On May 7, 2026, Bloomberg published the headline conclusion of a fresh Morgan Stanley research note on China’s humanoid-robot industry: China’s early lead in humanoids will help drive the next leg of its global manufacturing and export dominance, the way EVs and batteries did in the previous cycle. The same note, dug into one paragraph deeper, contained the dryly funnier finding: of the Chinese corporate buyers Morgan Stanley actually surveyed, only 23 percent said they were satisfied with the humanoid products available today.
Both things are true at the same time. The supply-side story is a Chinese industrial-policy victory in slow motion. The demand-side story is a customer base saying almost-but-not-yet with a polite cough.
The piece is worth pulling apart, because it is the most precise public picture we have of where the China humanoid market actually is in May 2026 — and what is going to happen in the next 18 months.
The numbers
The Morgan Stanley survey, led by China industrials analyst Sheng Zhong, polled Chinese corporate buyers across manufacturing, logistics, and consumer-services categories. The findings, in the order they hit:
- 62% of respondents said they are likely to adopt humanoid robots in some form within three years.
- Only ~10% are currently evaluating or running pilot projects.
- Only 23% of buyers who have actually used or evaluated current products said they were satisfied.
- 92% said the unit price needs to fall below ¥200,000 (~$28,000) before mass-adoption math works.
- 150+ Chinese humanoid-robot companies are now operating in the market.
- ~14,000 humanoid units were delivered in China in 2025.
- Morgan Stanley’s 2026 forecast: 28,000 units (+133% YoY).
- China holds ~90% of the global humanoid market by both unit shipments and supply-chain capacity, per the same note.
The shape that comes out of those numbers is unambiguous. The intent is high. The current product is not. The price has to come down by roughly 2-3× before the addressable market broadens. And there are roughly an order of magnitude too many vendors chasing the available demand.
Sheng Zhong’s own framing — quoted in the note — is that “2026 is a critical year as humanoid integrators strive to reach commercialization and build up their ecosystems,” with an explicit warning of a coming shake-out as IPOs and venture capital outpace actual end-customer demand.
Why both stories are true at once
The Bloomberg headline (“humanoid robots will drive next leg of China export dominance”) and the actual survey finding (“only 23% are satisfied”) are not in tension. They are sequential. The 2018-2024 China EV playbook ran the same way:
- Stage 1: Many companies, generous capital, weak product-market fit, low buyer satisfaction.
- Stage 2: Vicious price competition forces engineering breakthroughs and supply-chain efficiency.
- Stage 3: A small number of survivors emerge with cost structures that make the technology globally exportable.
- Stage 4: Export dominance.
The May 7 note is essentially Morgan Stanley pattern-matching against that playbook and saying: this is what early-stage looks like. The 23% satisfaction number is the Stage-1 signature, not a refutation of the export thesis. It is what you would expect to see in 2026 if China is going to dominate humanoids in 2030 the way it dominates EVs in 2025.
What that pattern-match misses, of course, is that Stage 2 is brutal for the 80% of companies that don’t survive it. The note is internally consistent on this — it explicitly forecasts a shake-out window in 2026 as IPOs outpace demand. The math is straightforward: 150+ vendors, ~14,000 units of demand in 2025, ~28,000 units in 2026. Even at the optimistic forecast, the total addressable market for the year is on the order of 200 units per vendor on average. Most of these companies will not exist in 2028.
The price wall is the real story
The 23% satisfaction number is a soft datum — buyers will say all kinds of things in surveys. The hard datum is the ¥200,000 price ceiling. Ninety-two percent of respondents converged on that number as the threshold below which mass adoption becomes mathematically possible.
Where are current Chinese humanoids priced?
- Unitree’s H2 commercial (full-size humanoid): $40,900 (~¥295,000).
- Agibot’s G2 (wheeled humanoid for industry): mid-six-figure RMB.
- XPeng IRON: not yet public, expected to land in the same range.
- Apptronik Apollo (US): ~$50,000 list, more in practice with services.
- Figure 03 (US): not publicly priced, target unit-economics around $30k–$40k at scale.
Unitree’s R1 budget line starts at $4,290 — but that is an entry-level R&D platform, not an industrial humanoid that does machine tending or material handling at a price-per-shift that beats human labor.
The gap from $40,000 down to ¥200,000 (~$28,000) is roughly 30%. Hitting that price for a real full-size humanoid (≥1.5m, full hands, multi-hour battery, certified for factory floor) within 18 months is the actual technical-and-supply-chain question facing Chinese vendors. Schaeffler’s Q1 humanoid commitments — 45 partners, 30 prototype orders, 5 series-production contracts — and Hexagon’s AEON deployment at Fill Maschinenbau are the European tier-1 supplier ecosystem putting hard P&L numbers on this question. The Morgan Stanley survey is the buyer side of that same conversation, and the buyer side is saying: get cheaper, work better, then talk to us.
Who survives the shake-out
The note doesn’t name names, but the shape of the survivors is implicit in the data:
- Vendors with mass-production cost structures already in motion: Unitree (5,500+ units in 2025), Agibot (~5,168 units), Leju (Guangdong 10k/yr line opened March 2026).
- Vendors with deep-pocketed strategic backers that can subsidize the price-down curve: Tesla Optimus (vertical), Figure (Nvidia/Intel/Qualcomm/Salesforce-backed), 1X (OpenAI), XPeng (carmaker-funded).
- Vendors with specific industrial-customer wedges already locked: Boston Dynamics-Hyundai (Atlas), Apptronik-Mercedes/GXO/Jabil (Apollo), AEON-BMW/Schaeffler/Fill (machine tending).
The vendors most exposed to the shake-out are the 100+ Chinese startups with no production volume, no strategic deep-pocket, and no anchor industrial customer — which is, per Morgan Stanley’s headcount, most of them. Some will be acquired into the Unitree/Agibot/Leju spheres of influence. Most will quietly close.
What to watch through Q4
- The 2026 unit number. Morgan Stanley’s 28,000-unit forecast is the central estimate. If actual deliveries land closer to 20,000, the shake-out is faster and harsher than baseline. If they land at 35,000+, the demand-side may be ahead of the satisfaction number’s implication.
- Price prints from Unitree H3 / R2 and Agibot G3. Whether the next-generation full-size models breach ¥200,000 is the single biggest variable for the 2027 demand curve. Watch the Q3/Q4 announcement cadence.
- The first announced Chinese humanoid IPO failure or down round. This is the canary for the shake-out actually starting. Watch for it in Q3.
- Western buyer-side surveys. Morgan Stanley ran this on the China side; the equivalent number on the US/EU side is not yet public. If US satisfaction is also in the 20s, the issue is product, not market. If it’s higher, the issue is the specific Chinese vendor field.
The honest summary
Two sentences from the May 7 note that are doing all the work:
- From the Bloomberg-summarized version: China’s humanoid lead is the next manufacturing/export wedge.
- From the survey body: the products that buyers can actually purchase in May 2026 are not yet good enough, at the price they’re sold at, for most buyers to be satisfied.
Both true. The first is the 2030 statement. The second is the 2026 statement. Between them sits the shake-out — the year in which several dozen Chinese humanoid vendors that exist today will not exist in 2028, and the survivors will compress 30-50% of unit cost out of the bill of materials so the ¥200,000 number becomes real. The 23% satisfaction figure is not a problem with the thesis. It is the price of the thesis.
The thing to remember about the China EV story is that it took six years between “buyers aren’t satisfied yet” (2018) and “BYD outsells Tesla” (2024). The Morgan Stanley note is implicitly betting humanoids run the same arc on a similar clock. The next 18 months are the first compressing leg.