From renovation software to AI infrastructure: how “Qunhe Technology”’s 479 million model builds a moat with an 82% gross profit margin—what’s the business logic?

Ask AI · How does Manycore Technology leverage massive 3D data to empower the development of physical AI?

Amid the wave of companies vying for the capital markets in “Hangzhou Six Young Dragons,” Manycore Technology’s progress has drawn particular attention. According to the Hong Kong Exchanges and Clearing Limited (HKEX) disclosure on March 29, ManycoreTech Inc. (abbreviated as “Manycore Technology”) has submitted to HKEX for a listing hearing on the HKEX Main Board, with JPMorgan and Jianyin International as joint sponsors.

Manycore Technology was co-founded by Huang Xiaohuang, Chen Hang, and Zhu Hao in 2011. Since its establishment, it has received investments from well-known investment institutions including IDG Capital, Hillhouse Capital, Shunwei Capital, and Temasek. According to the information materials after the listing hearing, from 2023 to 2025, Manycore Technology recorded revenue of RMB 664 million, RMB 755 million, and RMB 820 million, respectively. In the same period, it reported annual losses of RMB 646 million, RMB 513 million, and RMB 428 million.

Data source: Public disclosures by Manycore Technology

Just this February, the company received the China Securities Regulatory Commission’s overseas issuance and listing filing notice (commonly known as the “route permit”). As the “top seed” for an IPO among this cluster of hard-tech enterprises, it is steadily advancing through the final gate for a Hong Kong stock listing, aiming to become the “world’s first space intelligence” company.

Its co-founder and CEO, Chen Hang, has a brilliant metaphor: Manycore Technology aspires to become the “water seller” of the physical AI era. Supporting this strategic ambition is the core asset moat the company has built over many years—by the end of 2025, it had accumulated more than 479 million 3D models and more than 500 million structured 3D spatial scenarios.

At the application layer, the company has already established strategic partnerships with leading players in embodied intelligence, such as Yuan Intelligent Robotics and Galaxy General. As NVIDIA founder Huang Renxun emphasized, “the next wave of AI is physical AI,” and the massive structured 3D data that Manycore Technology has mastered is the key foundational infrastructure to help physical AI move from grand blueprints to widespread deployment.

The “water seller” of the physical AI era: Building core assets with high barriers

In the fiercely competitive technology track, Manycore Technology’s role positioning goes beyond being a single product or solution provider; instead, it is committed to becoming an underlying enabler for the entire physical AI ecosystem.

Chen Hang’s “water seller” metaphor vividly reveals the company’s business philosophy: during a gold rush, adventurers who look for gold mines are replaced more frequently, but the “water seller” business that provides essential tools and services to all gold prospectors carries more enduring and stable value. The foundation of this strategic positioning lies in the company’s data assets accumulated through more than a decade of deep cultivation—assets that are almost impossible to replicate in the short term.

The core of these assets is massive and high-quality structured 3D data.

By the end of 2025, the company’s database already contained more than 479 million 3D models and more than 500 million structured 3D spatial scenarios. The preciousness of this data lies in its “physical correctness.”

Unlike image or video data commonly used on the internet for visual training, each model from Manycore Technology embeds rich physical attribute labels, such as the exact dimensions of objects, material density, structural load-bearing capacity, and motion hinge relationships. This means that AI models trained on these data can not only “see” an object’s shape, but also “understand” the object’s physical essence and the interaction rules in space. For example, if a robot needs to learn how to open a door, it not only needs to recognize the door’s image, but also needs to know the rotation principle of the door hinge, the force application points on the handle, and the inertia and weight of the door leaf.

This “physically correct” data is precisely the scarcest resource in the development of embodied intelligence today. Training a robot that can safely and efficiently operate in complex, unstructured real-world environments requires it to undergo massive, diverse trial-and-error experiences and learning in the virtual world. Building a simulation environment capable of accurately simulating physical laws such as gravity, friction, collisions, and material deformation is fundamentally based on structured 3D data annotated with precise physical parameters. With its long-term and deep industrial focus in space design, Manycore Technology has systematically accumulated a “data mine” of this kind, thereby securing an irreplaceable “water selling” position in the wave of physical AI.

