The Chinese capital markets have witnessed a rare industrial narrative this year, with a company making significant moves in two critical sectors within a single week. This series of strategic actions aptly outlines the commercial landscape being reshaped by Xunce (03317).

On June 8th, Xunce announced the deployment of its TokenOS operating system into intelligent connected vehicles, the largest testing ground for physical AI. On June 15th, the company again positioned itself at the industry's forefront by announcing strategic cooperation agreements with three major domestic GPU manufacturers: MetaX Integrated Circuits, Tianshu Zhixin Semiconductor, and Biren Technology.

Signing deals with three domestic GPU players simultaneously is an uncommon event. This series of agreements is anchored in specific vertical scenarios, follows a path of integrated hardware and software, and aims for out-of-the-box usability, with each collaboration targeting a clear industry sector and a deliverable product form.

These moves signify a deep coupling of Xunce's "computing power + data" strategy and explore a replicable practical path for the domestic AI industry to overcome implementation bottlenecks and achieve scaled industrial application. Viewed within the larger narrative of Xunce's continuous strategic moves, this likely represents a critical step in its comprehensive leap from a "data service provider" to an "integrator of domestic AI infrastructure."

The "Last Mile" Challenge in the Domestic GPU Race

With the widespread application of AI and big data technologies, China's computing power scale is experiencing rapid growth, and the domestic GPU market has seen explosive expansion over the past five years. According to Frost & Sullivan forecasts, the global GPU market is expected to reach 3.61 trillion yuan by 2029. Within that, China's GPU market is projected to reach 1.36 trillion yuan by 2029, with its global market share expected to rise from 15.6% in 2024 to 37.8% in 2029.

However, behind these prosperous figures lies a common, deep-seated industry dilemma. While the issue of domestic GPUs being "usable" is largely resolved, a significant gap remains in making them "user-friendly," with three core pain points.

First, fragmented software ecosystems. The domestic GPU field is flourishing with many players, but software ecosystem compatibility and industry adaptation are uneven, migration costs remain high, and customers lack both the willingness and capability to switch.

Second, a long-standing disconnect between computing power and data. Computing power providers lack data support from vertical scenarios, while industry clients lack the integrated capability for efficient computing power and data governance.

Third, prohibitively high "last-mile" costs. Even after procuring domestic GPUs, enterprises still need to undertake extensive engineering work like model adaptation and business debugging themselves, leading to long AI implementation cycles and uncertain outcomes.

A deeper change is that GPUs are shifting from being generic computing suppliers to scenario-customization experts, with miniaturization and scenario-specific reverse design becoming new trends. This reverse innovation, which deduces chip architecture from business challenges, requires vast amounts of scenario data as fuel—precisely where Xunce's core capability lies.

Xunce has deep experience in over a dozen vertical fields including finance, telecommunications, power, advanced manufacturing, smart cities, and embodied intelligence, accumulating substantial scenario data governance capabilities and industry implementation know-how.

MetaX, Tianshu Zhixin, and Biren, as core players in the domestic GPU race, possess respective core technical advantages in software ecosystem compatibility, general computing performance, and large-scale computing cluster deployment. The precise complementarity of their capabilities with Xunce's directs the cooperation from the outset towards genuine industrial implementation needs, rather than conceptual publicity stunts.

A Systematic Approach with Three Parallel Chips

It is noteworthy that Xunce is not betting on a single chip but has simultaneously partnered with three companies to build diversified computing power channels. In the wave of domestic substitution, different GPU manufacturers are forming differentiated competition, and betting on a single one implies a single point of ecosystem risk. Xunce, in its role as a data operating system, is designed to be compatible with all upstream computing infrastructure.

Xunce is matching the technical characteristics and advantage areas of the three GPU manufacturers with corresponding vertical sectors, forming a targeted implementation layout with clear scenarios, ultimately aiming to deliver integrated hardware/software solutions that are "ready to use out of the box."

The cooperation with MetaX anchors in two high-value sectors: smart manufacturing and tech finance. The partners will explore full-stack intelligent computing end-to-end technical paths, jointly build an efficient and stable computing power foundation, conduct deep validation in areas like vertical application agent development, structured data model fine-tuning, and mainstream model adaptation/debugging to promote product scaling and achieve "out-of-the-box" deployment goals for clients.

