Mid-term results of Quncore Technology (00068): Revenue from new AI applications surged 177%, and the card is the target of scarce physical AI in Hong Kong stocks

Qun Nuclear Technology (00068) handed over its first interim report card after listing: revenue of 405 million yuan, gross margin climbed to a historical peak of 83%, adjusted net profit of 55.42 million yuan, an increase of 211% over the previous year, and is close to the level of the full year of last year. The market is more concerned that in the first half of the year, the company's revenue from new AI applications and products increased 177% year-on-year.

This means that Qun Nuclear Technology's spatial intelligence infrastructure strategy is continuing to be implemented, and the commercialization process has been implemented at an accelerated pace.

Chairman Huang Xiaohuang said, “Spatial intelligence is a key puzzle for AGI, and it is also a bridge for AI from the digital world to the physical world. In the past three years, Qun Nuclear Technology has advanced from the spatial intelligence application layer to the bottom of the technical infrastructure. In the future, we will continue to increase investment in world model research and development, continue to broaden application boundaries, and be a 'water seller' for infrastructure in the era of spatial intelligence.”

1. AI applications have become the second growth curve, the digital low-level circuit of the physical world of cards

Specifically, LuxReal, a professional AI video creation platform launched by the company in May, is connected to SpatialGen, a self-developed spatial generation model, and focuses on the long-standing “spatial consistency” pain points in AI videos — that is, coherent images and self-consistent physical rules. The month after launch (June), its registered users increased by more than 270% month-on-month, revenue increased by 470% month-on-month, and the double climbing speed of users and payments verified the initial establishment of product-market matching (PMF). This is also a typical example of spatial intelligence as an underlying ability that empowers new scenarios.

Furthermore, the order amount for SpatialVerse in the first half of the year, which serves physical AI training, has exceeded that of the whole of last year. SpatialVerse has reached cooperation with domestic physical AI companies such as Zhiyuan Robotics, Galaxy GM, Qiongche Intelligence, and Hesai, and has formed close industrial and academic cooperation with many overseas tech giants. The core value of this product is to generate “high-fidelity, physically-correct synthetic virtual data”, which is currently the scarcest “fuel” for models and physical intelligence training in the world.

Unlike language models that rely on Internet text, physical AI requires massive amounts of 3D environmental data that conforms to mechanical, optical, and collision rules, and the 3D data and 3D reconstruction and generation capabilities accumulated over many years of core constituted a first-mover advantage and competitive barrier in the physical AI era.

The strategic significance of this business is not short-term revenue, but rather the “data infrastructure” of the next generation of AI. The core of the group has entered the lower level of the digitization of the physical world.

In addition, the company integrated low-level capabilities such as space reconstruction, space generation, space editing, and spatial understanding on the Aholo spatial intelligence platform and opened it up to the public in the form of products and models, breaking out of industry restrictions and further allowing spatial intelligence to enter thousands of industries such as entertainment media, smart body training, cultural tourism, industrial twins, games, and space design.

2. The “Computing Power - Model - Application” flywheel is accelerated, or included in Hong Kong Stock Connect in early September

What supports these is the company's “computing power-model-application” full-stack architecture. Quncore Technology built its own GPU cluster, processing an average of 12.9 million computing tasks per day in the first half of the year, forming a computing power base. During the reporting period, the company's capital expenditure increased 124% year on year. This signal was viewed by some short-term investors as an erosion of profits, but from a strategic perspective, it is an inevitable choice to “put computing power first.”

In terms of models, in May, the company's large spatial intelligence model completed the generative AI service filing. It is one of the first large spatial models in China to complete compliance filing. Recently, Qun Nuclear Technology launched the 3D generation model Lux3D, initiated internal testing of next-generation world model products, and further enriched the model matrix. Currently, the Hong Kong stock AI circuit focuses on the big language model application layer. The “3D explicit route” world model direction pioneered by Group Core provides the market with scarce differentiated assets.

For the first time, the company disclosed “average daily token consumption” — 2.4 billion. It's not that impressive numerically, but its rapid rise means that whether it's an internal product or an external API call, the model inference layer is forming large-scale traffic entrances. This corresponds to the “water seller” logic of the core infrastructure layer: it not only serves as a technical foundation for internal products, but also provides computing power, models, and data services to all enterprises that require spatial intelligence capabilities. If the token billing model succeeds, it will become another revenue engine after software subscriptions.

From computing power to models to applications, data accelerates layer by layer in the flywheel. The company's CFO Shen Bei summed it up as a self-reinforcing cycle of “applying massive 3D data, data feeds back model training, and the model further empowers the product”.

In terms of capital markets, Qunnac Technology was recently officially selected as a constituent stock of the Hang Seng Composite Index. The relevant adjustments will officially take effect on September 7, 2026. Inclusion in the Hang Seng Composite Index is a core prerequisite for entering the Hong Kong Stock Index, which means that improvements in passive capital allocation and southward liquidity will arrive simultaneously.

Of course, there are also risks: spatial intelligence is still in the early stages of the industry; the Token business model needs to verify the will to pay on a large scale; competition among overseas giants in the world model field is also intensifying. But at least at the current point, Qun Nuclear Technology has used a solid track record to prove that spatial intelligence is no longer a long-term vision, but is being transformed into a quantifiable, compoundable, and priceable commercial reality. For the Hong Kong stock market, this is probably where the real scarce value of “the first global spatial intelligence stock” lies.