How Chinese is Tesla's Optimus, really?

The short answer

Direct answer: Less confidently than the headlines suggest. The widely repeated "~70% Chinese" figure is an estimate by Chinese securities firms of Optimus's Gen-3 bill-of-materials value — not a Tesla disclosure, and not audited. Tesla names no Optimus suppliers; two of the most-cited "suppliers" have publicly denied their reported orders. The structural dependency is real — rare-earth magnets above all — but the confirmed supplier map is far thinner than the concept-stock lists imply (the lists of shares traders buy on a theme before any real orders are confirmed). For a buyer, the risk lives in the bill of materials, not the badge — the brand name on the outside.

~90%China's share of global rare-earth magnet processing — the one genuinely hard dependency in a humanoid's bill of materials, with no near-term substituteSource: McKinsey; CNBC, 2025

Where does "70% Chinese" come from?

From Chinese sell-side research, repeated across retail media — not from Tesla. Tesla does not disclose Optimus suppliers, and its Chinese partners are under non-disclosure agreements; in their own filings they reference "a major North American customer," never Tesla by name. The "~70%" figure is a blended estimate by Chinese securities firms (CITIC, Kaiyuan, Guotai Junan among them) of the share of the Gen-3 bill-of-materials value — not component count — that Chinese suppliers would take. It is directional, not audited.

The number that gives it weight is a cost estimate, widely reported and attributed to Morgan Stanley (via the South China Morning Post): re-sourcing an Optimus Gen-2 unit away from China would raise its bill of materials from roughly $46,000 to about $131,000 — a 2.8× jump, driven mostly by the mechanical "body" parts. That figure is a reported estimate, not a directly published Morgan Stanley document, so it belongs in quotation marks around the source, not stated as fact. But its direction is not seriously disputed: the Chinese cluster is where humanoid parts are cheap.

$46K → $131KEstimated Optimus Gen-2 bill of materials if every Chinese part were re-sourced elsewhere — a ~2.8× increase (reported, attributed to Morgan Stanley via SCMP)Source: Morgan Stanley, as reported by SCMP, 2026

What's actually confirmed — and what's concept-stock?

Very little is confirmed. Here the reading has to be careful, because the same names appear in two different registers: as real, disclosed Tesla car suppliers building genuine robot-component capability, and as robot suppliers — where the evidence thins to analyst inference and stock-market positioning.

What is solid: several of these firms are real, verified Tesla EV suppliers with disclosed robot-component divisions and capacity investments. Tuopu (chassis, since 2016) has a robot-actuator division; Sanhua is a confirmed Tesla EV thermal supplier; Xusheng has supplied Tesla castings for over a decade. That much is not fiction. What is not confirmed is the leap from "capable EV supplier building robot lines" to "confirmed Optimus supplier at X% share." That leap is made almost entirely by securities analysts and concept-stock traders — the shares people buy on a theme before any real orders are confirmed — and two data points contradict the hype:

• Sanhua issued a clarification denying a reported roughly $685M Optimus actuator order (it did not deny that a Tesla collaboration exists under NDA — but the headline order figure is company-refuted).
• Wolong Electric Drive walked back its own Tesla-order signal: after a reported ~200,000-motor Optimus order drove its stock up, Wolong clarified that its Tesla work is limited to industrial-robotics production-line automation — "a small proportion" of its business — not humanoid motor supply.

And Elon Musk himself has said the Optimus ramp forced Tesla to insource nearly every component — which is hard to reconcile with a "70% externally China-sourced" story. Both can be partly true; neither is settled.

So the only honest way to present the supplier map is as reported candidates, with the denials shown as prominently as the claims:

None of these is a confirmed design win — a part actually selected into the product. A "confirmed 70% China-sourced Optimus" headline repeats securities-desk marketing as fact.

Where the dependency is real — and where it isn't

This is the distinction that separates analysis from alarm. China's position in the humanoid supply base is genuine but uneven, and the difference decides how much a buyer should worry.

The hard dependency is rare-earth magnets. China processes roughly 90% of the world's rare-earth magnets and produces the overwhelming majority of the high-performance NdFeB (neodymium) magnets that every humanoid's motors need — a chokepoint with no near-term substitute. It is not hypothetical: on Tesla's April 2025 earnings call, Musk said China's magnet export controls had held up Optimus, and that Beijing wanted end-use assurances:

"China wants assurances that these are not used for military purposes, which obviously they're not. They're just going into a humanoid robot."

— Elon Musk, Tesla Q1 2025 earnings call, April 23, 2025

Almost everything else is a cost choice, not a chokepoint. Reducers, bearings (China ~40%), motors (~35%), power electronics — alternatives exist in Japan, Korea, Germany, and the US. China leads on price, not exclusivity; Chinese harmonic reducers run roughly 40–60% below the Japanese incumbents, which is why they win on cost. A buyer should treat those categories as second-sourceable — available from an alternative supplier at higher price — and reserve real concern for the magnet content.

One number worth correcting while we're here: reducers are often called "a third of a humanoid's bill of materials." That is an industrial-robot figure. For Optimus, the Chinese-brokerage BOM breakdown puts actuators at roughly 35%, screws 18%, reducers 15%, motors 12%, and sensors plus everything else 20% — with about 14 harmonic reducers in the body (28 body actuators: 14 rotary, 14 linear). Reducers are ~15% of an Optimus, not a third.

Is this just a Tesla thing?

