By RWT Robotics Editorial Team Published
CONFIRMED
Huawei Cloud’s own event page scheduled a November 16 livestream to introduce CloudRobo WM1.2, which the company describes as the first Ascend-native world model in its CloudRobo lineup. Separately, an AI Insider roundup headline claims Shanghai humanoid maker UniX AI raised a $75M pre-A round, a figure we could not match to any UniX AI announcement or funding database. One of these is a documented product push from the company itself. The other is, for now, a single unverified headline.
What we know
Start with the part that has a paper trail. Huawei Cloud’s official event page scheduled a livestream for November 16, 2026, to introduce CloudRobo WM1.2. Huawei describes it as the first Ascend-native world model in the CloudRobo lineup. According to the same page, the agenda covered the model’s architecture and an ecosystem push on Ascend cloud.
That is the full extent of what the source supports. We found no independent coverage of WM1.2’s technical details. The event page tells us what Huawei planned to talk about. It does not tell us how the model performs, and we have not reviewed what was actually presented.
Huawei Cloud’s CloudRobo product page fills in some of the surrounding stack. It lists embodied-AI models including CR-VLA-Dual, which Huawei describes as a dual-system vision-language-action model, and Lerobot-ACT, an imitation learning model. Those are the building blocks a robot developer would actually touch: a policy that maps camera input and instructions to actions, and a way to learn behaviors from demonstrations.
None of this came out of nowhere. Huawei Cloud released the Pangu World Model and the CloudRobo Embodied AI Platform at Huawei Developer Conference in June 2025. At that point CloudRobo also proposed an open “Robot to Cloud” (R2C) protocol, meant to standardize how robots connect to cloud-side models. WM1.2 is the latest step in that platform play, not a new direction.
On the UniX AI side, the background is thinner but real. UniX AI is a Shanghai-based humanoid maker with two product lines: the wheeled Wanda series and the bipedal Panther. Dealroom, a funding aggregator rather than an official source, lists a $28M round for the company in May 2025 and a $41M round in December 2025.
What is rumored
The $75M pre-A round rests on one item: a headline in an AI Insider roundup. We rate it grade D. AI Insider is a secondary aggregator here, and the claim appears only in a headline, with nothing we could trace back to a primary announcement.
We tried to match it. We found no UniX AI announcement of a $75M round. Dealroom does not show it. The funding history we could find for UniX AI shows no round of that size, and no round carrying the “pre-A” label.
One more detail makes us cautious. The only search result we found for a “$75M Pre-A” round belonged to Unitary Quantum, a different company with no connection to UniX AI. We cannot say whether the roundup headline confused the two. We can say that the one matching figure we found points somewhere else entirely.
Until UniX AI, an investor, or a major outlet with reporting of its own confirms the round, treat the $75M figure as unverified.
What we don’t know yet
- What WM1.2 actually does. Huawei has not published, in anything we found, benchmarks, training data, model size, or results on physical robots. “Ascend-native” is a claim about the platform the model runs on, not about how well it plans or predicts.
- What happened on stream. The event page shows the livestream was scheduled for November 16. We have not verified what was shown or said.
- Whether demos were autonomous. We reviewed no demo footage from Huawei or UniX AI. Whether any robot shown alongside these announcements was running autonomously or under teleoperation is unverified. In this field that is the first question to ask, because it changes what a demo proves.
- Whether R2C has traction. We found no data on how many robot makers, if any, have adopted the R2C protocol since June 2025.
- Whether the UniX round exists. If it does, we don’t know the investors, the valuation, or what the money is for.
What it means
The angle everyone wants is that Chinese humanoid and robot-brain funding keeps growing, even at the pre-A stage. Parts of that story have support. Others don’t yet.
The support: if Dealroom’s listings are accurate, UniX AI went from a $28M round to a $41M round within about seven months in 2025. That is a company raising larger checks in quick succession, which fits a hot market for Chinese humanoid makers. Dealroom is an aggregator, so treat even that as indicative rather than final.
The gap: a $75M round would extend that trend neatly, and that neatness is exactly why it needs checking. Right now it is one headline from a grade D aggregator, with no primary source and a stray “$75M Pre-A” match pointing at a quantum computing company. We are not going to build a funding trend on it.
The Huawei piece is more solid and arguably more important. Huawei is not building robots here. It is positioning its cloud and Ascend stack as the layer robot makers plug into: world models, VLA policies, imitation learning, and a connection protocol. If that works, many hardware companies would run their robot brains on Huawei infrastructure. That is a platform bet, and it is a bigger one than any single funding round.
The engineering caveat is the usual one. A world model is valuable only if it helps a physical robot plan or act better in the real world, measurably and repeatedly. So far, the public record for WM1.2 is an agenda. The things to watch for are published evaluations, results on real hardware, and named robot makers shipping on CloudRobo. Until then, this is a confirmed launch and an unconfirmed raise. Neither one is evidence of progress yet.
Credits and sources
By RWT Robotics Editorial Team
- Published









