Chinese humanoid robotics: funding drops 40% in one month, the shakeout begins

Chinese humanoid robotics: funding drops 40% in one month, the shakeout begins

🔎 The tap shut off in thirty days

For eighteen months, Chinese humanoid robotics ran like a money machine: 288 funding rounds, 226 companies, more than 46 billion yuan committed in the first half of 2026 alone. In September, everything came to an abrupt halt. Physical AI funding plunged 40.1% in a single month — from 19.4 to 11.6 billion yuan (NE Times, relayed by Seoul Economic Daily, October 27, 2026).

This is no market accident. It's a political decision. The CSRC, China's securities regulator, passed a simple directive to investment banks: no more IPOs for companies unable to prove sustainable revenue and a path to profitability (The Information, September 9, 2026).

The result boils down to three words: the shakeout begins. More than 150 manufacturers are now fighting over a market that Beijing just sorted out with a stroke of regulation. And while China freezes its mega-rounds, the West keeps funding in its own way — Machenta, a spinout from Imperial College London, just raised $40 million in seed funding for instant-learning robots.

The Essentials

• -40.1% in one month: physical AI funding drops from ¥1.94 B (August) to ¥1.16 B (September 2026), according to NE Times, as reported by Seoul Economic Daily and STCN (October 26).
• The CSRC locks down the market: "window guidance" issued September 9, 2026, requiring sustainable revenues, shrinking losses, and substantial technological innovation.
• Unitree, the spark: stock multiplied more than 5x at its Shanghai IPO, then down -55% from its peak (Reuters, September 21, 2026); IPO priced at 219 times earnings.
• Galaxea AI stalled: IPO timeline derailed despite a valuation of over ¥2 B (WSJ, September 11); Noetix files confidentially in Hong Kong (Bloomberg, October 22).
• The strong absorb: AgiBot raises ¥100–150 M at a valuation of around ¥5 B.
• This is not the death of the sector: early-stage deals are growing (24 vs. 19) and September remains above the monthly average of H1 2026 (~¥770 M).

Recommended tools

Following this shakeout requires sorting press releases from facts. Here are the four tools I use for my physical AI monitoring:

My advice: don't collect tools. A well-built monitoring agent and a publishing channel are enough — I detail the method at the end of the article.

September's numbers: a correction, not a crash

Chinese physical AI funding fell from 19.4 billion yuan in August to 11.6 billion in September 2026, a drop of -40.1% in thirty days. That's the headline number. The superficial reading — "Chinese robotics is collapsing" — would nevertheless be wrong.

Let's look at the details:

Three contradictory signals in the same table. Yes, the total is collapsing. But concentration is falling too: the ten largest projects captured 67.7% of the money in August, versus 59.4% in September. Translation: it's the mega-rounds that are freezing, not startup money.

And the most telling detail: the number of early-stage deals rose from 19 to 24 in one month. Chinese venture capital isn't fleeing physical AI. It's simply moving back down the chain, to where ticket sizes are small and valuations are still reasonable.

Now let's put September in perspective. With more than 46 Bn¥ disclosed in the first half of 2026 (IT Juzi), the monthly average was around 7.7 Bn¥. Even after its 40% drop, September remains above that average. We're correcting an overheating; we're not closing up shop.

STCN wrote on October 26 that "the funding frenzy around humanoid robot companies has entered a cooling phase." That's the right word: cooling. Crash would be a misuse of language — for now.

For Chinese founders, the consequence is immediate: raise early, raise small, prove revenue. The window for multi-billion-yuan rounds is closed for everyone except established leaders.

The CSRC Shuts Off the Stock Exchange Tap

Since September 2026, a "window guidance" from the CSRC has required investment banks to only bring to market companies capable of demonstrating three things: sustainable revenue, a shrinking loss trajectory, and substantial technological innovation (The Information, September 9, 2026).

For those unfamiliar with the system: a window guidance is an informal but binding instruction that the regulator passes on to banks. No legal text, no press release. A phone rings, and a door closes.

First documented victim: Galaxea AI. The company had filed a confidential application with the Hong Kong Stock Exchange after a Series B+ round in April 2026 that valued it at over 20 billion yuan, with Ant Group, Meituan, and Hillhouse Capital among its shareholders (Bytivo). Its listing timeline derailed because it failed to meet the tightened criteria (WSJ, September 11, 2026).