No — and this is where a fair assessment differs from an easy attack on Tesla. The dependency is structural to the whole field: China holds roughly 63% of the humanoid mechanical-component base by value, on Morgan Stanley's estimate. But the amount varies sharply, and the leaders are diverging:

• Boston Dynamics (Atlas) is building on a Korean supply base — Hyundai Mobis actuators — positioning away from China.
• 1X (Neo) builds motors, transmissions, and sensors largely in-house in California.
• Figure is vertically integrating hands and actuators, but still buys from the same regional suppliers its Chinese competitors use, at lower volume.
• Agility (Digit) and Apptronik (Apollo) assemble in the US, but their component provenance isn't publicly documented — so we don't claim it.

"Every Western robot is secretly Chinese" is false; Atlas and Neo disprove it. The accurate statement is that provenance ranges from near-total (Optimus) to deliberately minimal (Atlas), and the map keeps changing — Tesla is deepening Optimus's China sourcing near-term even as it strips Chinese parts from its US cars, two separate stories that are easy to conflate.

What should a buyer, lessor, or insurer do about it?

The reason this matters to Rovara's readers is a gap in how the rules work. US restricted-list regimes — the FCC Covered List, the DoD's 1260H list of Chinese military-linked companies, the proposed GUARD Act — target where a robot is assembled and which finished units or entities are listed. They do not, by default, reach the Chinese content inside a Western-branded robot. With one exception that already applies in practice: if a specific component supplier is itself listed.

That is not hypothetical. RoboSense and Hesai — two Chinese LiDAR makers — are on the 1260H list. A US-branded robot built with their sensors carries listed-entity exposure that the brand name on the outside hides. The rule keys on the brand; the risk lives in the bill of materials.

For a buyer, lessor, or insurer, that turns into a short, concrete checklist:

1. Map the badge against the BOM. Country of assembly is not country of content. Ask for the origin of the cost-dominant subsystems: actuators, reducers, motors and magnets, sensors and LiDAR.
2. Screen the actual suppliers against the live lists — 1260H, the FCC Covered List, the Entity List — not just the brand.
3. Separate chokepoint from cost-choice. Rare-earth magnet content is a hard dependency; most of the rest is second-sourceable. Ask whether the vendor has a qualified second source, not just where it buys today.
4. Price serviceability and lead time. A single-sourced Chinese joint plus export licensing is a tail risk — a rare but severe risk — to uptime and residual value; a part's resale-limiting question is whether it is re-buyable, not just how worn it is.
5. Re-check at every model refresh. A 2026 BOM is not a 2027 BOM; reshoring (domestic magnets, Korean actuators) is real and ongoing.

FAQ

Is Tesla's Optimus really 70% Chinese? The '~70%' figure is an estimate by Chinese securities firms of the share of Optimus's Gen-3 bill-of-materials value that comes from Chinese suppliers — not a Tesla disclosure and not an audited number. Tesla names no Optimus suppliers, and its Chinese partners reference only 'a major North American customer' under non-disclosure agreements. Treat it as a directional broker estimate, not a fact.

Which Chinese companies actually supply Tesla's Optimus? None is confirmed by Tesla or the companies themselves. Chinese securities research names candidates — Tuopu and Sanhua for actuators, Leaderdrive for harmonic reducers, Moons' and Zhaowei for hand components — but these are reported/candidate relationships, not confirmed design wins. Two named firms have since qualified their reported orders: Sanhua denied a roughly $685M order figure, and Wolong clarified that its Tesla work is industrial-robotics automation, not humanoid motor supply.

Where is Optimus genuinely dependent on China? Rare-earth magnets are the hard dependency: China processes roughly 90% of the world's rare-earth magnets, with no near-term substitute, and Elon Musk publicly said China's 2025 export controls held up Optimus. Most other categories — reducers, bearings, motors — are cheaper from China but second-sourceable from Japan, Korea, Germany, or the US. China leads on cost there, not exclusivity.

Do other Western humanoids also depend on China? It varies. The dependency is structural to the industry — China holds roughly 63% of the humanoid mechanical-component base by value — but the leaders diverge. Boston Dynamics is building Atlas on a Korean supply base (Hyundai Mobis actuators), and 1X builds Neo largely in-house in California. Provenance ranges from near-total (Optimus) to deliberately minimal. 'All Western robots are secretly Chinese' is false.

Why does component provenance matter to a robot buyer? US restricted-list rules target where a robot is assembled and which finished units or entities are listed — not the Chinese content inside a Western-branded robot, unless a specific supplier is itself listed. That already applies in practice: RoboSense and Hesai, two Chinese LiDAR makers, are on the Pentagon's 1260H list, so a US-branded robot using their sensors carries listed-entity exposure the brand name hides. Verify the bill of materials, not the brand.

Sources

The "~70%" estimate and Optimus BOM breakdown: Chinese securities-firm research via 36Kr and BigGo Finance — blended sell-side estimates, treated as directional. Cost penalty: attributed to Morgan Stanley via SCMP, as reported by Tech Digest and Interesting Engineering. Supplier clarifications: Bamboo Works — Sanhua denies the order figure; Wolong's reported ~200,000-motor order and its later clarification (min.news; Pandaily). Rare-earth magnets and Musk's quote: The Register — Q1 2025 earnings call; McKinsey — humanoid supply chain. Other humanoids: Korea Herald — Boston Dynamics/Hyundai Mobis; Forbes — US makers reducing China dependence. 1260H LiDAR listings (RoboSense and Hesai): WilmerHale; HSF Kramer. Component-count and status figures are drawn from securities research and trade trackers, not Tesla filings; and because Optimus volumes have not reached the levels these projections assume, any figure derived from unit counts should be read as provisional.