The same market, in the same month, produced two contradictory datasets — and both are accurate.
According to a November 16 report by China Business Journal (中国经营报) citing IT Juzi (IT桔子) data, embodied intelligence deals in November stood at 15 financing rounds worth RMB4.01 billion (about US$557 million) as of November 11 — a fraction of September's 60 rounds and RMB11.03 billion (about US$1.53 billion), or October's 33 rounds and RMB10.11 billion (about US$1.40 billion). Meanwhile, IT Juzi's "death company library" has quietly added names since September: Yinghe Robotics (盈合机器人), Smart Orange Dynamics (智橙动力, SMORO-BOT) and Wujing Dynamics (无境动力).
The fall from the first half's pace is steep. IT Juzi recorded 322 embodied-AI rounds worth RMB93.5 billion (about US$13 billion) in the first six months of 2026 — versus RMB57 billion (about US$7.9 billion) for all of 2025.
A Full Order Book, an Empty Payroll
The Wujing Dynamics case, as documented by the Journal, deserves a longer look, because it breaks the usual narrative of failures. Founded February 11, 2026, the company signed nearly RMB500 million (about US$69 million) in framework orders with several central state-owned enterprises within four months, entered delivery, held close to another RMB500 million in regional pipeline orders, and had raised RMB110 million (about US$15 million).
On paper: a going concern with a full order book. In practice: employees told media they had not been paid since August, and the company has been listed among the fallen. The lesson is uncomfortable — framework orders and revenue bookings are not the same as cash flow, and in a hardware business where delivery consumes working capital, an impressive contract wall can coexist with payroll failure. Investors reading deal flow announcements are, in effect, reading the top of an income statement whose bottom they cannot see.
From "Face Coverage" to "Point Concentration"
Why has deal count fallen while headline rounds like UniX AI's RMB500 million (about US$69 million) Pre-A and RoboScience's nine-figure A2 round still get done? The Journal's most quoted analyst, Li Chenwei of Wanlian Securities, offered a structural read:
- Exit channels have lengthened. The previous wave was supported by "listable-in-theory plus technology narrative." When exits take longer and diligence shifts from technology foresight to commercial delivery proof, small concept-stage projects lose funding first.
- The head pattern is set. Leading players have emerged and new-project supply is shrinking, so round count falls — not because the industry fundamentals turned, but because there are fewer new things to fund.
- Single-deal intensity is rising. Even as breadth cools, individual checks are getting bigger, flowing toward leaders and toward the bottleneck layers — perception, compute, data and models — rather than whole-robot integration.
Li's summary line captures the shift: this is not a bubble-versus-no-bubble question but a repricing, in which companies with revenue, delivery and repeat customers will now get fundamentally different valuations from companies with only a story.
The Founder's Version
Zhang Zhengtao, founder and chairman of Zhongke Huiling (中科慧灵), put the founder-side view to the Journal: "The heat in the primary market is still there, but not every company can get money anymore." Capital, he said, is visibly concentrating into leaders and companies with proven scenario delivery.
Other voices in the report fill in the commercial logic. Zhang Yun, global CEO of Ries Strategy, dated the cooling to mid-year: capital rushed the track early, but embodied AI has not yet reached maturity, and money ultimately follows demonstrated value. Aitu Technology CEO Zhang Hanwen was blunter about order quality — estimating that genuinely repeat orders in the sector run under 1 percent — and argued that industrial customers, who calculate payback on line retrofits and equipment purchases, are the buyers where real cash flow exists; his own garment-sewing robot claims payback periods as short as a year in some scenarios.
The consensus the report builds: the industry is not retreating, it is repricing. Capital that once spread across the whole category is converging on demonstrable commercial closure — and the companies that could only fundraise are now discovering which side of that line they were on.
Honest limitations
- Monthly data-provider counts (IT Juzi, Gaogong Robotics) use different methodologies and cutoffs; the numbers indicate direction, not precision. IT Juzi's November figure covers only 11 days of the month, which mechanically depresses it.
- The Wujing Dynamics wage-arrears allegations rest on employee statements and unanswered calls from the Journal's reporters; the company's side is not on record.
- Analyst quotes are individual sell-side and industry views, not market consensus.
- "Death company library" listings mark reported distress, not confirmed liquidations; corporate statuses change.
What readers can do now
- If you invest in the sector, ask two questions of any funded embodied-AI company: what is the repeat-order rate, and what is the cash conversion cycle — the Wujing case shows neither appears in funding announcements.
- Distinguish two declining metrics: deal count, which reflects fewer new entrants and a set head pattern, versus total capital, which is concentrating upward; they signal opposite things.
- Watch the bottleneck-layer funding — edge compute, data, perception, models — as the cleaner read on where the industry believes value will accrue, since that is where the larger checks are now landing.
