Chinese Planetary Roller Screws vs GSA & Rollvis: The 80% Import Gap Inside Every Humanoid Leg

A planetary roller screw is the strongest way known to turn a spinning motor into a straight push. Instead of steel balls rolling at points, a ring of threaded rollers makes line contact with the shaft — so a roller screw of the same size as a ball screw carries far more load, runs stiffer, and survives repeated impact. That is exactly what a knee, hip or ankle does when a 50 kg humanoid lands a step. The roller screw is the single most expensive component category in a humanoid's bill of materials — reducers and roller screws together account for roughly 33% of a robot's BOM, and one humanoid-grade screw costs an estimated $1,350-2,700 (J.P. Morgan, 2025).

For decades, almost nobody outside a handful of Swiss, Swedish, German and American factories could make one to spec. That wall is what the Chinese supply chain is now climbing — with a speed and a price curve that are rewriting who gets to sell the humanoid era its legs.

The screw that decides how strong a robot is

Tesla's Optimus Gen 2, unveiled December 2023, is the most documented example: 28 custom actuators — 14 rotary (frameless torque motor + harmonic drive) and 14 linear (frameless torque motor + planetary roller screw), distributed through arms and legs. Teardown and supply-chain research report that the linear actuators use inverted-type roller screws supplied by Switzerland's GSA, deployed across upper arm, lower arm, thigh and lower leg. Figure AI, Agility Robotics and 1X Technologies also build roller-screw linear actuators into their robots, and Chinese humanoid programs — Unitree, UBTECH, AgiBot, Kepler and dozens more — now specify them as standard.

The manufacturing bar explains the bottleneck. Each screw and roller is case-hardened alloy steel, ground after heat treatment so shaft and rollers hold thread-lead error to roughly 3 micrometers — about one-twentieth the width of a human hair — on specialized multi-axis thread grinders that very few factories on Earth own. That is why output has not scaled with demand, why the market is growing at over 30% a year, and why Jefferies sized the global planetary roller screw market at roughly $1.8 billion with a 5-year CAGR above 30% as humanoid programs move from prototypes to production runs.

The Swiss stranglehold

Planetary roller screws are the most concentrated precision component in robotics. Switzerland's GSA (Gewinde Satelliten Antriebe AG, founded 1982, linked to the century-old thread-grinder Gewinde Ziegler) and Rollvis (founded in Geneva 1970, acquired by GSA around 2016) are consistently reported as holding more than half of global supply. Sweden's Ewellix — the former SKF linear-motion division, sold to Triton in 2018 and acquired by Germany's Schaeffler in 2022 for about EUR 582 million — is the third pillar. America's Moog builds both standard and inverted roller screws at its Presezzo plant near Bergamo, Italy; Bosch Rexroth sells PLSA screw assemblies; Tolomatic integrates them into US actuator lines.

For China the math is stark: domestic market still roughly 80-85% import-dependent, with GSA, Rollvis and Ewellix holding about 70% of China's own planetary roller screw market, according to Chinese industry research. In other words, the country that makes more humanoid robots than anywhere else was paying Swiss prices for the component that decides whether those robots can stand.

The comparison: Swiss incumbents vs Chinese challengers

Dimension	GSA / Rollvis (Switzerland)	Nanjing Process Equipment (China)	Shuanglin / NOUS ROBOT (China)
Global position	>50% combined global supply; GSA acquired Rollvis ~2016	Most-cited Chinese producer with mass-production capability	Shuanglin: 100,000-set line planned 2026; NOUS: world's smallest 1.5mm screw in volume
China market share	GSA + Rollvis + Ewellix ≈ 70% of China; imports 80-85% overall	Leading domestic brand; import substitution reference	Sample / early-volume stage, ramping 2026
Price per unit	~$1,350-2,700 (humanoid-grade, J.P. Morgan 2025)	~¥2,000-3,000 typical domestic (2025, down from ¥10,000)	NOUS micro screw <¥100; Shuanglin mini screw 3.7g
Precision	C3-C5 with decades of field data; ~3 µm lead tolerance	C5-class grinding, closing gap (2-3 grades behind)	NOUS C5-level mass production; Shuanglin 0.003mm precision claimed
Load & life	Benchmark dynamic load; 8-9/10 yield; 100k+ units/month	~30% lower dynamic load; life 20-50% shorter; yield 60-70%	Shuanglin claims 300% load gain; NOUS 3M-cycle life on 1.5mm screw
Humanoid exposure	Optimus Gen 2 linear actuators (reported GSA); Figure/Agility/1X	Sampling with domestic humanoid programs	NOUS 1.5mm linear joints adopted by leading domestic humanoids; Shuanglin samples to NEV-backed humanoid maker

The honest gap is in volume and consistency, not existence. Chinese makers have crossed the "we can make it" threshold; the question now is whether they can make 100,000 nearly identical screws a month — which is precisely where the automotive supply chain that entered this market excels.

