Ling Universe Raises 200 Million in New Funding to Build an AI World Operating System

"A leap forward." By Qian Ren

Edited by Zhiyan Chen

AnYong Waves has learned that Ling Universe (灵宇宙) recently completed a RMB 200 million Pre-A funding round, with participation from Guofang Innovation (under Shanghai International Group), Guotai Haitong, GF Xinde, Didi Chuxing, Koala Fund (under Lakala), Runjian Shares, and other financial institutions and listed companies, while existing shareholders made additional investments beyond their pro rata rights.

Ling Universe is a human-computer interaction intelligence technology company. Its founder, Jiawei Gu, has long been a familiar name in China's venture capital circles. Gu earned his bachelor's degree at Tongji University before being admitted to Tsinghua University for graduate studies. After graduating in 2010, he joined Microsoft Research Asia, then in 2014 moved to Baidu's Institute of Deep Learning as head of human-computer interaction. At 28, he was among the first cohort of Baidu's "Young Marshal Program" for AI talent, leading the development of frontier AI products including BaiduEye, DuBike, and DuLight.

After leaving Baidu in 2016, Gu spent time building out robotics industry investments at an A-share listed company while also founding Ling Technology (物灵科技). The company gained prominence through its AI reading companion robot Luka, which became an early consumer-grade smart hardware product to go global at a time when mobile tool apps dominated the overseas push. To date, Luka has sold nearly 10 million units across 18 countries.

"If AI is going to enter people's lives, it has to start by understanding humans in home and on-person scenarios — and globalization is the natural direction for this category of products," Gu told AnYong Waves. During the Luka era, capital was abundant and everyone was rushing into robotics. But the underlying models weren't ready, hardware was too expensive, and building end-user devices meant "walking a tightrope against the wind." Now everything has changed: models have matured, foundational costs have come down, and interaction paradigms have shifted. "AI is finally entering the physical world."

Based on this assessment, Gu founded Ling Universe after the arrival of large models — not as a simple second startup, but as a relaunch of his decade-long commitment to the core philosophy of "animism in all things" (万物有灵) on a new foundation built for the large model era. Rather than pursuing another single hit product, the aim is to build an operating system ecosystem for next-generation AI terminals, giving machines true "spirituality" in perception, empathy, and proactive interaction.

At six months old, Ling Universe completed three consecutive funding rounds with shareholders including 37 Interactive Entertainment, Xueda Education, SenseTime, Glory Ventures, Linear Capital, Tsinghua SEE Capital, and Yinxingu Capital, spanning large models, semiconductors, education, and entertainment. In 2025, it launched its first product, the portable AI terminal "Ling Universe Cube," which ranked first in AI toys during the 618 shopping festival. During the recent Double 11 period, sales grew more than 230% compared to 618.

But AI hardware is not Gu's ultimate goal. He wants to push validated models toward scale — moving beyond companion machines to build a human-computer interaction system that can operate in households around the world: from the single-point breakthrough of AI-assisted reading to a systematic embodied intelligence experiment for global families.

Ling Universe's key engine is its self-developed LingOS interactive operating system. LingOS's core value lies in its transferability — it is not firmware locked to specific hardware, but an "AI spirituality" or "robot soul" that can be infused into terminals of different forms, from portable devices to home consumer robots, continuously evolving through real-world interaction data it collects. In Gu's vision, LingOS will not be limited to intelligent responses in single scenarios, but will become a "universal AI spirituality interface" that transcends geography, culture, and age.

In Gu's view, "hardware is just the interface; the system is the core." And from a development path perspective, the value of that system needs to be validated and amplified in larger markets — especially overseas markets.

As the company announced this new funding round, AnYong Waves sat down with Gu at Ling Universe's Beijing headquarters.

The following conversation has been edited by AnYong Waves —

Part 01

The South Slope Route

AnYong: More than a year has passed since your last round. Compared to fundraising in the early days of the company, what is the biggest change at Ling Universe?

