In April 2025, on Tesla’s Q1 earnings call, Elon Musk told investors that Optimus, the company’s humanoid robot, had been held up by a problem few outside the rare earth industry had been watching. “Optimus was affected,” he said, “by the magnet issue from China because the Optimus actuators in the arm do use permanent magnets” (Adamas Intelligence).

Days earlier, Beijing had placed seven heavy and medium rare earths on its export control list. The admission mattered beyond Tesla. It was the first time a Western humanoid robot programme had publicly conceded that a single Chinese policy move could reach inside its production line, and the problem it exposed will outlast any one shipment.

‍

What Rare Earths Do Inside a Robot

Rare earths show up in a modern robot in three places. The powerful compact permanent magnets inside every motor and actuator are made from neodymium-iron-boron (NdFeB), often with praseodymium alongside. Where motors run hot, dysprosium and terbium are added in small amounts so the magnet keeps its strength, and samarium bonded with cobalt (SmCo) is used where temperatures would defeat NdFeB.

Beyond motors, erbium and ytterbium support fibre-optic signalling in advanced sensors, and europium, terbium and yttrium produce the colours on the displays operators use to monitor the machines. Permanent magnets alone account for more than 80% of global rare earth demand by value, per McKinsey data cited by S&P Global, because the compact high-torque motors that depend on them are now the default across robotics, electric vehicles and defence hardware.

‍

Why Humanoids Change the Scale

A standard EV traction motor uses one or two large magnets. A humanoid uses dozens of smaller, high-performance ones. Modern designs typically have 40 or more degrees of freedom, each driven by its own motor, per Adamas Intelligence cited in Sprott’s October 2026 report. That is why a single humanoid can contain multiples more magnet material than a passenger EV.

The intensity is landing on top of a sector that is already scaling. The International Federation of Robotics reported in September that industrial-robot installations crossed 600,000 units for the first time in 2025 (+11%), bringing the global operational stock past 5 million units, a stock that IFR President Jane Heffner noted is “more than double the number seven years ago.” Separately, Goldman Sachs expects the humanoid-robot market alone to grow from about $3 billion today to $38 billion by 2035.

‍

‍

Humanoids Are Already in Commercial Deployment

Commercial humanoid deployments are now live, not experimental. Agility Robotics’ Digit has handled more than 100,000 totes in commercial operation at GXO’s Flowery Branch, Georgia warehouse, where it fulfils orders for Spanx. Figure AI’s 11-month pilot at BMW’s Spartanburg plant, which contributed to the assembly of more than 30,000 vehicles, has moved into a commercial agreement for staged deployment at the same facility. Apptronik’s Apollo is already working at Mercedes-Benz, backed by a $350 million Series A raised in February 2025 with Google among its investors.

The scale shows up most clearly in two places. Hyundai plans to deploy more than 25,000 Boston Dynamics Atlas humanoids at its Georgia Metaplant from 2028, per Yonhap News Agency reporting on investor briefings. And in China, Chinese manufacturers accounted for more than 97% of global humanoid shipments in the first half of 2026, roughly 19,100 units in six months, led by AgiBot (around 8,400) and Unitree (around 5,900). The magnets powering those robots still come, overwhelmingly, from one country.

‍

The Magnet Plants Being Built for Them

Washington is backing an ex-China magnet build-out at a sustained pace. Per Sprott citing Fastmarkets, close to $2.9 billion in direct federal funding went to rare earth and magnet supply chains outside China in June 2026 alone.

Two commercial magnet plants matter specifically for robotics. USA Rare Earth commissioned its first sintered NdFeB production line at Stillwater, Oklahoma in March 2026, with first customer orders in Q2 2026 and 600 tonnes per year of run-rate capacity targeted by year-end; its stated end markets include defence, physical AI and robotics. And Noveon Magnetics, which describes itself as the only operational sintered-NdFeB producer in the United States, closed a $215 million Series C in January 2026 to expand capacity beyond 2,000 tonnes per year at its San Marcos, Texas plant, alongside a multi-year supply agreement with General Motors.

‍

What to Watch Next

The real test is deployment, not policy. Figure AI’s BMW Spartanburg commercial agreement runs on, Tesla’s Optimus is targeting pilot production in 2026, and Agility Robotics is scaling its Oregon factory. Each new humanoid deployment moves the sector a step further from demonstration into recurring magnet demand.

The next IFR World Robotics report, due in September 2027, will show whether the +11% installation growth of 2025 held. Between now and then, the question for rare earths is whether robotics stays an incremental use case, or whether it becomes, as Adamas forecasts, the single largest driver of NdFeB demand by 2040.