The Chinese charge: gearmakers, then automakers

China's roller screw push started with precision-machine veterans — Nanjing Process Equipment (the most frequently cited domestic producer with genuine mass-production capability) and Shandong's Best Precision. Then the money arrived. In October 2024, automotive-parts maker Shanghai Beite Technology announced a roughly RMB 1.85 billion (~$260 million) investment in a dedicated planetary roller screw production base. Industry tracking shows A-share companies announcing hundreds of billions of yuan of screw capacity since 2025, including Wuzhou New Spring's 98,000-set planetary roller screw plan in Xinchang, Zhejiang.

The most interesting entrants are automotive suppliers, because they bring what roller screws actually need: batch manufacturing discipline, automotive-grade quality systems, and cost control learned over decades.

Shuanglin Co. developed a self-designed inverted planetary roller screw it says carries 300% higher load with 0.003mm precision for humanoid use, delivered its first sample batch for a humanoid joint-module program in 2025, and is scheduling a 100,000-set mass production line for 2026 after a 1,500-set pilot line. Best Precision entered through Germany's KISSLING acquisition and targets batch supply in 2026. Hengli Hydraulic is going one level up, building integrated "screw + motor + encoder" linear actuators on its hydraulic-cylinder distribution channels. Dingzhi Technology took the micro route — hollow micro roller screws aimed at fingers and wrists, an under-served corner of the market.

And then there is NOUS ROBOT (Shenzhen, founded 2023 out of a family business with 30+ years of screw manufacturing). Its claim to fame is the world's smallest planetary roller screw — 1.5mm shaft diameter, 5.5mm nut — mass-produced at C5 precision, with a 3-million-cycle life and output locked at under ¥100 per unit, roughly two orders of magnitude below prior sample prices. Its micro linear actuators have been adopted by leading Chinese humanoid makers for next-generation dexterous hands, and its investor list tells the strategic story: SAIC Capital (Angel+, Dec 2025), then a Series A over RMB 100 million led by Shanghai Semiconductor Industry Investment with Lenovo Ventures and Amber Capital (Mar 2026), then a near-100-million RMB A+ round with Shunwei Capital (2026). The same technology platform is being pushed into automotive drive-by-wire chassis — robots and cars sharing one screw.

The three bottlenecks that decide the race

1. Grinding machines. The highest-threshold step — grinding the nut's internal multi-start thread — still depends on high-precision thread grinders imported from Japan and Europe, with long lead times and unit costs in the millions. Domestic candidates include Qinchuan Machine Tool (which builds both grinders and screws), Huachen Equipment and Zhejiang Hademann. Whoever secures stable grinders, or builds their own, controls the speed of the entire race.

2. Yield at scale. Overseas incumbents run 80-90% yield; Chinese heads are at 60-70%. A 20-point yield gap is a 20-point cost gap on a part where every micron costs money. Yield, not samples, is the real proof of manufacturing maturity.

3. Monthly capacity. Chinese leaders are still at 10,000-20,000 units a month; the incumbents are above 100,000. With humanoid programs planning tens of thousands of robots a year — each needing 14+ screws — capacity, not performance, will decide who takes the first volume wave.

The price collapse is already happening

This is the number that changes everything: planetary roller screw pricing has moved from early 10,000-yuan sample levels to roughly 2,000 yuan for many specs in 2025, with some sizes quoted at a few hundred yuan, depending on spec, precision, batch and delivery. At the micro end, NOUS locked its 1.5mm screw under ¥100. At the humanoid program level, the arithmetic is simple: a robot like UWORLD U1 at ¥119,800 or Unitree's G1 at $13.5K cannot absorb $1,350-2,700 Swiss screws in every joint — so every humanoid maker in China is dual-sourcing, and Chinese suppliers are getting design wins by the quarter.

None of this means the Swiss incumbents are finished. GSA and Rollvis carry 40+ years of field data, the yield, and the marquee customers; the entire premium industrial world still specs them. But the economics of a consumer-priced humanoid are brutal, and the direction of travel is unmistakable: the roller screw is following the same arc as RV reducers, harmonic drives and force sensors — from import monopoly to Chinese commodity in about three years.

What it means for buyers and investors

For a humanoid maker, the strategy is dual-source and qualify now: keep a Swiss/EU primary for the first production batches, but put Chinese screws through full cycle testing immediately — 2026-2027 capacity is being booked now, and the ¥2,000 price point is real margin. For an investor, the pattern from every prior component war applies: the winner will be the company that converts a design win into disclosed volume at high yield, not the one with the best sample. Watch Shuanglin's 100,000-set line ramp, Beite's ¥1.85 billion facility, NOUS ROBOT's automotive-robot dual engine, and which of them lands in a named humanoid program's published BOM. The Swiss monopoly is not broken yet — but it is now priced in years, not decades.