Gu: We're no longer just a hardware company. Previously, investors were buying into my track record of "having built Luka and sold nearly 10 million units" — that was proven consumer-grade AI deployment experience, plus imagination around AI terminalization. What we're building isn't AI toys — we're attempting an AI embodied intelligence "south slope route," and our investors recognize this.

To put this judgment more bluntly: look at the Little Genius watch — RMB 2 billion in annual profit, 12 million units sold. This category will definitely be rebuilt in the AI era. And among companies in this category, we're the one that most resembles a "new-form AI terminal entry point."

AnYong: How do you understand the "south slope route"?

Gu: It's like climbing Mount Everest — you can take the north slope or the south slope, one is a leap, the other is gradual.

The north slope route is what many peers are taking: race to build a humanoid demo, even if it can't be deployed yet, show the future form first. The south slope route is what Tesla took: deploy hardware first, accumulate real-world data, iterate models, then feed that back into products.

We chose the second path. Because the core of embodied intelligence isn't the demo, it's "world understanding." You need real data volume, sufficiently long sequences, and enough household scenarios before you can even talk about embodied intelligence. Some people didn't understand this at first, but once products started shipping at scale and the data loop began working, they realized that by taking the south slope, we were the ones walking more steadily.

AnYong: Do investors view you as a "cash-flow consumer product" or an "AI-native hardware company"?

Gu: Investors actually fall into two categories. One type looks at whether you can make money immediately, whether you're consumer electronics. The other looks at whether you can build systematic capability, whether you're a next-generation entry point. The reason we can continue raising funds is that value is visible on both the cash flow and future ends: products sell, demand is real; and we're not a single product — we're building the next-generation real-world human-computer interaction system.

What I show them is actually quite simple: the nearly 10 million household data points accumulated during the Luka era; the world's largest multimodal dataset in children's picture book content (image + voice + text + emotion); low-latency technical capabilities in spatial understanding and intent recognition; the data loop of proactive interaction systems; and the cultural understanding built through overseas distribution and household interviews. These can't be bought with money — they're built up over ten years.

AnYong: Ling Technology, which you founded in 2016, also claimed to be building robots, but ultimately your best product was Luka. Why will Ling Universe be different this time?

Gu: The timing was genuinely too early then. The underlying foundation models weren't ready, hardware was expensive, and interaction was just simple command-response — far from achieving the sensory experience of "spirituality." I can feel that the present moment is completely different from that time. The underlying operating system has finally matured.

This maturity isn't a breakthrough in single technology, but a complete "ecological foundation" that can support AI-native terminals. The core characteristics are clear: First, multimodal large models are truly usable now — they can process text, image, and voice, and also understand dynamics and causality in the physical world. Second, AI infrastructure has fully taken shape: cloud computing costs have dropped dramatically, and edge NPUs can support large model local inference, allowing complex AI functions to run efficiently on consumer-grade hardware. Third, the "democratization" brought by open-source ecosystems — just as Android's openness spawned countless terminal innovations, open-source large models mean we don't have to build from zero at the foundation layer, and can focus on scenario-based innovation and interaction reconstruction.

Today, I believe we're standing at AI's "Android moment." I'm absolutely certain: the optimal window for AI terminals comes only once, and that time is now. Wait another two years, infrastructure becomes cheaper and easier to replicate, and you won't be able to build a first-mover advantage barrier.

In recent years, AI's explosion has been entirely inside screens — people open browsers, input prompts, get answers. But the relationship between humans and machines should actually be more direct, more natural. If AI only stays on screens, it will always be a "tool." But when it enters reality and perceives the world, it becomes humanity's "symbiotic intelligence."

AI has finally, truly begun to "land in the physical world." What Ling Universe wants to do is transform the world from a prompt (word as prompt) into an interface (world as interface). This isn't simple hardware upgrading — it's a paradigm revolution in human-computer interaction: in the future, it will no longer be "humans seeking services" but "services actively finding humans," and LingOS is the underlying protocol supporting this paradigm.

AnYong: How specifically will this be achieved?