The collaboration with Tianshu Zhixin will involve joint chip-level adaptation and optimization for vertical scenarios to create efficient, integrated hardware/software computing solutions. They will conduct joint chip-level testing and deep optimization for typical industry scenarios, forming integrated solutions tailored to specific business needs. They will also jointly research heterogeneous computing network solutions for complex business scenarios, aiming to achieve unified management and dynamic allocation of diverse heterogeneous computing resources by building a "cloud-edge-end" collaborative computing power scheduling system, accelerating the scaled application of intelligent computing in areas like financial risk control, industrial quality inspection, and smart energy.

Furthermore, the two parties will jointly conduct R&D for the application of embodied intelligence (physical AI) in scenarios like smart manufacturing, intelligent services, and smart inspection, jointly exploring commercialization paths for physical AI.

The partnership with Biren Technology focuses deeply on the core urban management sector, aiming for scaled implementation through computing clusters and industry-specific integrated appliances. The two will collaboratively develop end-to-end intelligent computing solutions for urban management, jointly build domestic intelligent computing clusters, and integrate GPU computing power with data computing platform capabilities.

They will jointly create dedicated integrated appliances for urban management, standardizing hardware/software adaptation capabilities to significantly lower client deployment barriers, while deepening cooperation in areas like agent development and model debugging/validation to jointly cultivate the industrial ecosystem.

It is evident that Xunce's cooperation with the three GPU manufacturers establishes a deep collaborative model of "computing power provider + data integrator." From scenario-specific customization and optimization at the chip level, to deep adaptation at the system level, to integrated appliance consolidation at the product level, and finally to joint promotion at the market level, the parties have formed a closed-loop, full-chain collaboration.

This model breaks down the barriers between computing power and data, enabling chip capabilities to directly match industry scenario needs, while also allowing industry data to drive computing power optimization in reverse. This truly achieves a two-way fusion of "computing power moving downward, data moving upward," exploring a new paradigm for collaborative implementation within the domestic AI industry.

Investment Value: A Triple Logic for Valuation Reassessment

Assessing the commercial value of this strategic move, a triple logic supports Xunce's long-term investment value.

First, a quantitative reassessment of market scale. Artificial intelligence is entering a new era driven by computing power, with the global computing power scale continuing to expand at high speed. A Deloitte China report, "Technology Trends 2025," indicates the global chip market is expected to reach $576 billion in 2024, with AI chips accounting for 11%, surpassing $57 billion. By 2025, the next-generation AI chip market is projected to exceed $150 billion; by 2027, the global AI chip market is expected to grow to $400 billion, with a conservative estimate also reaching $110 billion.

Frost & Sullivan also forecasts the global GPU market will reach 3.6 trillion yuan by 2029, with China's GPU market expected to be 1.36 trillion yuan, its share rising to 37.8%, highlighting the dual opportunities of high-speed growth and domestic substitution.

Second, continuous optimization of revenue structure. Xunce is transitioning from a traditional project-based model to a fee model based on Token consumption. In April 2026, the quarterly recurring revenue (ARR) for Token grew 300% sequentially, with the proportion of paid Token revenue already exceeding 5% and expected to rise to 20%-30% by year-end. This triple signing will further unlock incremental Token usage in the AI field while smoothing the seasonal volatility of project-based income, structurally enhancing revenue quality and predictability.

Third, a deep reconstruction of valuation logic. As AI enters the physical world, the enterprise that can define the metrics and profit-sharing models for physical interactions occupies a structural position in the industry chain. Xunce's scenario Token system, built on the TokenOS operating system, is precisely an early mover attempting to become this player in the era of physical AI.

As the large-scale deployment of the GPU system deepens, the potential for long-term revenue scale to continuously expand exists, and the market's valuation logic for Xunce will be reconstructed.

If the smart vehicle sector represents the first ticket Xunce secured for the scaled implementation of physical AI, then the domestic GPU sector represents a systemic strategic positioning locked in at the computing power supply chain level.

The combination of the two forms a clear closed loop: the TokenOS operating system provides domestic GPUs with data Tokenization capability, enabling chips to leap from "usable" to "user-friendly"; large-scale implementation scenarios provide the TokenOS operating system with high-frequency, high-density Token usage as fuel.

With the TokenOS operating system at its core and the three major domestic GPUs as its foundation, Xunce is constructing a complete ecosystem for the synergy of data, computing power, models, and applications.

When computing power and data are no longer disconnected, and when the TokenOS operating system can run seamlessly on a diverse chip foundation, Xunce's positioning as a "key integrator of domestic AI infrastructure" will be continuously validated and strengthened in future market expansion, with its industrial and commercial value gradually being realized.