ILUVATAR COREX (Stock Code: 09903.HK) recently disclosed its first financial report since going public. While the company's revenue continues to show high growth, its annual loss has further intensified, accumulating over 3.2 billion yuan in losses over the past four years.
It was noted that the revenue growth for the company's inference product series was achieved by proactively reducing prices, sacrificing short-term profits, which subsequently led to a decline in the series' gross margin. Soaring research and development expenses, coupled with a sharp increase in share-based payment costs from equity incentives, further compressed profit margins. As business scale expanded, the company's inventory, trade receivables, and notes receivable rose simultaneously, negatively impacting its cash flow performance.
Facing a new phase in the AI industry driven by efficiency and the competition between general-purpose and specialized chip pathways, the company is seeking breakthroughs through hardware-software co-design and the deployment of products for new scenarios. However, its ability to balance generality with specialized efficiency and convert technological advantages into market advantages remains to be seen.
The gross margin for the inference product series declined. ILUVATAR COREX's general-purpose GPU products cover different needs across two core application scenarios: training and inference, corresponding to the "Tiange" and "Zhikai" product series respectively.
With growth in shipment volume and customer numbers, the company's revenue has shown an upward trend in recent years, increasing from 189 million yuan in 2022 to 540 million yuan in 2024. However, alongside revenue growth, the company's losses have continued to widen. Net losses for 2022, 2023, and 2024 were 554 million yuan, 817 million yuan, and 892 million yuan respectively.
Based on 2024 revenue data, ILUVATAR COREX holds approximately 0.3% market share in China's general-purpose GPU market, with market shares for both training and inference products each below 1%, indicating room for further market share improvement.
The loss situation has not improved since 2025. The financial report shows the company achieved revenue of 1.034 billion yuan in 2025, a year-on-year increase of 91.6%; however, the annual loss was 1.004 billion yuan, expanding by 12.5% year-on-year. Calculated cumulatively, the company's net losses since 2022 have reached 3.267 billion yuan.
By product line, both the "Tiange" and "Zhikai" series showed significant growth, with corresponding revenues of 584 million yuan and 339 million yuan, accounting for 56.5% and 32.8% of total revenue respectively.
It was observed that while revenue for the inference series products grew substantially, their gross margin declined. In 2025, due to the company's strategy of proactive price reductions to sacrifice short-term profit for market share gains and accelerated inventory sales, the gross margin for its "Zhikai" series products was 39.2%, a decrease of 7.5 percentage points year-on-year.
From a technological pathway perspective, ILUVATAR COREX has chosen a route deeply compatible with NVIDIA's CUDA ecosystem. The advantage lies in compatibility with mainstream AI frameworks and the CUDA ecosystem, lowering customer migration barriers and facilitating scalable deployment.
As key technological innovation deepens, customers are focusing not only on chip peak performance but more on effective computing power in practical business scenarios, migration costs, development experience, and ecosystem compatibility. The ability to deeply optimize and promptly adapt to mainstream AI frameworks and inference engines, provide comprehensive toolchains and developer support, and collaboratively build an open ecosystem with upstream and downstream industry players have become critical factors determining market position.
Notably, although ILUVATAR COREX's DeepSpark community has gathered over 600 mainstream algorithm models, its ecosystem maturity is difficult to rival CUDA's in the short term. Combining industry views, whether Moore Thread's MUSA or ILUVATAR COREX's software stack, while each is developing its own programming model, they have not yet formed a concerted force.
Apart from general-purpose GPU products, the company's AI computing power solutions business was significantly affected by project delivery cycles and fluctuations in the supply cycle of supporting equipment markets. Revenue for this segment fell 42.2% year-on-year to 96 million yuan.
High R&D expenditures are eroding profits. Due to the characteristics of the AI chip and general-purpose GPU market, which involve substantial upfront investment and long product commercialization cycles, losses are a common current reality for domestic GPU companies.
Persistently high R&D investment is the main reason for ILUVATAR COREX's losses. From 2022 to 2025, the company's R&D expenses were 457 million yuan, 616 million yuan, 773 million yuan, and 974 million yuan respectively, accounting for 241.8%, 213.1%, 143.2%, and 94.2% of revenue during the same periods.
In addition to R&D investment, the company's sales and distribution expenses, as well as administrative expenses, increased to varying degrees. In 2025, these two expense items were 152 million yuan and 482 million yuan, increasing by 23.9% and 87.3% year-on-year respectively. The company's revenue scale was insufficient to cover these expenses. ILUVATAR COREX attributed the significant increase in administrative expenses mainly to increased share-based payments to administrative personnel and growth in professional service fees.
The 2025 financial report shows that due to the implementation of equity incentives leading to recognition of large share-based payment expenses, the company's share-based payment costs increased from 248 million yuan in 2024 to 526 million yuan in 2025, a surge of 112%, further squeezing the company's profit margin. Excluding this factor, the company's adjusted net loss was 438 million yuan, narrowing by 32.1% year-on-year.
The company mentioned in its financial report that the AI industry currently faces challenges in large-scale implementation. The development of large models is no longer limited to simple parameter scaling but has entered a new phase driven by efficiency. How to significantly reduce training and inference costs while maintaining model capabilities has become the core proposition of industry competition. This shift places new demands on computing infrastructure.
Notably, regarding the choice between general-purpose and specialized chip pathways, the industry presents a pattern of "each having strengths and weaknesses, mutually contending." While general-purpose chips offer advantages like high flexibility, general architectures have limitations in cost-effectiveness for specific scenarios. Specialized chips, despite long development cycles and closed ecosystems, offer clear advantages in energy efficiency and unit cost for specific AI inference tasks.
In response to changing industry trends, ILUVATAR COREX is taking measures such as implementing hardware-software co-design and launching the "Tongyang Series" robot brain products targeting robotics and edge computing scenarios. This is seen externally as the company's attempt to find a balance between generality and specialized efficiency. However, whether it can successfully bridge the path between general-purpose and specialized approaches and achieve sustained conversion of technological advantages into market advantages still requires time to verify.
It was also noted that as business scale expanded, ILUVATAR COREX's inventory, trade receivables, and notes receivable climbed simultaneously, affecting the company's cash flow performance. In 2025, the amounts for these two indicators were 710 million yuan and 577 million yuan, increasing by 107% and 53% year-on-year respectively. During the same period, the net cash used in operating activities was -1.162 billion yuan, a decrease of 88.02% year-on-year, with cumulative negative cash flow exceeding 3.1 billion yuan over four years.
Comments