Gu: Ling Universe is not a company that "sells devices" — it's a company that "builds interaction systems." We're creating a new operating system, LingOS, which is a closed-loop system that simultaneously supports two AI forms: one is "on-person AI," like the current Cube device; the other is "off-person AI," such as home consumer robots or desktop devices like Luka.

These two lines have a double-helix relationship: one handles perceiving and interacting with the world, the other handles executing operations in the world. Together they will form the next-generation human-computer interaction ecosystem. At the application layer, LingOS has an "Agent-Net" — we no longer need an App Store. Every AI character is an Agent that can be "added" rather than "downloaded." Parents can add a "Teacher Cao," kids can add an "Elon Musk" or a "Li Bai."

These intelligent agents proactively come to you — for example, if you've been learning English, it will push English stories; if you like science, it will recommend exploration tasks. From language companion, to reading companion, to family growth companion, and finally becoming an all-around helper for childcare and parenting.

AnYong: "Proactively coming to you" is quite interesting.

Gu: The future of AI isn't humans operating machines — it's machines proactively finding humans. For example, if you photograph a plant, it can automatically summon Darwin to explain it, or invite an AI Miyazaki to draw it as a manga; at a museum, it can make bronze artifacts "speak" and tell their stories. In October this year, we partnered with Zhejiang Provincial Museum, using Ling Universe's AI learning companion Cube as a digital docent and interactive experience officer — dozens of units rented daily, demand was extremely high. Because the experience of "artifacts speaking for themselves" is incomparable to traditional guided tours. The product roadmap we're following is essentially using LingOS's continuously evolving spatial interaction and intent understanding capabilities to materialize my original vision of "everything is teaching material, the world is the classroom" into this kind of revolutionary edutainment experience.

This philosophy also originated from Luka. Before Luka, children's smart devices were almost all single-modal — either voice intercom or image recognition. Luka was the first to fuse "camera + voice + multimodal interaction" together — it could read picture books, could accompany children in reading stories.

The pain point Luka solved was very real: in the evening, parents need to read picture books to their children, seven or eight times until their throats are hoarse. We realized AI could share this burden for parents, and thus the entire category was opened up. That was the first time we realized that AI products must land in deep scenarios.

AnYong: There must be more than one company with operating system ambitions. How do you make LingOS become "Android" rather than one of those dead nobodies?

Gu: We're not rushing to be "the operating system for the entire industry." Ling Universe will choose to prioritize making "our own closed loop" solid, continuously producing hit products that users recognize.

The ecological position of an operating system is never "declared" — it's "grown." Our thinking is to first tighten the "data-experience-commerce" loop of our own hardware, and after accumulating certain deterministic results, let embodied enterprises actively choose us. This is the key to gradual progress and surviving cycles. The foundation of an operating system is "solving real problems," not having empty technical architecture.

Part 02

"Globalization Was Always the Starting Point"

AnYong Waves: Global development is an important path for AI and embodied intelligence companies today. Your previous product, Luka, was among the earlier consumer-grade smart hardware products to go overseas. Does Ling Universe have globalization plans?

Gu: This funding round is a turning point — it allows us to push validated models toward scale. Not just this year, but for years to come, the keyword will be globalization.

Going forward, I hope our overseas sales exceed 60%. I believe LingOS can become a globally common interaction layer, regardless of hardware form — any AI terminal, including glasses, robots, and so on, can connect.

AnYong: How did Luka's globalization begin back then?

Gu: Actually from day one, we never saw ourselves as a company only for the domestic market. Going overseas wasn't a sudden decision, but a natural extension based on demand, product, and validation — every step was taken solidly.

While Luka was still in R&D, we simultaneously planned multilingual content and global compliance. After validating core value domestically, we selected 30+ localization partners — such as industry leaders in Germany, maternal and infant channel operators in Southeast Asia — leveraging their local resources to rapidly reach users and reduce market education costs.