A leap from design software to spatial intelligence infrastructure

The evolution trajectory of Manycore Technology clearly shows how a tech company can start from vertical applications, gradually upgrade its core capabilities into a foundational platform—and it also reflects the “lucky wind” opportunity bestowed by the times.

The company’s starting point is spatial design software. Its flagship product, “Kuhome,” has become one of the largest spatial design platforms in the world, covering more than 200 countries and regions. This business not only brings stable cash flow and a customer base to the company, but also inadvertently completes the most critical original accumulation for the physical AI era—by serving industries such as household interior decoration, commercial projects, and real estate, it generates a huge amount of 3D spatial data that correspond to the real world through design tools.

Precisely based on this deep accumulation, Manycore Technology has achieved a strategic upgrade from an application software provider to a spatial intelligence infrastructure provider. The company systematically builds a business flywheel of “spatial editing tools—spatial data—spatial large models.” The tools continuously generate and enrich data, and the data is used to train and optimize large models; in turn, more powerful models can empower the tools, improving their intelligence level, thereby forming an ever-strengthening closed loop. Guided by this strategic direction, in recent years the company has successively launched key platforms and models.

The company has built a spatial intelligence platform called SpatialVerse for robot training. The platform converts the massive spatial data accumulated from design software such as Kuhome into highly realistic, physically correct synthetic virtual datasets, specifically for simulation training for robots, AR/VR, and embodied intelligence systems. At the same time, the company released its self-developed spatial language model SpatialLM and spatial generation model SpatialGen, endowing machines with the ability to understand, reason about, and create 3D space. To open these capabilities to a broader ecosystem, the company launched the Aholo spatial intelligence open platform, which provides core capabilities such as spatial reconstruction, generation, editing, and understanding to the outside world in the form of APIs or SDKs. This series of deployments indicates that Manycore Technology has already formed a complete spatial intelligence technology stack covering data, platforms, models, and an open ecosystem—far beyond the role of an initial design software company.

Commercial validation and future imagination

According to its prospectus, the company’s revenue has maintained steady growth. In 2025, it achieved revenue of RMB 820 million, and its gross margin rose to 82.2%. Even more noteworthy is that in 2025 the company achieved a turnaround to profitability for adjusted net profit, proving that its core SaaS business has healthy self-sustaining cash-generating capabilities. The vast majority of its revenue comes from software subscriptions, especially enterprise customer subscriptions; the business model therefore has a high degree of certainty and sustainability.

However, the capital markets’ expectations for technology companies always center on growth potential. Although the company’s traditional spatial design software market remains firmly in the leading position, its scale is relatively limited. Therefore, the company is betting its future growth engine on the emerging field of spatial intelligence. Currently, its spatial intelligence business has entered an early harvest stage. By the end of 2025, the SpatialVerse business had gained 16 customers, achieving revenue of RMB 5.2 million, and carried out substantive cooperation with leading companies such as Yuan Intelligent Robotics. These collaborations are not simply tailored projects; they are based on deep integration of the company’s platform capabilities, validating the company’s value as a provider of a robot “virtual training ground.”

NVIDIA CEO Huang Renxun has said that physical AI defines “the next wave of AI,” a “physical AI” era that holds opportunities worth tens of trillions of dollars.

For physical AI to truly mature, it must solve the perception, decision-making, and interaction problems of intelligent agents in the physical world. The “physically correct” massive structured 3D data that Manycore Technology possesses, and the SpatialVerse platform built on it, are precisely the key infrastructure to meet this need. Cooperation with Yuan Intelligent Robotics and Galaxy General is only the beginning of this transformation.

With the booming development of industries such as embodied intelligence, autonomous driving, and virtual reality, demand for high-precision spatial simulation and training data will grow exponentially. With its early-mover advantages in building data and platform moats, Manycore Technology is positioned at a crucial node connecting the digital world and the physical world. The value of its “water seller” role is expected to continue to be amplified in the gold rush of physical AI. And the company’s ultimate room for growth will be confirmed by the market.