Second move: Noetix Robotics, which filed confidentially for a Hong Kong listing (Bloomberg, October 22, 2026). The city is becoming the sector's pressure valve — a more permissive channel than Shanghai or Shenzhen for companies that don't fit within the CSRC's requirements.

Why Hong Kong? Because the exchange has different listing rules, attracts international capital, and offers funds a dollar-denominated exit. For a company blocked on the mainland, it's the obvious plan B.

Why the tightening? Reuters documented it on September 21: regulators want to protect retail investors, who have rushed into humanoid stocks. Beijing has seen what happens when a company valued at 219 times its earnings hits the stock market with millions of retail investors on board. The response was regulatory, not monetary.

My take: this is triage, Chinese-style — brutal, opaque, but remarkably effective. The West regulates through transparency; China regulates through the window. The message is the same on both sides: free money based on promises is over.

Unitree: the spark that lit the fire

It all started with the IPO of Unitree, the global leader in humanoids. And contrary to what you read everywhere, Unitree is not a demo company — it's the trigger of a confidence crisis.

The fundamentals are real:

(Source: Embodied AI Frontier, October 2026.)

Read those numbers again: 60% gross margin, 5,500 humanoids delivered, 591 million yuan in profit. Unitree is probably the most profitable humanoid robotics company in the world. The problem is not the company. The problem is the price.

219 times earnings is the price tag of a company expected to triple for ten years without a hitch. The stock more than quintupled on its first day of trading in Shanghai, then has slid 55% from its peak (Reuters, September 21, 2026). The market delivered its verdict in six weeks.

It was this volatility — millions of retail investors buying the top — that triggered the CSRC's window guidance. Unitree is not the cause of the cooling. It is the proof of it.

The shakeout: 150 companies, and an NDRC blowing the whistle on playtime

More than 150 Chinese companies claim to be humanoid manufacturers, and the NDRC itself has warned against blind investment and overcapacity. Mathematically, there is no room for 150 winners.

The sector has entered its most predictable phase: the purge. STCN talks about a frenzy that is cooling; Yicai and NE Times document the decline month by month; the NDRC, China's powerful planning commission, has warned against speculation on humanoids. When the Chinese state blows the whistle, public and semi-public money disappears first.

The state of play as of October 2026:

The investors' assessment is blunt. Zhang Zhikuan, of Co-Stone Capital, an investor in AgiBot: "the valuations of many physical AI companies have already reached high levels, and decisions on additional investments require greater caution." When a sector investor says that publicly, it means he has already reduced his positions.

But note the nuance: the shakeout does not hit everyone equally. AgiBot is raising an additional 1 to 1.5 billion yuan at a valuation of around 50 billion. The strong absorb the capital that the weak lose. The correction does not shrink the sector — it concentrates it.

My prediction: by the end of 2027, half of the current 150 companies will have pivoted, merged, or shut down. The survivors will sell robots, not videos. Component makers — actuators, sensors, dexterous hands — will fare better than the demo assemblers.

I had set the scene in our analysis of the Automate 2026 reality check: China had already moved to mass production while the West was still lining up demos. The shakeout is the logical next step — you don't mass-produce 150 different brands.

And the West? Machenta raises $40M while Beijing sorts things out

While China freezes its mega-towers, the West continues to fund small-scale research: Machenta, a spinout from Imperial College London, has raised $40 million in seed funding for instant-learning robots (The AI Insider, October 27, 2026).

The contrast in scale is dizzying: $40M on one side, towers worth ¥1 to 1.5 billion on the other. But the models differ. Machenta is raising on a technological promise — robots that learn a task instantly, without months of training. China is raising on an industrial promise: mass-producing low-cost humanoids.

Should we conclude that the Chinese correction opens a window for Europe? No — and that's my most opinionated point in this article. China is correcting at ¥11.6 billion per month. Europe has never been at that level. A competitor that slows down by 40% still catches up with you, as long as it's running twice as fast.

The Chinese machine, moreover, remains intact on the technological front: the country has just released the first 100% domestic humanoid, with LimX and Kyland circumventing restrictions on components — I detail the implications in this article on the first 100% Chinese humanoid. The financial shakeout and the rise in technological autonomy are advancing at the same time. This is not a contradiction: it's the strategy.

What the West should take away from this: don't chase after complete hardware, but occupy the layers above — software, simulation, learning models, integration. That's exactly Machenta's niche. And the only one where $40M is still enough to matter.