Luka solved a pain point that transcends culture: parents' time and ability to accompany children in reading is limited. Luka ultimately entered households in 18 countries, 60+ cities. In markets including the US, South Korea, Singapore, and Germany, Luka has become many families' first AI partner — not because it's a high-tech Chinese product, but because it's a "companion" that genuinely shares parenting pressure and stimulates children's reading interest. The value it provides is universal and perceptible.

Children need empathy; parents need liberation. This demand isn't limited to China. On this foundation, Ling Universe embeds the capability of "being understood" at the system level, with the goal of making AI operate globally.

AnYong: How do you judge which country should be the first stop for overseas expansion?

Gu: I don't think it's GDP, population, or friendliness — it's real demand density. For example, picture book reading was originally imported from abroad, with decades of history in Europe, America, and Japan, where family education is more mature. Or take AI enlightenment education — many countries have stronger demand than China, because they see it as a new foundational capability.

So when we select countries, we mainly look at three dimensions: AI enlightenment demand density, alignment with family education philosophy, and "fit" with culture and interaction patterns.

I'll sit in living rooms with local parents and children, observing how they interact with technology. Actually any country is possible — the key is starting with the first batch of users, gradually moving toward mass market, and discerning in the details whether they truly care about the value you provide.

AnYong: Why is the household scenario so important?

Gu: The household is the most universal usage scenario. No matter the country, parents want their children to grow up happily and well, and they want technology to help share the burden. Luka validated this, and Ling Universe extends this logic to a broader scope — household interaction includes not just learning, but also emotional companionship, life collaboration, entertainment and creation, and more.

We want devices to understand "the language of home." This understanding isn't keyword triggering, but the system mastering rhythm, habits, and tone through continuous learning. When it understands a household sufficiently, it qualifies to become a "family member." Intelligence doesn't mean high IQ — it means empathy.

AnYong: Under massive global cultural differences, can empathy still be achieved?

Gu: That's a good question. Cultural difference is the biggest challenge for AI going overseas.

We used to say AI is universal, but we've actually found that cultural differences have enormous impact on interaction. For example, in educational content, American children are more accustomed to inquiry-based questioning, while Asian families are more results-oriented. So we adjust AI interaction strategies in different versions — some are more like exploration guides, others more like teaching assistants.

Voice recognition is equally complex. While the system supports multiple languages, accents, speech rates, and semantic habits all require independent training. This means we must build a cultural adaptation layer beyond algorithms. What LingOS is doing is giving the system cultural understanding capability — recognizing emotional expression, tone changes, and behavioral patterns. Understanding is the deepest capability for globalization.

AnYong: This sets a very high bar for training.

Gu: Luka's users represent nearly 10 million households, behind which are tens of billions of picture book reading and interaction data points — a larger data volume than America's Epic or Audible. Multimodal data including image, language, emotion, and behavior forms the foundation for our large model training. Now our system can learn and narrate a picture book it's never seen before on the fly. This data also gives us fast response speed: average latency of 200-300 milliseconds. We've built a three-layer intent understanding architecture at the foundation (gating + Lora optimization + RAG search enhancement), ensuring both speed and stability.

AnYong Waves: What is your ultimate goal?

Gu: My ultimate goal has always been "animism in all things, symbiosis between humans and machines."

In reality, humans have long relied on various technologies. Going forward, most humans will no longer have independent thinking — developing the habit of consulting AI on everything can make decisions more professional. AI dependence is the key distinguishing mark of future humans; the deeper the dependence, the more excellent. Our children as "AI natives" will be stronger than us, because from a young age they'll have the strongest intellectual advisor, delegating repetitive information screening and professional knowledge support to AI. AI is no longer a cold tool, but an inseparable "symbiotic partner" for new generations of humans.

I believe one day in the future, LingOS will empower all embodied devices and AI terminals — it can be installed in next-generation immersive AR glasses, making cultural artifacts "come alive"; it can empower humanoid robots to become understanding companions in the home. At that time, "animism in all things" will no longer be a slogan.

Image source | Unsplash

Recommended Reading

The flow of money, the rise and fall of people