Robotics Shutdowns 2026: Funding Is Not Delivery Capability

In the first half of 2026, China’s embodied-AI sector raised about RMB 93.5 billion across 288 deals — nearly RMB 600 million a day. In the same six months, eight robotics companies went into bankruptcy, liquidation or shutdown, and a ninth, K-Scale Labs, had already folded three months earlier in Silicon Valley.

Both facts are real, and they are not in tension. They are the same fact seen from two ends: capital is abundant, commercial closure is not. For anyone sourcing robots or robot components, the casualty list is far more informative than the funding tally.

The list

Compiled from court announcements and public reporting; the tally is incomplete (more than 20 companies by one count) and “shutdown” spans formal liquidation, voluntary wind-down, bankruptcy review and asset disposal.

One detail deserves more attention than the list itself: most of these companies had raised a new round, or carried a well-known investor’s name, within a year of failing. Money was not the missing input.

Three ways to die

The exhibit mindset — DeepBlue. Court-supervised liquidation began in May 2026, with 102 employee claims totalling more than RMB 23 million; wage arrears had surfaced as early as 2023. DeepBlue ran cleaning, delivery and disinfection lines in parallel while its parent touched a dozen sectors and controlled 91 companies. The demos worked. The deployments did not: a disinfection order announced at RMB 700 million with LG delivered only a fraction once pandemic demand receded, and staff described units that “only worked as exhibits”, tripping on site details that never appear in a booth. The parent booked RMB 120 million of revenue in 2022 against a net loss of RMB 92 million. Capital was converted into capacity and prototypes, not cash flow.

Anchorless expansion — Yinghe. Five business lines, a announced RMB 10-billion industrial base, and a peak valuation of US$500 million. The filings tell the rest: zero social-insurance contributors at its Ji’an subsidiary in 2024; the Shunde base never entered production; full-year 2025 revenue of RMB 4.33 million against a net loss of RMB 67 million. Panda Capital’s Li Lun has described after-tax executive packages above RMB 3 million, RMB 28 million spent on office fit-out, and a RMB 90 million order with a 47-day delivery window that was breached and triggered claims. The company sized its organisation to its funding round rather than to its validated demand.

The process gap — K-Scale Labs. This one is the most instructive for component buyers. K-Scale shipped roughly ten K-Bot prototypes, each costing more than US$100,000 to build, against a launch price of US$8,999 for the first hundred units (list price US$16,000). Closing that gap required switching from machined parts to die casting — and the tooling money never arrived. The company raised about US$6.25 million across three rounds, failed to secure the US$10–15 million it sought in Q3 2025, and shut down in November with roughly US$400,000 left, refunding deposits and open-sourcing everything. Founder Benjamin Bolte’s summary is the cleanest sentence anyone has written about this phase of the industry: the supply-chain gap cannot be filled with enthusiasm and open source.

Different causes, identical ending — and the same sentence describes all three: raising money is not the same as building it; building it is not the same as selling it; selling it is not the same as surviving it.

Why capital cannot substitute for cash flow

Robotics is capital-heavy and slow. A funding round buys runway; it does not buy a business. Where cash generation is absent, each round is a transfusion, not a cure — and when the market stops believing, as it did for K-Scale in the autumn of 2025, the runway ends within weeks.

The institutions tracking this are not optimistic. CAICT expects the embodied-AI sector to complete its first shakeout through market competition within two to three years. Xinding Capital has put it more sharply: of roughly a thousand humanoid robot companies in China, perhaps ten to twenty will remain.

An eight-point supplier due-diligence checklist

The list above is public information. The harder question is what a buyer checks before a supplier appears on it. Eight things worth verifying, all of them obtainable from public filings, a site visit, or a contract clause:

1. Headcount against social-insurance records. Declared team size is a slide; social-insurance contributors are a filing. Yinghe’s Ji’an subsidiary showed zero.
2. Accepted deliveries, not signed intent. Ask for the acceptance certificate numbers of the last three installations. An announced order is marketing until it is accepted on site.
3. Revenue against cash burn. Yinghe’s RMB 4.33 million of revenue against a RMB 67 million net loss is a fifteen-fold gap. Ask what funds the difference and for how long.
4. Executive cost structure. Compensation and non-operating spend out of line with revenue is a signal about how the last round was used.
5. Repeat-purchase rate and customer concentration. A single customer above roughly half of revenue is a red line; one-off demonstration orders are not revenue.
6. A manufacturing route to volume, named in writing. Which parts move from machining to die casting or stamping, at what annual volume, funded how. K-Scale never got past this question.
7. Spares commitment in years, not intentions. Write the support window into the contract. K-Scale’s customers got refunds; they did not get a parts supply.
8. Tooling ownership and retrieval rights. If the supplier fails, who holds the moulds, fixtures and test rigs for your part — and can you collect them?

Protecting the order itself

Due diligence reduces the probability of a bad outcome; contract structure limits the damage when it happens anyway.

• Milestone payments, not mobilisation fees. Tie advances to acceptance events rather than to order signing.
• Retention held past the warranty period. A percentage released only after a defined run-hours threshold.
• Escrow for drawings, firmware and test procedures. For custom parts, a third party holds what you need to re-source.
• Tooling clauses that survive insolvency. Moulds paid for by the buyer should not enter the estate.

None of this is unusual in mature manufacturing. It is unusual in robotics, where most buyers are still treating a young vendor the way they would treat an established one — and where, by the sector’s own forecasts, most of those vendors will not be established in three years.

Sources: shutdown figures and company details from CMR Industry Alliance, New Strategy Mobile Robot Industry Research Institute, court announcements and public reporting compiled by the WeChat account 铁卫士 (19 October 2026); K-Scale Labs details from 36Kr and Humanoids Daily; sector shakeout forecasts from CAICT and Xinding Capital. Company names are romanised from Chinese; revenue and loss figures follow the source’s basis of preparation. The due-diligence checklist and contract recommendations are Latest Robot News editorial.

Related reading

• Embodied AI funding, September 2026: big rounds cool, money moves upstream — the other half of the picture: where capital went when it stopped chasing platforms.
• Robot component sourcing: from prototype to production — qualifying a supplier once the numbers check out.
• Humanoid robot market H1 2026: 36,000 units, top five all Chinese — the volume these vendors were competing for.