Domestic GPUs welcome a key puzzle piece! XUNCE (03317) "three-core advancement" opens up the "computing power + data" closed loop, leveraging the trillion-dollar AI landing market.

In just a week's time, XUNCE Technology, a key player in the industry, made consecutive moves in two important race tracks, a rare occurrence in the capital market this year. However, this series of moves aptly outlined the commercial landscape that XUNCE Technology (03317) is reshaping. On June 8th, XUNCE Technology announced the landing of the TokenOS operating system in the smart connected car, the largest physical AI test field; on June 15th, the company once again took the lead in the industry by signing strategic cooperation agreements with three domestic GPU manufacturers - Muxi Integrated Circuit, ILUVATAR COREX Semiconductor, and BIREN TECH. It is rare for a company to collaborate with three major domestic GPU manufacturers at once. These series of agreements are anchored on vertical scenarios, combining software and hardware, and aiming for plug-and-play solutions. Each collaboration corresponds to specific industry tracks and deliverable product forms. Behind these moves, it signifies XUNCE's deep coupling of "computing power + data," and provides a replicable practical path for domestic AI industries to address the pain points of implementation and explore scalable industrial applications. If we look at these collaborations within the broader context of XUNCE's continuous strategy, it is likely a key deployment in XUNCE's full leap from a "data service provider" to a "domestic AI infrastructure integrator." The dilemma of the "last mile" in the domestic GPU track With the widespread application of AI and big data technology, China's computing power is rapidly increasing, and the GPU market has experienced explosive growth in the past five years. According to Frost & Sullivan's forecast, the global GPU market is expected to reach 3.61 trillion yuan by 2029. By 2029, the market size of China's GPU market is expected to reach 1.36 trillion yuan, with a market share increasing from 15.6% in 2024 to 37.8% in 2029 among global markets. However, behind the prosperity of data lies the industry's deep-seated problems. While the issue of "usable" domestic GPUs has been mostly resolved, the gap in "good usability" still exists, with three core pain points: first, fragmented software ecosystems. While there is a variety of domestic GPUs available, the software ecosystem lacks consistency and compatibility with the industry, resulting in high migration costs and low willingness and capacity from customers; second, the long-term separation of computing power and data. Computing power manufacturers lack vertical scenario data support, while industry customers lack the integration capabilities for efficient computing power and data governance. Third, the high cost of the "last mile." Even if companies purchase domestic GPUs, they still need to complete a considerable amount of engineering work, such as model adaptation and business debugging, resulting in a long AI deployment period and uncertainty in effectiveness. Moreover, a more profound change is that GPUs are transitioning from general computing power suppliers to scene customization experts, with miniaturized and scenario-based reverse design becoming a new trend. This reverse innovation, which traces chip architecture back from business challenges, requires a massive amount of scenario data - which happens to be the core capability of XUNCE. Having developed deep expertise in more than ten vertical fields such as finance, telecommunications, electricity, high-end manufacturing, smart cities, and embodied intelligence, XUNCE has accumulated profound capabilities in scenario data governance and industry implementation. Muxi, ILUVATAR COREX, and BIREN TECH, as core players in the domestic GPU race, each have core technological advantages in software ecosystem compatibility, general computing power performance, and large-scale computing cluster deployment. The precise complementarity of their capabilities allows cooperation to address real industry needs from the very beginning, rather than being solely conceptual. A Systematic Layout of "Three Cores Moving Forward" It is worth noting that XUNCE is not betting on a single chip but has simultaneously signed agreements with three, establishing diverse computing channels. In the wave of domestic substitution, different GPU manufacturers are forming differentiated competition, with a single focus implying ecological singular risk. As a data operating system, XUNCE is compatible with all computing infrastructure. XUNCE has matched the unique technical characteristics and strengths of the three GPU manufacturers with corresponding vertical tracks, forming distinct grounding layouts with clear focuses, ultimately leading to the delivery of an integrated software and hardware solution: With Muxi, it anchors intelligent manufacturing and technology finance as two major high-value tracks. Both parties are exploring a full-stack intelligent computing end-to-end technology path, building an efficient and stable computing base to deeply validate vertical application intelligent body development, structured data model fine-tuning, mainstream model adaptation and debugging, and promote product scaling implementation, achieving the deployment goal of "plug-and-play" for