The artificial intelligence computing landscape is transitioning from cloud-based training to edge-level inference and large-scale deployment, with memory chips riding a wave of sustained demand driven by the exponential growth in large language model parameters. Embodied intelligence is simultaneously accelerating the need for on-device processing power. Against this backdrop of converging growth drivers, INGDAN (HKG: 00400) delivered a robust first-half financial report on August 31st, marked by simultaneous growth in both volume and profitability.

According to the financial results, the company generated revenue of RMB 13.249 billion in the first half, a substantial 98.4% increase year-over-year. Fueled by strong demand for AI computing power and storage chips, operating profit reached RMB 517 million, up 87.7% annually, while net profit surged 96.3% to RMB 373 million. All three key metrics recorded robust growth, showcasing accelerating profit elasticity and tangible progress in its AI transformation strategy.

Meanwhile, the AI Token Factory is approaching a pivotal revenue realization phase. Previously disclosed letters of intent for service orders exceeding USD 1 billion have been converted into formal service contracts. These agreements are slated to contribute over USD 1 billion in service fees cumulatively over the next five years, with deliveries commencing in the fourth quarter of 2026. Market valuation logic for the company is shifting towards its new identity as a "Token computing power operator."

Second Growth Curve Taking Shape

With three decades of deep immersion in the chip industry, INGDAN serves tens of thousands of clients both domestically and internationally, spanning over a hundred industry scenarios across AI computing centers and intelligent terminals. Its robust partnerships with leading global chip manufacturers such as NVIDIA, Intel, AMD, and SanDisk form a solid foundation that not only anchors current performance but also provides a ready-made customer base and supply chain capabilities for new business ventures.

Leveraging this foundation, the company is converting its chip resources into deliverable computing services. Responding to the demand for intelligent computing center construction, it has developed a comprehensive product matrix covering core chips and software ecosystems through its close ties with top-tier chip suppliers. This strategy transforms upstream chip access into full-stack computing services, from chip selection to cluster deployment and maintenance, shifting revenue quality from sheer shipment volume towards service depth.

In the realm of embodied intelligence, the company is strategically anchored to edge computing platforms like NVIDIA's Jetson series. It has completed its product roadmap—from development kits to mass-production solutions—for intelligent terminals including robotics and drones. Under its Kepler Lab initiative, self-developed SOM-level products are already shipping in volume to clients in sectors like customs and banking. The company offers complete solutions catering to the "big brain and small brain" architecture of robots, covering everything from cloud-based large models to real-time on-device inference, having secured marquee clients such as Meituan, Baidu, and WeRide in autonomous driving and robotics. As downstream clients accelerate their production schedules, this segment is poised to become a new source of medium-to-long-term growth.

While its full-stack service capabilities for AI computing are established and its positions in memory and embodied intelligence are secured, the real future lies with the AI Token Factory. The core innovation of this facility is its consumption-based billing model: each time a customer's AI application invokes computing power, INGDAN earns a revenue share. This ties the company's income directly to the activity levels of its clients' operations, transitioning revenue generation from one-off deliveries to recurring commissions.

This model aligns perfectly with the rising demand for computing in vertical industries. Humanoid robots and drones require continuous on-device inference, demanding a steady supply of edge computing. Since most vertical enterprises lack both the capability and the inclination to build their own computing clusters, third-party computing services become an essential solution. INGDAN is building a vertical-industry Token computing scheduling platform, employing a dual-track deployment strategy combining its own computing clusters with domestic and international third-party data centers. This asset-light approach to the computing operations sector sidesteps the financial strain of heavy capital expenditure. Its focus on delivering lightweight Token computing power to vertical industries such as humanoid robotics and drones reuses its existing downstream customer ecosystem, giving it a significantly lower marginal customer acquisition cost compared to competitors.

On August 24th, a five-year contract locking in over USD 1 billion in service fees was signed, with service delivery commencing in Q4 2026. The planned computing network spans multiple Asian countries, with a total capacity exceeding 100 megawatts, targeting tier-1 cloud service providers and the robotics vertical market. Remarkably, the journey from announcing the intention in mid-July to signing the formal contract took just over a month. For the computing services sector, this swift transition from intent to agreement within such a timeframe signifies substantial customer validation of both the business model and the company's delivery capabilities. More critically, the five-year revenue recognition cycle of this contract marks the beginning of a fundamental structural change in the company's income streams.

