The Rise of Embodied AI: Humanoids Step Into Reality

A new era of Robots is around the corner. The autonomous machines resembling humans are transitioning from science fiction to real-world reality, powered by rapid AI advancements. Unlike fixed industrial robots or simple consumer devices like Roomba, these "new robots" use AI models to learn tasks autonomously in virtual worlds, enabling them to pack boxes, sort parts, perform household chores, or handle repetitive/dangerous work.

The robotics industry is poised to become a multitrillion-dollar market by 2050, rivaling cars or smartphones:

• Morgan Stanley's Adam Jonas: $25 trillion in combined robot revenues (including cars, drones, humanoids, etc.), ushering in the "third Industrial Revolution."
• RBC's Tom Narayan: 350 million units sold annually at ~$25,000 each → $9 trillion market.
• UBS's Phyllis Wang: 300 million humanoids → $1.4–1.7 trillion addressable market.
• Broader estimates: Up to 1 billion humanoids in operation by 2050 (from near-zero in 2030 to explosive growth in 2040–2050).

Robots already operate in factories and warehouses, but widespread home use remains distant due to production scale issues and high costs ($100,000–$200,000 per unit currently). AI progress is accelerating timelines, and embodied AI will amplify trends like demand for AI chips, data connectivity, power generation, and U.S. manufacturing revival. Investors see this as a long-term portfolio essential, even if peak adoption is decades away.

Current vs. New Robots: The Shift to Mobile, Learning Humanoids

The dominant players today are industrial giants with fixed robots:

• Kuka (owned by China's Midea), ABB (selling to SoftBank), Fanuc, and Yaskawa Electric (Japan).
• ~500,000 units/year sold, growth tied to slow auto industry demand.

The exciting "new robots" are mobile, AI-trained, and capable of human-like tasks with exponentially improving capabilities:

• Boston Dynamics' Atlas (majority-owned by Hyundai): 6'2", ~200 lbs, lifts 110 lbs, 360° vision, operates in -4°F to 104°F. Billed as the world's most dynamic robot; customers recover costs in ~2 years via efficiency gains.
• Tesla's Optimus (Gen 3 in development): ~5'8", 125 lbs. Evolved from suit-dancing demos to remote-controlled serving and learning kung fu. Musk claims it will learn by observing humans or verbal instructions; aims for $20,000 production cost at ~1 million units/year scale. Tesla is repurposing Model S/X lines for robot production.
• Figure AI's F.03: ~5'8", 135 lbs, carries 44 lbs; positioned for home tasks like laundry, cleaning, and dishes (like Rosie from The Jetsons); home testing planned by year-end.
• Agility Robotics' Digit: ~5'9", carries 35 lbs; already deployed in Amazon and GXO Logistics warehouses; ~10,000 units/year capacity.
• Chinese players (Unitree, UBTech) are advancing fast, showcased at the 2025 World Humanoid Robot Games in Beijing, creating intense U.S.-China competition.

Timelines are shortening: Self-driving tech (Tesla FSD, Waymo) trains AI similarly, and Musk calls autonomous cars "robots on wheels." Virtual-world training has replaced manual programming, yielding "physical AI" that interacts with the real world.

Challenges and Path Forward

Key barriers: High costs, lack of mass manufacturing, and supply chain gaps (e.g., falling AI compute costs needed). Auto makers like Tesla and Hyundai have advantages in sensors, actuators, batteries, metals, and high-volume production expertise. Boston Dynamics benefits from Hyundai's manufacturing prowess and partnerships (e.g., DeepMind, Nvidia).

Investment Opportunities: No Pure Plays Yet, But Strong Beneficiaries

The sector is early in the hype cycle; ETFs like KOID and BOTZ are suboptimal (heavy in rare earths or legacy industrial robots).

Top picks:

• Nvidia (NVDA): Backbone for robot chips, training (Groot model, Jetson Thor, Cosmos platform); extends existing AI business.
• Hyundai Motor: Owns ~88% of Boston Dynamics (~$25B hidden asset); shifting from slow-growth auto to embodied AI; attractive valuation ex-Boston Dynamics.
• Tesla (TSLA): Valued at $1.5T as AI/robotics play (200x 2026 earnings); huge upside from Optimus, but expensive—wait for better entry.
• Mobileye Global (MBLY): Acquired Mentee Robotics; proven autonomous tech edge.
• MP Materials (MP): Rare earths for motors/actuators.
• Aptiv (APTV): Auto parts adaptable to robots.

In summary, the robot revolution is real and transformative, driven by AI and auto-manufacturing expertise. Challenges remain in cost/scale, but opportunities in chips, components, and production leaders could reshape portfolios and industries.

Written by Antonio Costa, our Global Head of Equity Investments Alcral Research