Author｜Qian Jing

Editor｜Chen Xiaoran

Five-Year Locked-In Order: Performance Progress Yet to Be Determined

On the evening of October 19, an announcement by Xiechuang Data (300857.SH) garnered attention as a computing power contract valued at RMB 37.49 billion officially came into effect. The company has secured its largest business order to date, opening up growth prospects for the next five years while simultaneously putting its funding, delivery, and payment collection capabilities under scrutiny.

A Colossal Order

1 According to the announcement, Xiechuang Data's wholly-owned subsidiary, Xiechuang Intelligent Computing, signed a Computing Power Cloud Service Agreement with a client. The estimated total amount of the contract does not exceed RMB 37.49 billion, with a service period of 60 months starting from the day after the order passes acceptance and the client's acceptance report is obtained. The agreement is now in effect, and the company will bill on a monthly basis after the service passes final acceptance.

Image source: Xiechuang Data announcement

A straightforward calculation reveals that the contract's upper limit translates to an average annual revenue of approximately RMB 7.498 billion and an average monthly revenue of about RMB 625 million. This figure represents 61.28% of Xiechuang Data's total annual revenue of RMB 12.236 billion in 2025 and is roughly 2.7 times the revenue of RMB 2.761 billion generated from intelligent computing power products and services in the same year. Within the company's scope, this is undoubtedly a massive order.

However, the phrase 'not exceeding' is pivotal. The RMB 37.49 billion figure represents the upper limit calculated based on the service term, computing power quantity, and monthly price. The final revenue hinges on purchase orders, actual delivery quantities, acceptance results, and settlement progress. It does not constitute confirmed performance on the day of signing.

The contract also imposes a stringent threshold: fees can only be billed after successful delivery, inspection, and acceptance. If performance, deployment quality, operational stability, or compliance fail to meet standards and remain unsatisfactory after rectification, the client may terminate the order or agreement without being liable for breach of contract.

The client's identity is exempt from disclosure due to trade secrets, so the outside world cannot independently assess client concentration, genuine demand, or payment capacity at present.

Payment methods are not confined to cash; bank or commercial acceptance bills are also acceptable. If bills are discounted in advance, who bears the cost must be separately agreed upon. In other words, the value of this contract cannot be judged solely by the total amount but must also take into account subsequent purchase orders, server arrivals, acceptance reports, and payment collection quality.

A large order is not a money-printing machine but rather akin to a five-year construction blueprint. The blueprint is ready; whether the building can be completed on time is the next major focus.

Assets and Transformation

Xiechuang Data was not originally a pure computing power company. Initially, the company focused on data storage devices and AIoT smart terminals, offering products such as SSDs, memory sticks, NAS, smart cameras, and smart doorbells. It later expanded into server and peripheral remanufacturing, then advanced into GPU clusters, cloud computing power, and industry applications.

Today, Xiechuang Data operates in four business segments: intelligent computing power products and services, data storage devices, server and peripheral remanufacturing, and AIoT smart terminals. Among these, the computing power segment not only sells servers and cluster solutions but also operates GPU clusters, providing resource scheduling, model deployment, operation and maintenance, and performance optimization services through the FCloud platform. Application scenarios encompass model training and inference, graphics rendering, autonomous driving, biopharmaceuticals, and cross-border e-commerce.

This transformation is already evident in its financial statements. In 2025, Xiechuang Data achieved revenue of RMB 12.236 billion and net profit attributable to shareholders of RMB 1.164 billion, representing year-over-year growth of 65.13% and 68.32%, respectively.

Revenue from intelligent computing power products and services reached RMB 2.761 billion, up 1,727.17% year-over-year, accounting for 22.57% of total revenue, with a gross margin of 25.53%. Revenue from data storage devices was approximately RMB 4.493 billion, accounting for 36.72% and remaining the largest revenue source.

By the first half of 2026, the transformation accelerated further, with revenue reaching RMB 12.523 billion, up 153.29% year-over-year, already surpassing the total for 2025. Net profit attributable to shareholders was RMB 1.838 billion, up 325.51% year-over-year. Revenue from intelligent computing power products and services was RMB 3.216 billion, also exceeding the total for the previous year.

Net cash from operating activities turned positive at RMB 1.825 billion, compared to -RMB 1.560 billion in the same period last year, indicating that some projects have begun entering the delivery and payment collection stages.

The challenges are also apparent in the same financial statements. As of the end of June 2026, Xiechuang Data's fixed assets reached RMB 22.024 billion, compared to RMB 9.283 billion at the end of 2025. Short-term and long-term borrowings totaled approximately RMB 17.362 billion, with the asset-liability ratio rising to 86.79%.

In the first half of 2026, Xiechuang Data's financial expenses were RMB 586 million, up 714.48% year-over-year. Net cash outflows from investing activities were RMB 12.443 billion, while net cash inflows from financing activities were RMB 13.853 billion.

This set of data is clear-cut: the company's profits are growing, but its reliance on financing is also on the rise. The value of the RMB 37.49 billion long-term order lies in linking heavy asset investments with five years of service demand. The risk is that servers must be purchased upfront, interest must be paid in advance, and revenue can only be recognized monthly after acceptance.

If delivery is delayed, financing costs increase, or client payment collection is prolonged, significant mismatches may emerge in the income statement and cash flow.

Dividends and Dangers

The industry winds are still favorable.

According to the 2026 China AI Computing Power Development Assessment Report jointly released by IDC and Inspur Information, China's intelligent computing power scale is expected to reach 2,576.5 EFLOPS in 2026, up 87.9% year-over-year, and may increase to 10,524.3 EFLOPS by 2030, with a compound annual growth rate of 50.3% from 2025 to 2030.

The driving force is shifting from large-scale model training to intelligent agents, multimodal applications, and high-frequency inference.

This shift is good news for computing power service providers. Training projects tend to be phased, while inference demand is more sustained, more dispersed, and requires greater elastic scheduling. Enterprises that purchase GPUs and build data centers themselves not only face high investments but also bear equipment iteration, utilization fluctuations, and operation and maintenance costs.

On-demand leasing is therefore more akin to utilities: clients pay for usable computing power, not a room full of machines.

Xiechuang Data's opportunity lies in not merely subleasing server racks. The company already possesses storage, server remanufacturing, cluster delivery, cloud platform, and operation and maintenance capabilities, and it has deployed computing power centers in multiple locations.

If it can truly integrate 'computing, storage, connectivity, and management,' long-term contracts will not only generate revenue but also dilute procurement and operation and maintenance costs, improve equipment utilization, and accumulate client entry points for subsequent model services and industry applications.

However, the industry will not always rely on the notion that 'having cards means making money.' While high-end computing power remains in short supply, mid-to-low-end resources may face structural oversupply. Chip iterations are rapid, and rental income and residual value of old equipment will decline. Power, networking, data center utilization rates, and financing costs will all erode gross margins.

The true competitive barriers are shifting from equipment scale to client quality, cluster stability, scheduling efficiency, and capital costs.

Therefore, while this contract warrants attention, it is not yet time to celebrate based solely on the total amount. The focus should not be on stock price speculation but on four key indicators: the scale and delivery time of the first batch of orders, client acceptance progress, monthly payment collection methods, and the company's new borrowings and financial expenses.

If these four aspects proceed smoothly, Xiechuang Data may complete a crucial leap from a hardware manufacturer to a comprehensive computing power service provider. If funding and delivery cannot keep pace, the RMB 37.49 billion contract may transform from a growth driver to a stress test on the balance sheet.

True large orders are never merely signed; they are earned through the delivery of server after server, acceptance report after acceptance report, and cash collection after cash collection.