GLOW embodied AI model gets $74M boost as Noin Intelligence readies KNOWIN-X1 home robot

Noin Intelligence, a Chinese startup founded by a former Huawei generative-model lead, has closed a 500 million yuan (US$74.1 million) Angel++ round announced this week to fund the GLOW embodied AI model and its first home robot, KNOWIN-X1. Matrix Partners China led the round. Redpoint Ventures, SenseTime Guoxiang Capital, Walden International and L2F Lighthouse Founders' Fund also participated, and Lighthouse Capital acted as exclusive financial advisor.

Founded in August 2025, Noin also operates under the English name Knowin. Its founder, Li Yinchuan, previously led generative-model work at Huawei's Noah's Ark Lab. Noin employs more than 190 people, 90 percent of them graduates of top-tier universities, and plans to use the proceeds for global recruitment, full-stack R&D, product validation and the push toward mass production.

The Angel++ label is a Chinese convention for a late angel-stage round, and closing one barely a year after founding tracks with how quickly the team has scaled. The investor mix is another sign of how mainstream the embodied-AI thesis has become: Matrix Partners China leading at the angel stage, with SenseTime Guoxiang Capital, Walden International and L2F Lighthouse in the cap table, places the deal inside the Chinese AI investment establishment.

Inside the GLOW embodied AI model

The GLOW embodied AI model runs as a closed loop of three subsystems. A data-generation module called KnowinDream feeds a spatial-perception layer, KnowinWorld, which in turn supplies the task planner, KnowinBrain/Agent. Keeping data, perception and planning inside one pipeline is the point of the design, because the seams between those layers are where household robots typically fail.

A closed loop also speeds up iteration: improvements in the planner feed back into data generation, which retrains spatial understanding, in a cycle that open pipelines cannot reproduce. That matters for the timeline, because the model has well over a year of development ahead before KNOWIN-X1 is expected to ship.

KNOWIN-X1, the first product built on the system, is a 40cm foldable robot with dual arms offering 23 degrees of freedom and a combined 6kg payload. It works in lying, kneeling and standing positions and is built around what the company calls "cultivation-style" interaction, in which owners demonstrate chores and the robot learns them instead of executing fixed routines. The robot is approaching engineering verification, the stage where hardware startups test mechanical reliability and manufacturing feasibility before committing to tooling, with a target price in the tens of thousands of RMB and a planned 2027 release.

A funding wave with a software thesis

The round lands in a heavy stretch for Chinese robotics capital. PokeBot closed a nine-figure Pre-A round this month, Discover Robotics took in more than $100 million, and Zeroth raised $74 million in Pre-Series A funding in July for its home-companion humanoid Jupiter. China-based robotics companies had already pulled in $5.6 billion across 176 deals through mid-May, a pace matching the 2021 peak, and developers released 13 embodied AI models in June alone. Competition in the category now centers on software intelligence instead of hardware specs, and IPO momentum in the sector is building, which gives late-stage backers a clearer exit path.

The investor thesis behind much of that money is cross-embodiment intelligence, or models that transfer skills between robot bodies. FieldAI, Skild AI, Physical Intelligence, Generalist AI, Dyna Robotics and Rhoda AI have all raised around that idea this year, and the funding cluster now extends to X Square Robot, Spirit AI, GigaAI, RobotEra, EngineAI, LimX Dynamics and Noetix Robotics, each of which has closed meaningful late-stage capital. On the data side, Noitom Robotics raised several hundred million yuan to sell embodied-AI training data, a reminder that real-world collection is the bottleneck Noin's synthetic pipeline is designed to bypass. Most of the recent deals sit at the early stage: an Angel++ round for Noin, a Pre-A for PokeBot and a Pre-Series A for Zeroth, with Discover Robotics' nine-figure raise the exception.

The trade-offs of a 40cm robot

Size is the strategic bet. A 40cm foldable frame costs less to build, ship and store than a full-size humanoid, which lets Noin target the tens-of-thousands-of-RMB price band. That sits well below the consumer tier occupied by robots such as 1X's NEO, priced at $20,000 or $499 per month with a $200 deposit, and it changes the risk profile of the company, because a compact machine is also a limited one.

The contrast with the full-size tier runs deeper than price. Norway's 1X, backed by OpenAI, is raising around $1 billion at a $10 billion valuation to scale NEO's data collection and production, and Zeroth is positioning Jupiter as a long-term home companion. Noin is attacking the same market from the small end with a fraction of that capital, and the entire wager rests on whether the GLOW embodied AI model can generalize a task from a few demonstrations in an unfamiliar home.

The limits are concrete. Twenty-three degrees of freedom and a 6kg payload cap what the hardware can physically accomplish, so the product's value depends almost entirely on the software. There is also a cost question inside the closed loop: generating synthetic data in KnowinDream and grounding it in KnowinWorld avoids expensive real-world collection, but household chores are long-tail and environment-specific. The cultivation-style model shifts the training burden onto owners, which keeps data costs low while asking a great deal of the first customers.

The 2027 timing carries its own risk: the round has to fund more than a year of development with no product revenue, which is why the company is prioritizing mass-production readiness over near-term sales. Li's background adds a talent dimension. Recruiting a former generative-model lead from Huawei's Noah's Ark Lab is evidence that Chinese robotics startups now compete for LLM researchers, the same pool that powered the country's large-model boom.

For buyers, the practical takeaway is the price anchor: if KNOWIN-X1 lands near the tens-of-thousands-of-RMB range, it undercuts full-size humanoids by a wide margin and becomes the first credible consumer-scale option in the category. For the industry, the round is another data point that Chinese embodied AI is consolidating around software differentiation rather than hardware alone. The next milestone is engineering verification of the robot, followed by a 2027 launch that will test the cultivation-style thesis in real homes.

Why this matters

Noin is testing the cheapest credible route into home robotics: a compact, teachable machine backed by a closed-loop model rather than a full-size humanoid. If the GLOW embodied AI model generalizes from user demonstrations, the company lands at the price-performance point the category has lacked. Engineering verification of KNOWIN-X1 is the first checkpoint, and the 2027 release will show whether the closed loop holds up outside the lab.

Related Articles

• GigaAI Debuts SeeLight S1 as China's First General-Purpose Home Robot
• Generalist AI Funding Round Secures $400 Million to Advance Physical AGI
• OpenAI Smart Speaker: Donut Shape, Moving Parts, $300-$400 [Update]

✔Human Verified

Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.