clients. With ILUVATAR COREX, they will jointly carry out vertical scene chip-level adaptation and optimization, creating an efficient computing solution. They will conduct joint testing and deep optimization for typical industry scenarios at chip level, forming an integrated software and hardware solution tailored to specific business needs. They will also research heterogeneous computing network solutions for complex business scenarios, accelerating the large-scale application of intelligent computing in finance risk control, industrial quality inspection, smart energy, etc. Additionally, they will jointly develop applications of embodied intelligence (physical AI) in intelligent manufacturing, intelligent services, smart inspection, etc., exploring the commercial landing path of physical AI together. With BIREN TECH, they will deeply cultivate the core track of city management, achieving scaled implementation through computing clusters and integrated industry machines. Both parties will collaborate on developing end-to-end intelligent solutions in the city management field, building a domestic intelligent computing cluster, integrating GPU computing power and data computing platform capabilities, standardizing the adaptability of software and hardware, greatly reducing customer deployment thresholds. They will also deepen cooperation in intelligent body development, model debugging and verification, fostering the industry ecosystem together. It is clear that XUNCE's collaboration with the three GPU manufacturers has established a deep collaborative model of "computing manufacturers + data integrators": from scene-customized optimization at the chip level, to deep compatibility at the system level, and eventually to integrated solutions at the product level, jointly promoted in the market. They have formed a closed-loop collaboration across the entire chain. This model breaks down the barriers between computing power and data, allowing chip capabilities to directly match industry scene demands and industry data to drive computing power optimization in reverse, truly achieving the dual fusion of "computing power sinking, data rising," and exploring a new paradigm for collaborative domestic AI industrial landings. Investment value: Three-fold logic to reconstruct valuation From the evaluation of commercial value, there are three-fold logics supporting XUNCE's long-term investment value: First, a reevaluation of the market space's magnitude. Artificial intelligence is entering a new era driven by computing power, with global computing power expanding rapidly. Deloitte China's "Technology Trends 2025" report shows that the global chip market is expected to reach $576 billion by 2024, with AI chips accounting for 11%, surpassing $570 billion. By 2025, the market for new generation AI chips is expected to exceed $150 billion; by 2027, the global AI chip market is expected to grow to $400 billion, reaching a conservative estimate of $110 billion. Frost & Sullivan also predicts that the global GPU market will reach 3.6 trillion yuan by 2029, with China's GPU market size expected to reach 1.36 trillion yuan by 2029, representing 37.8% of the market, highlighting the high-speed growth and dual opportunities of domestic substitution. Second, continuous optimization of the revenue structure. XUNCE is transitioning from traditional project-based billing to a billing model based on Token consumption. In April 2026, Token's recurring revenue (ARR) grew by 300% quarterly, and Token's paid revenue share has exceeded 5%, expected to rise to 20%-30% by the end of the year. This triple signing will further open up incremental Token calls in the AI field, while smoothing out seasonal fluctuations in project-based income, structurally enhancing income quality and predictability. Third, a deep reconstruction of valuation logic. When AI enters the physical world and enterprises define measurement methods and profit distribution models for physical interactions, they occupy a structural position in the industry chain. XUNCE's scene Token system built on the TokenOS operating system is attempting to be a pioneer in this role in the era of physical AI. With the continuous deep deployment of the GPU system, there is a potential for sustained enlargement of long-term income scale, and the market's valuation logic for XUNCE will be reconstructed. If the smart car track has given XUNCE the first ticket for the scaled implementation of physical AI, then in the domestic GPU track, XUNCE has strategically positioned itself within the computing supply chain. The combination of the two forms a clear loop: TokenOS operating system provides data tokenization capability for domestic GPUs, enabling chips to transition from "usable" to "usable." Large-scale scene deployments provide high-frequency, high-density Token call fuel for the TokenOS operating system. With TokenOS operating system at its core and the three major domestic GPUs as its base, XUNCE has built a complete data-computing-model-application collaborative ecosystem. When computing power and data are no longer separate, when the TokenOS operating system can seamlessly run on a range of chip bases, XUNCE's positioning as a "critical integrator of domestic AI infrastructure" will continue to be verified and strengthened in the future market expansion, gradually releasing its industrial and commercial value.