← Back to blog
NewsAbout 6 min read

China's Humanoid Robot Funding Didn't Collapse. It Consolidated.

Published Nov 7, 2026
China's Humanoid Robot Funding Didn't Collapse. It Consolidated.

The headline number out of China this week looks like a crash. Funding deals in embodied AI and humanoid robotics fell to 33 in October, worth a combined 10.1 billion yuan, about $1.4 billion. In September there were 60 deals worth 11 billion yuan. In June, the peak, 55 deals worth 30.9 billion.

The count nearly halved. The money barely moved.

That gap is the story, and most coverage of the IT Orange data, as reported by 21st Century Business Herald on October 31, led with the drop in deal count. Run the arithmetic and something else appears. Average deal size went from roughly 183 million yuan in September to about 306 million in October, a 67% increase. Compared with June's 562 million average, October is still down, but the direction of the last month is toward fewer, larger cheques.

What "delivery pricing" means in practice

The report's own framing is that the sector is moving from imagination pricing to delivery pricing. In plain terms: valuations used to be set against prototypes and concept demos. Now they are set against confirmed orders and repeat enterprise customers.

That is a different underwriting standard, and it changes which companies can raise. A team with a good demo and a paper on dexterous manipulation could raise a seed round in 2025. In October 2026, the same team needs a customer who has taken delivery twice.

The timing lines up with two other pressures. Publicly listed robotics names have performed poorly since listing, which removes the exit story that early-stage investors underwrite. And the Hong Kong exchange has tightened scrutiny of new listings, which removes the fast path to liquidity that several Chinese hardware companies were counting on.

The counterexample in the same month

Kuaishou's Kling, the video generation unit, closed a round of about 19.8 billion yuan, roughly $2.9 billion, on October 30, at a post-money valuation above 120 billion yuan. Investors included Alibaba, Tencent, Baidu, a national AI industry fund and several state-backed vehicles. Two investors missed payment deadlines and had their allocations reassigned.

Kling is not a robotics company, so the comparison is not apples to apples. But it shows where Chinese AI capital went in October: to units with visible revenue. Kling reported more than 850 million yuan in Q2 2026 revenue, up over 200% year on year, with an annualised run rate near $500 million and around 50,000 enterprise clients.

The pattern across both data points is the same. Capital is moving toward assets that can show a purchase order, and away from anything that cannot.

The first public benchmark arrived on November 5

Three days after the funding data, Agility Robotics completed its merger with Churchill Capital Corp XI and began trading on Nasdaq as AGLT. The transaction and related financing brought in roughly $465 million in gross proceeds, with Foxconn leading a $201 million private placement.

The filings that came with it are the first public financial statements for a humanoid robot company, and they are useful precisely because they are unglamorous. Agility reported about $1.78 million in net sales for 2025 against a net loss of roughly $138 million. Its Digit robots have logged more than 65,000 hours at customer sites. Digit 5 carries a three-year intent order for 1,000 units worth $300 million.

Read those three numbers together. A $300 million intent order against $1.78 million of trailing revenue is what a pre-scale hardware business looks like: commitments far ahead of shipments, and a burn rate that requires the commitments to convert.

Two stacked brass weights on pale marble, the smaller one casting a long shadow.

Public markets will now price humanoid robotics quarterly, on units shipped and cash consumed. Private investors in China will read those filings, because they are the only comparable set available.

Demand is not the problem

The same week, JD.com laid out its physical AI plan at the Wuzhen summit. Over five years, JD Logistics intends to purchase 3 million robots, 1 million unmanned vehicles and 100,000 drones. The company is building data collection centres and plans to capture more than 10 million hours of real human task video and over 1 million hours of robot operation data within two years, covering more than 100 production and service scenarios.

It also reported that its JoyAI-Sim simulation engine matches real-robot evaluation results about 90% of the time at 3.2 times the speed.

That is a buyer with a budget and a schedule. It is also a buyer that will specify requirements, negotiate price, and expect service. Which is exactly the condition that separates companies able to raise in this market from those that are not.

What the shift does to the supply chain

For Western buyers of Chinese robotics components, the repricing has a practical consequence that has nothing to do with valuations.

When capital was cheap and prototype-driven, component vendors could sell on specification. A harmonic reducer with better backlash numbers won the design slot. When capital is allocated against delivery, the customer is a systems integrator who has to hit a unit cost, and the questions change: what is your yield, what is your lead time, and can you hold a price for twelve months.

That pressure propagates. Companies that raised on a demo now have to fund production tooling out of a smaller round, and the ones that cannot are the acquisition targets for the survivors. Consolidation at the deal level usually shows up as consolidation at the vendor level about two quarters later.

There is also a demand-side read. The buyers placing the large orders are Chinese platforms and logistics operators, and they are specifying for their own workflows rather than for a general market. That produces robots optimised for tote handling and warehouse aisles, which is a real market and a narrower one than the pitch decks described.

What to watch

Three indicators will tell you whether October was a pause or a repricing.

The first is whether average deal size keeps climbing while deal count stays low. That would confirm consolidation rather than retreat.

The second is Agility's first two quarterly disclosures. If shipments ramp against the Digit 5 order, the public benchmark improves for everyone. If cash burn stays near $35 million a quarter with flat revenue, the private market will reprice faster than any analyst report.

The third is whether the Chinese state funds that anchored Kling's round extend the same logic to robotics. So far they have concentrated on AI software with revenue. The robotics thesis needs one portfolio company to show repeat orders at volume to earn the same treatment.

None of this is a verdict on the technology. The engineering progress in the last eighteen months has been real, and the deployment data from warehouse and logistics customers is the most credible evidence the sector has produced. What changed in October is the standard of proof.

For three years, the pitch was a video of a robot doing something a person used to do. The pitch that raises money now is an invoice.

Related articles