A Shift in the Valuation Anchor

Following the Token Factory's establishment, INGDAN's revenue model is undergoing a transformation, rendering the old framework of hardware distribution pricing obsolete for three key reasons. First, the efficiency of the old distribution model has reached its ceiling. Consider the industry benchmark: as of the end of 2025, inventory stood at approximately RMB 772 million with a turnover period of about 21 days, translating to roughly 17 turns per year. This is an industry-leading efficiency level, leaving little room for further optimization within the traditional framework. When the market continues to price the company based on distributor logic, the growth potential remains trapped within the model's boundaries, failing to account for the value of new business ventures—this is why a valuation logic shift is necessary.

Second, the nature of demand itself is being restructured. Computing power is evolving into a metered, subscribable, and renewable standardized service. Consumption patterns are shifting from centralized procurement to pay-per-use. Consequently, computing expenditure is transitioning from a predictable capital investment to a day-to-day operating cost that fluctuates with business activity. For the company, this changes the visibility and sustainability of its revenue streams, necessitating a corresponding shift in the metrics used to assess its value.

Third, the contract structure itself has changed. The five-year AI Token Factory agreement, distributing over USD 1 billion in service fees across its term, is significant not just for its monetary value, but for providing a replicable template for an entire computing operations model. As successive contracts are signed and the customer pool expands, the revenue flywheel will continue to accelerate. For context, leading overseas computing operators like CoreWeave boasted a contract backlog of USD 66.8 billion by the end of 2025, roughly 13 times their annual revenue for that year. Once the model proves successful, contract growth is not linear but compounds. The company's revenue model is being reshaped, a point already corroborated by its financials: operating profit of RMB 517 million and net profit of RMB 373 million, up 87.7% and 96.3% year-over-year, respectively.

Contract Finalization Triggers Value Reassessment

The finalization of the five-year service contract distributes over USD 1 billion in service fees, roughly USD 200 million annually, over the next five years. With service delivery slated to begin in Q4 2026, the revenue recognized within this fiscal year will be limited, with substantial contributions beginning to flow from 2027 onwards. For a computing operator, this represents a clear timeline for value realization—a schedule the market can price, as it brings predictable future revenue into sharper focus.

Comparing similar compute-related businesses, the valuation gap between distribution and operations is an order of magnitude. Traditional hardware distributors handle supply chain circulation and ancillary services, where chip flow heavily relies on distribution networks, and success is measured by scale and efficiency. Leading overseas distributors typically command low price-to-sales (P/S) ratios, reflecting their high-volume, low-margin electronic component distribution. In stark contrast, AI computing operators like CoreWeave offer income that is contractually locked, predictable, and sustainable, transforming revenue from "point-in-time" to "linear" and boasting far superior gross margin structures compared to distribution.

As a prime example, CoreWeave's contract backlog reached USD 66.8 billion by end-2025, about 13 times its annual revenue, and climbed to roughly USD 104 billion by June 2026. Currently trading at a P/S ratio exceeding 6 times—over 20 times higher than traditional distributors—the market is paying a clear premium for this "foreseeable future."

This logic applies equally to INGDAN. The company shares a similar business foundation with CoreWeave: upstream, it draws from the same original equipment manufacturer resources; downstream, it has tens of thousands of existing customers, many of whom are themselves significant consumers of computing power. Converting these hardware buyers into recurring computing service clients is the direct pathway for the Token Factory to capitalize on its established downstream ecosystem. INGDAN's starting point for this transition is higher than most peers. Its three decades of supply chain experience and massive existing customer base mean it doesn't need to build a computing ecosystem from scratch; it can secure manufacturer authorizations upstream and find computing power buyers downstream, with the intermediary scheduling platform and asset-light dual-track approach mitigating the financial pressures of heavy capital expansion.

The pace of transformation has also exceeded expectations—the leap from intention disclosure in mid-July to contract signing on August 24th, a span of just over a month, sealed a USD 1 billion five-year agreement. Its differentiated positioning towards vertical industries like humanoid robotics and drones allows it to avoid head-on competition with general-purpose cloud giants. With the first revenue schedule secured by the five-year contract, the next crucial step is consistently expanding the contract backlog thickness. Once the market begins to measure the company using the benchmarks for computing operators, the shift in its pricing reference point will be complete. INGDAN's revenue structure is already migrating towards a service-based model, and its order book is consolidating towards a contract-based framework. As this pricing reference transition occurs, revaluation is no longer a mere expectation but a process that will manifest quarter by quarter.

Looking ahead to the second half of the year, customer demand remains robust and the order backlog is substantial, with management projecting continued high-speed business growth. The AI Token Factory is scheduled to begin deliveries in Q4, with computing nodes coming online progressively, extending the revenue mix from one-off hardware deliveries towards recurring computing services. From delivery to recurring commissions, the viability of this business model has been validated. INGDAN's competitive moat will therefore no longer rest on any single contract, but on the strength of its entire customer network and computing scheduling architecture—that is the true strategic layout the company is executing.