NeuroAI NEUROAINEUROAI.SITE
China AI, reported clearly
1,300+ verified stories on models, robotics, BCI and the business of AI.
Browse stories
ESC

Robots raised RMB 90 billion in six months — and the term sheets now ask for order books

Chinese embodied-AI startups raised more than RMB 90 billion in H1 2026, and in the period to late May deal value ran more than triple a year earlier.

2026-11-16 · 1182 words · NeuroAI
Robots raised RMB 90 billion in six months — and the term sheets now ask for order books

AI² Robotics (智平方) and HKC (惠科) describe a display-panel line in Huizhou where a wheeled robot with two arms lifts a glass substrate onto a trolley, checks the placement, and rolls to the next station. It is not a demo, they say: it runs on a shift schedule, beside people paid by the panel.

The company behind it, AI² Robotics (智平方), signed a three-year plan in 2025 to put more than 1,000 such machines into the production sites of HKC (惠科) — an order it values at close to RMB 500 million (≈US$69M / HK$543M) and describes, citing a Morgan Stanley assessment, as the largest single order for productivity-grade robots. In June 2026, AI² closed a round of nearly RMB 5 billion (≈US$694M / HK$5.4B) at a valuation above RMB 20 billion (≈US$2.8B / HK$21.7B).

That deal explains most of what changed this year: the money no longer buys the demonstration. It buys the purchase order.

The size of the wave

By the count kept by ITJUZI (IT桔子), a Chinese startup database, domestic robotics companies closed 434 primary-market deals worth RMB 74.6 billion (≈US$10.4B) in the period through 26 May 2026 — against 235 deals and RMB 22.1 billion (≈US$3.1B) in the same window a year earlier. By the end of June, reporting by China Economic Net put cumulative embodied-intelligence (具身智能) financing above RMB 90 billion (≈US$12.5B) for the half.

Deal count roughly doubled. Deal value more than tripled. The money concentrated in later-stage companies.

The largest rounds:

  • AI² Robotics (智平方): close to RMB 5 billion in June at a valuation above RMB 20 billion, with investors spanning state funds, suppliers and Moutai (茅台).
  • Galbot (银河通用): RMB 2.5 billion in March 2026, backed by the National AI Industry Investment Fund (国家人工智能产业投资基金), Sinopec (中国石化) and CITIC Investment Holdings, at a post-money valuation above RMB 20 billion.
  • Star Era AI (星海图): a RMB 1 billion B round in February and a nearly RMB 2 billion B+ in April, taking its valuation to about RMB 20 billion.
  • X Square Robot (自变量机器人): four rounds from B to C completed by 30 June, more than 30 investors, post-money valuation above RMB 20 billion.
  • Robot Era (星动纪元): RMB 2.5 billion across two rounds inside two months, after starting 2026 already above a RMB 10 billion valuation.

Four of those five sit on the same number: RMB 20 billion. Robot Era, the fifth, started the year above RMB 10 billion and is not in that cluster. When four private companies in one sector converge on the same valuation within five months, that is less a coincidence than a pricing convention — a bracket investors use until someone lists and proves otherwise.

Who is actually writing the cheques

The investor mix shifted more than the totals did. In the 2025 cycle, the lead cheques were mostly financial venture capital. In 2026, three types of money arrived together.

Industrial capital now sits at the top of the table. Battery makers, display manufacturers, liquor groups, telecom carriers and internet platforms are investing in robots they may also buy from. For a supplier like HKC, a stake in AI² converts a customer relationship into a supply guarantee.

State and local guidance funds are the second layer. The National AI Industry Investment Fund appears repeatedly, alongside provincial vehicles such as Guangdong's AI fund and city-level platforms in Shenzhen, Hangzhou and Beijing. Their mandate is less about a 10x return and more about anchoring production capacity and supply chains in a specific region.

Financial VCs sit third. That ordering changes diligence: they cannot outbid strategic investors, so they compete on speed and on structures that strategics avoid.

The consequence is a filter: a founder with a good model but no factory customer has a much harder time in 2026 than in 2025, because the largest cheque-writers evaluate against their own procurement plans.

The evidence investors now demand

The due-diligence questions have moved from papers to production lines. Three data points recur in the 2026 rounds:

  • Units shipped, not units demonstrated. Robot Era says it began thousand-unit deliveries in the second quarter and runs in more than ten logistics centres, reaching about 85% of human efficiency in some tasks, around the clock. Company-supplied figures — the kind investors now underwrite.
  • Repeat orders. A single flagship deployment is treated as a pilot. The valuation argument needs a second customer in the same scenario.
  • Gross margin at the system level. Selling a robot at a loss to win a logo is acceptable in a seed round; at a RMB 20 billion valuation, investors ask what the service contract earns after installation.

That is why the HKC order matters. It is a three-year commitment in a single, brutal scenario — handling glass panels of varying sizes, colours and reflectivity — where classical automation fails. If the same configuration sells to a second panel maker, the story becomes a product. If it needs a bespoke engineering team each time, it stays a project.

The exit door opened, slightly

Late-stage money can underwrite these prices because an exit finally exists. Unitree (宇树科技) priced its STAR Market IPO in October 2026 at RMB 150.80 a share, raising about RMB 4.2 billion. LimX Dynamics (逐际动力) closed a pre-IPO round of nearly US$200 million (≈RMB 1.44 billion / HK$1.56 billion) at roughly a US$2.2 billion valuation (≈RMB 15.8 billion / HK$17.1 billion) in November. AgiBot (智元机器人) has been working toward a Hong Kong listing.

A listed comparable gives private investors a mark — and a ceiling. Once Unitree trades, every RMB 20 billion private valuation gets compared with a daily quoted price — and if the public market values humanoid revenue at a fraction of the private mark, the next round gets harder.

Honest limitations

The deal counts and totals come from ITJUZI as reported by Chinese state-affiliated financial media; the database relies on disclosed and reported rounds, and Chinese primary-market deals are frequently announced without amounts, so totals are floors rather than censuses. Valuations for private companies are company or investor statements, not audited figures, and post-money numbers can include liquidation preferences that make the headline number larger than the common equity value. The HKC order value, unit counts and efficiency percentages are company disclosures or investor-day claims; none has been independently verified. No company named here publishes audited segment profitability for its robot business. This is a description of capital flows, not investment advice.

What readers can do now

  • Discount the headline valuation by the structure. Ask whether a RMB 20 billion mark is post-money on preferred stock with a liquidation preference; in a down exit, common shareholders can end up far below it.
  • Follow the customer list, not the funding list. A robot company's next round is easier to predict from who is running its machines on a paid shift than from which fund signed the last cheque.
  • Watch the first post-IPO print. Unitree's trading multiple will reset price expectations across the entire private cohort within a quarter — including for companies that raised at RMB 20 billion and have not yet shipped a thousand units.

Related coverage

Cite this article

NeuroAI (脑机网), "Robots raised RMB 90 billion in six months — and the term sheets now ask for order books," 16 November 2026, https://neuroai.site/a/na-capital-embodied-funding-boom

Machine-readable: BibTeX · RIS · JSON · Markdown

More in “Capital & Markets” → · Back to home · Markdown version

China AI, reported clearly
1,300+ verified stories on models, robotics, BCI and the business of AI.
Browse stories