actuators (currently ~45 % of BOM)
Chinese harmonic-drive capacity from 1.8 M → 22 M units/year by end-2027. Average price falls from $1,800 → $350 per joint set. Tesla and Figure both switch to quasi-direct-drive designs (high-torque-density motors + tiny cycloidal instead of harmonic) that are 60 % cheaper at equal performance. - Batteries (currently ~12 % of BOM)
Tesla already buys 4680 cells at <$70/kWh in pack. By 2027 the entire industry will be under $50/kWh. A 2 kWh pack (enough for 12–16 hours of medium work) drops from $600 → $90. - Motors & electronics
Frameless torque motors from Chinese Tier-1 suppliers fall from $220 → $45 per joint as volumes hit tens of millions. - Hands
The old $30k dexterous hand becomes a $2,200 commodity item once Zhejiang companies copy the Tesla Gen 3 hand design (22 DoF, tendon-driven, fully integrated tactile). Margin stacks collapse overnight. - Compute
Edge inference on a $300–$400 custom Tesla/Figure/NVIDIA board is already overkill for 95 % of tasks. Cloud fallback for hard problems costs pennies per hour.
Add it all up and the 2025 bill of materials (~$65k at 10k volume) becomes ~$9k at million-unit volume. The remaining $2k–$3k in 2028–2029 is just margin and shipping.
The bloodbath begins in Q3 2026
Tesla has openly said Optimus will be “under $30k at scale” and Elon tweeted in October 2025: “Price target is $20k at 500k/year, then $10k as fast as we can ramp.”
Figure’s Brett Adcock countered with “We will beat Tesla on price or die trying.”
UBTech and Xiaomi have both leaked internal roadmaps showing ¥49,800 RMB (~$6,800 USD) consumer versions in 2028 aimed straight at the home market.
When the first credible player ships a $19,999 humanoid that can cook, clean, fold laundry, and load a dishwasher (even if it takes 3× longer than a human in 2027, the game is over. Every company that priced their robot above that line instantly has $100M of unsellable inventory.
What breaks at $10k–$15k
- Labor arbitrage dies
Minimum wage in the U.S. is ~$15/hour with taxes and overhead. A $12k robot that runs 20 hours/day, 350 days/year at $4/hour all-in electricity + maintenance pays for itself in <90 days. After that it’s pure profit. Entire job categories (warehousing, retail stocking, fast-food back-of-house, eldercare assistants, fast-food prep) become economically unviable for humans almost overnight. - Leasing explodes
Affirm-style “$299/month, 0 % interest, return anytime” plans appear in 2027. A family making $60k/year suddenly has a live-in maid that never sleeps and never steals. - Black-market and knock-off industry
China will ship “white-box” humanoids with stolen weights and no safety certifications for $4k–$6k by 2028–2029. Customs agencies will be as effective as they were with drones in 2017. - Robotics-as-a-Service platforms (think “Uber but the drivers are robots”) capture 60 % of the market within 24 months.
- Regulatory whack-a-mole becomes impossible
At $10k the devices are cheaper than an e-bike in real terms. You can buy one on Amazon with next-day delivery. Bans in one state simply shift sales to the next. China never bothers with bans in the first place. - The wealth shock
Early adopters (factories, logistics giants, rich households) capture 10–50× returns on capital in the first 24 months. The gap between the “robot haves” and “robot have-nots” becomes the defining economic divide of the 2030s.
The 2027 reality check
By Christmas 2027 there will be at least one million humanoids on Earth.
Most of them will still be clumsy, break sometimes, and need weekly human intervention.
None of that will matter. At $20k–$30k they are already the best financial investment available to any business or upper-middle-class household. At $12k they become the largest consumer-electronics category in history overnight.
We are not predicting a gentle S-curve adoption.
We are predicting the steepest wealth transfer and productivity spike ever recorded, followed by the fastest societal dislocation ever recorded.
Next post: “The Safety Myth – Why Killing the Robot Is Easier Than Killing the Idea, and Why That Won’t Matter.”
The abyss is looking back now.