How to follow this shakeout without getting burned

Three signals are worth more than a hundred press releases: units delivered, the Hong Kong window, and early-stage rounds. Everything else — demos, videos, keynotes — is noise.

Signal 1: actual deliveries. Unitree delivered more than 5,500 humanoids in 2025. That's the only kind of number that distinguishes a company from a video. Demand it for every company you follow.

Signal 2: the Hong Kong window. Who files, who passes, who gets blocked. Noetix filed confidentially; Galaxea was slowed down. Every decision by the Hong Kong Stock Exchange is a thermometer of the sector's real health.

Signal 3: early-stage deals. 24 versus 19 in one month: smart money is moving back down the chain. Follow where the seeds go, not where the mega-rounds go.

To automate this monitoring, I built an agent that runs 24/7 for €29/month — the full method is in this article on the €29/month AI agent. It scrapes funding announcements, IPO filings, and regulatory changes, then serves me a daily digest.

And if you want to turn this monitoring into income, the mechanics are the same as the ones I use: a website, a newsletter, consistency. I detailed the tools that earn me €300/month without coding in this article, and Hostinger remains my choice for hosting this type of site from ~€3/month (October 2026, check hostinger.fr).

One honest caveat: I don't give financial advice. I give signals. The difference is that signals survive the cycles.

❌ Common Mistakes

Mistake 1: confusing "funding decline" with "a dying sector"

The -40.1% should be read as a series, not in isolation: September remains above the H1 2026 average and early-stage deals are growing. The fix: always compare a crisis month to the six-month average, never to the previous record month.

Mistake 2: buying a humanoid stock because "China dominates"

Unitree has lost 55% from its peak despite having the sector's best fundamentals. If the global leader, with 60% gross margin, can collapse in six weeks, imagine the companies with no revenue. The fix: revenue, margin, deliveries — never the demo videos.

Mistake 3: believing the Chinese purge opens a window for Europe

The shakeout produces stronger survivors: AgiBot is raising at a ¥50B valuation while the small players die. Consolidation strengthens China, it does not weaken it. The fix: target software, simulation and integration, not full-stack hardware.

Mistake 4: mistaking demos for order books

More than 150 companies post spectacular videos; only one ships 5,500 units a year. The fix: demand units delivered and verifiable revenue before believing any announced growth trajectory.

❓ Frequently Asked Questions

Why has Chinese robotics funding dropped by 40%?

Because the CSRC has blocked investors' main exit route: the stock market. Its window guidance of September 9, 2026 requires sustainable revenue, a trajectory of shrinking losses, and substantial innovation. With no IPO prospect in sight, mega-rounds freeze up — hence the 40.1% drop in physical AI funding in a single month.

Is China's humanoid sector collapsing?

No, it's correcting. September 2026 (¥11.6 B) remains above the first half's monthly average (~¥7.7 B), and early-stage deals are up from 19 to 24. Capital is flowing back down the value chain rather than disappearing. This is selective sorting, not abandonment of the sector.

Is Unitree still a good deal?

The fundamentals are real: ¥1.71 B in revenue in 2025, ¥591 M in net profit, gross margin above 60%, and more than 5,500 humanoids delivered. But at 219 times earnings, the stock had priced in ten years of perfect growth — hence the 55% drop from its peak. The risk is the price, not the company.

What is the CSRC's "window guidance"?

An informal but binding directive from China's securities regulator to investment banks: only bring to market companies capable of generating sustainable revenue, reducing their losses, and demonstrating substantial technological innovation. No public text, immediate enforcement — Galaxea AI felt the effects as early as September 11, 2026.

Which companies will come out as winners from the shakeout?

Those already selling robots: Unitree (5,500+ units delivered in 2025), AgiBot (¥1 to 1.5 B being raised at a ~¥50 B valuation) and component manufacturers. The losers will be the dozens of companies living off demos and subsidies. Hong Kong will remain the preferred exit route for the survivors.

When will we know the shakeout is over?

When a new humanoid IPO goes through in Shanghai or Shenzhen without a market meltdown, when mega-rounds return beyond the leaders, and when the number of active players stabilizes below 100. My estimate, as analysis and not certainty: not before late 2027, probably 2028.

✅ Conclusion

China is not turning its back on humanoids — it has decided that only those who sell robots deserve a stock market listing, and the -40.1% of September 2026 is the price of that discipline. To understand where Chinese mass production is heading, read our analysis of the Automate 2026 reality check, then start your monitoring: the shakeout is only just beginning.