A robot company with a German headquarters and Chinese roots just raised the largest embodied-AI check of the week — and the interesting part is who wrote it and what they were buying.
On November 12, Agile Robots (思灵机器人), headquartered in Munich, announced a US$500 million Series F at an US$8.5 billion post-money valuation, led by SoftBank Group with Nvidia, Hitachi and T.Capital participating. The round takes the company's cumulative funding to about US$2 billion, with proceeds earmarked for capacity expansion and expansion into key markets — US market entry, more German production capacity, and preparation for a potential future IPO, per the company.
What Makes the Company Worth That Number
Agile Robots was founded in 2018 on technology out of the German Aerospace Center (DLR), and its core products are high-precision force-controlled robot arms, five-finger dexterous hands and medical robotics, running on the in-house AgileCore operating system. The positioning matters: rather than chasing the humanoid spotlight, the company has built its record in industrial manipulation — assembly, testing and precision tasks where force control, not bipedal walking, is the hard part. That is the profile Nvidia's investment points to: physical AI needs manipulators that can already do controlled work, and Agile's installed base in automotive and electronics manufacturing is exactly the kind of fleet such models learn from.
The US$8.5 billion valuation also reframes the humanoid-heavy conversation. The week's other large rounds went to companies building the layers under and around the robots — which is where the money's center of gravity is visibly shifting.
The Same Week, In China: Chips and Delivery Numbers
The funding roundups from Chinese securities weeklies (国联民生 and 国元 among them) tell a coherent story:
- Xixuan Chuangzhi (曦选创智), the embodied-AI edge-chip company co-founded by GPU maker MetaX (沐曦) and UBTech, closed a RMB500 million (about US$69 million) angel round led by GGV Capital and Matrix Partners China, at a RMB4 billion (about US$556 million) post-money valuation. The company is building a "big-brain-and-small-brain" robot SoC — integrating motion control, multimodal interaction and on-device inference — positioned against Nvidia's Thor, with a development platform planned for 2027 and a first chip targeted for 2028.
- UniX AI (优理奇) closed a RMB500 million (about US$69 million) Pre-A round on November 9, with Greenland Holdings (绿地集团) coming in as both strategic investor and downstream application partner across security, hotel and eldercare scenarios. The company reports monthly deliveries above 100 units since late 2025, peaking over 300 — commercial service robots shipping at volume, not lab demos.
- Qingtianzu (擎天租), a robot rental and services platform, was also listed in the weeklies as having closed a strategic round — part of a broader pattern in which deployment and operations businesses, not just robot makers, are raising.
Read together, the pattern is hard to miss. Capital is converging on two ends of the stack: the compute layer (edge SoCs that make embodied AI affordable at scale) and the commercialization layer (companies with auditable delivery and rental volume). The middle — yet another general-purpose humanoid chassis — is where financing has tightened.
The Physical-AI Logic Behind SoftBank's Check
SoftBank's lead, with Nvidia alongside, tracks the same thesis from the demand side. If embodied AI follows the trajectory the industry now expects, the bottleneck sequence runs: capable hardware, then data generated by deployed hardware, then models trained on that data. Companies already deploying precision manipulators at industrial scale sit on the first two. The investment case is less about a single robot product and more about who owns the interface between physical work and the models that will learn from it.
It also validates the ecosystem play around China's embodied-AI stack: an edge-chip joint venture backed by a GPU designer and a robot manufacturer (Xixuan), a deployment partner strategy by a delivery-stage robot maker (UniX), and a Munich-headquartered integrator with global ambitions (Agile) are different bets on the same eventual market — machines that do economically measurable work.
Honest limitations
- Round terms come from company announcements and securities-weekly roundups; final valuations and investor lists can shift between announcement and closing.
- UniX AI's delivery figures are company-disclosed, unaudited numbers.
- Xixuan Chuangzhi's SoC timeline (2027 platform, 2028 chip) is a plan stated at founding-stage, with execution risk typical of first-generation silicon.
- Agile Robots' IPO ambitions are preparation, not a filing; no exchange process has been announced.
What readers can do now
- Map the embodied-AI funding stack by layer — edge compute, manipulation-grade hardware, deployment services — rather than by robot form factor; this week's rounds suggest that is where pricing discipline now sits.
- Watch Nvidia's investment pattern as a signal: its strategic checks increasingly mark the companies whose fleets will feed physical-AI training loops.
- Track whether Xixuan's 2028 SoC keeps its date; a credible domestic edge chip against Thor would reshape cost assumptions for every Chinese robot maker.
