--- title: "Iluvatar CoreAI's inference business surges 238.2%, with 1000+ project verifications, accelerating commercialization." type: "Topics" locale: "en" url: "https://longbridge.com/en/topics/39726711.md" description: "Tianshu Zhixin's inference business surged by 238.2%, validated by over 1,000 projects, accelerating the commercialization and implementation of computing power. The value of computing power lies not in the parameters themselves, but in practical applications and tangible results. As artificial intelligence transitions from technological breakthroughs to large-scale implementation, computing power is becoming the core engine driving the intelligent transformation of all industries. In this global AI race, Tianshu Zhixin (9903.HK), a leading Chinese provider of general-purpose GPU products and AI computing power solutions, has demonstrated strong growth momentum, highlighting the resilience and commercial potential of domestic computing power companies. The recently released 2025 annual performance announcement of Tianshu Zhixin shows..." datetime: "2026-04-06T10:18:39.000Z" locales: - [en](https://longbridge.com/en/topics/39726711.md) - [zh-CN](https://longbridge.com/zh-CN/topics/39726711.md) - [zh-HK](https://longbridge.com/zh-HK/topics/39726711.md) author: "[亚太商讯那些事儿](https://longbridge.com/en/profiles/18018177.md)" --- > Supported Languages: [简体中文](https://longbridge.com/zh-CN/topics/39726711.md) | [繁體中文](https://longbridge.com/zh-HK/topics/39726711.md) # Iluvatar CoreAI's inference business surges 238.2%, with 1000+ project verifications, accelerating commercialization. Iluvatar's Inference Business Soars 238.2%, Over 1,000 Projects Validated, Commercialization Accelerates The value of computing power lies not in the parameters themselves, but in practical application and creating tangible results. As artificial intelligence transitions from technological breakthroughs to large-scale deployment, computing power is becoming the core engine driving the intelligent transformation of countless industries. In this global AI race, Iluvatar CoreX (9903.HK), China's leading provider of general-purpose GPU products and AI computing power solutions, has demonstrated robust growth momentum, highlighting the resilience and commercial potential of domestic computing power companies. The company's recently released 2025 annual results announcement shows that its annual revenue reached RMB 10.34 billion, a sharp increase of 91.6% year-on-year; gross profit reached RMB 5.58 billion, up 110.5% year-on-year, with the gross profit margin rising from 49.1% to 54.0%. While revenue grew rapidly, the company's adjusted net loss narrowed significantly by 32.1% to RMB 4.38 billion, indicating continuous improvement in profitability. **Core Business Revenue Surges 149.6%, Training and Inference Advance Simultaneously** In 2025, the company's general-purpose GPU products experienced strong growth, becoming the core driver of overall performance. During the reporting period, this segment contributed revenue of RMB 9.23 billion, a substantial increase of 149.6% year-on-year, with its share of total revenue climbing to 89.3%. Among this, revenue from the Tiange series for AI training scenarios reached RMB 5.84 billion, up 116.7% year-on-year; revenue from the Zhikai series, focused on cloud and edge inference, reached RMB 3.39 billion, a surge of 238.2% year-on-year, reflecting the market's strong recognition of the performance and cost-effectiveness of the company's self-developed GPU products. Building on this foundation, Iluvatar continues to invest in architectural innovation, driving computing and performance upgrades. In the training domain, the next-generation Tiange series will continue its R&D as planned, optimizing computational density, memory bandwidth, and cluster scaling efficiency for large-scale training scenarios. In the inference domain, the next-generation Zhikai products are under active development to meet the demands for cost-effectiveness, latency, and throughput across diverse inference scenarios from edge to cloud. Thanks to the continuous improvement in product capabilities, the company's project deployment pace has accelerated. By the end of 2025, Iluvatar's products and solutions had cumulatively been deployed in over 1,000 projects across key industries such as financial services, healthcare, and transportation, serving more than 340 clients, with the scale of commercialization continuing to expand. **Sustained Increase in R&D Investment, Significant Progress in Software Ecosystem Development** Alongside business expansion and steady growth in customer applications, the company continues to invest in core technologies such as large model inference PD separation, MoE large model training, mixed-precision quantization, and DSL programming languages, helping clients maximize the release of hardware computing power efficiency in real business scenarios. In 2025, Iluvatar's R&D investment reached RMB 9.74 billion, a 26.1% increase year-on-year. The R&D team expanded to over 530 members, of which more than one-third have over ten years of industry experience, and over 70% hold master's degrees or higher. High-intensity R&D investment is accelerating the transformation into competitive advantages. The company's newly released next-generation software development platform natively supports mainstream GPU programming models. After deep optimization, it achieves over 80% improvement in code migration efficiency, supporting "minute-level" smooth migration of operators. The self-developed core acceleration library system for all scenarios shows a general performance improvement of over 20% compared to the previous version. Simultaneously, a large model-specific acceleration toolkit, featuring a core equipped with lossless quantization technology, achieves a performance leap across the entire process while fully preserving model accuracy. Long-text processing efficiency is improved by over 50%, system computing power utilization is increased by over 60%, and chip computing power utilization ("MFU") reaches industry-leading levels. This means clients can not only obtain high-performance chips but also convert computing power into real business value with lower costs and higher efficiency. Stable financial performance also provides solid support for ongoing technological R&D and ecosystem development. By the end of 2025, the company's cash and cash equivalents reached RMB 15.05 billion, a nearly fourfold increase from the beginning of the year. The asset-liability ratio decreased from 59.1% to 39.8%, indicating a significantly optimized financial structure. Total interest-bearing bank and other borrowings amounted to RMB 10.09 billion, with overall debt levels under control. **Forward-looking Deployment, Edge Computing Opens a New Growth Frontier** As scenarios such as embodied intelligence, autonomous driving, and industrial internet mature rapidly, demand for intelligent computing at the edge is set to explode. The company's Tongyang series of edge computing products, targeting fields like robotics and smart terminals, brings the multi-computing fusion and high-concurrency inference capabilities from the cloud down to the edge. They deeply integrate CPU, GPU computing power, and various high-speed interfaces, achieving international mainstream levels in core metrics such as computing power, latency, and bandwidth, further expanding the intelligent boundaries of the physical world. While deepening its edge-side deployment, and with the large-scale deployment phase of large model applications underway, the company is expected to further deepen its penetration in key industries such as internet, finance, healthcare, and transportation, further consolidating its leading position in the domestic general-purpose GPU market. ### Related Stocks - [ILUVATAR COREX (09903.HK)](https://longbridge.com/en/quote/09903.HK.md)