Cloud Leopard Intelligent Co., Ltd. (hereafter "Cloud Leopard") has recently responded to the exchange's inquiry letter as it continues its push for a ChiNext IPO, aiming to raise 3.035 billion yuan.

As a potential candidate for the "first domestically produced DPU stock," Cloud Leopard's listing has drawn significant attention. However, several issues are worth noting: cumulative losses exceeding 2.4 billion yuan over three years, 95% of revenue dependent on a single related party, Tencent, a cumulative operating cash flow deficit of over 1.8 billion yuan, and drastic fluctuations in gross margin.

More critically, Tencent and its affiliates collectively hold 19.7792% of Cloud Leopard's shares. The directors nominated by Tencent possess veto power, significantly impacting the independence of the company's governance. Furthermore, 95% of Cloud Leopard's revenue is ultimately sold to Tencent, triggering two major "red flags": heavy reliance on a single customer and a lack of operational independence (related-party transactions exceeding 90%). Whether it can successfully navigate this remains to be seen.

Tencent Holds Nearly 20% Stake and Enjoys Veto Power

According to the prospectus, Cloud Leopard designs and develops DPU products. It relies on foundries and packaging/testing houses for chip production and board-level manufacturers for integration, ultimately delivering DPU boards to customers. A DPU evolves from traditional network interface cards (NICs), with mainstream technology moving from simple, non-programmable ASIC chips to complex, multi-core, heterogeneous SoC chips, with interim solutions based on FPGA chip designs.

From 2023 to 2025, Cloud Leopard reported revenues of 173,200 yuan, 36 million yuan, and 370 million yuan, respectively. Over the same period, net losses attributable to the parent company were 667 million yuan, 610 million yuan, and 1.189 billion yuan, totaling a cumulative loss of over 2.4 billion yuan. As of the end of 2025, the company's accumulated undistributed profits were -1.169 billion yuan.

Cloud Leopard attributed the losses to several factors: the low penetration rate of domestically produced DPUs with gradual market expansion, ongoing product introduction during the reporting period, persistently high R&D expenses, and the implementation of extensive equity incentives that resulted in significant share-based payment expenses.

The company's net cash flow from operating activities also remained negative, standing at -505 million yuan, -598 million yuan, and -706 million yuan from 2023 to 2025, respectively, with a cumulative outflow of over 1.8 billion yuan. Despite continuous massive losses and negative operating cash flow, Cloud Leopard plans to raise 3.035 billion yuan for DPU R&D and industrialization projects.

Filing documents show that actual controllers Xiao Qiyang and Zhang Xueli collectively control 29.8108% of the company's voting rights, less than 30%. Meanwhile, Tencent, through Shenzhen Tencent and its affiliates Guangxi Tencent and Suzhou Paiyi, holds a total of 19.7792% of the shares. The difference in voting rights between the actual controllers and the Tencent group is only 10.0316%.

Although there is a 10-percentage-point difference in voting rights, because the actual controllers' shareholding is below 30% and the director nominated by Shenzhen Tencent holds a board veto, the influence of the Tencent group on the company's governance independence is significant. Cloud Leopard stated that Shenzhen Tencent cannot unilaterally control the issuer based on this. During the reporting period, Shenzhen Tencent never actually exercised its veto power at the board or shareholder meetings and did not intervene in corporate governance or daily operations through this mechanism.

Dual Red Flags: Independence and Single-Customer Reliance

Tencent's influence on Cloud Leopard extends beyond governance to the core of its operational independence and whether its heavy customer concentration crosses the review "red line." From 2023 to 2025, sales to Cloud Leopard's top five customers accounted for 100%, 99.35%, and 92.87% of annual revenue, respectively. By end customer, direct and indirect sales to the largest end customer, Tencent (and its related parties), accounted for 0%, 93.16%, and 95.22% of annual revenue over the reporting period.

Cloud Leopard explained that the high customer concentration is related to the company's development stage and the concentration characteristics of the downstream market. According to the "Regulatory Rules Applicable Guidelines – Issuance Category No. 5," Article 5-17 on "Customer Concentration": "If an issuer's revenue or gross profit contribution from a single customer exceeds 50%, it is generally considered that the issuer has a significant dependence on that customer." Cloud Leopard's 93.16% in 2024 and 95.22% in 2025 far exceed the 50% threshold for "significant dependence."

Under these rules, review requirements for customer concentration include "whether the issuer has a related-party relationship with the major customer, whether the method of obtaining business affects independence, and whether the issuer has the ability to independently obtain business from the market." Furthermore, if the sponsor cannot fully verify and explain the reasonableness of the single-customer dependence, customer stability, or business continuity, it must prudently express a review opinion on whether the issuer has sustainable operations.

Interestingly, Tencent and Cloud Leopard have a related-party relationship, and the Tencent group's veto power gives it considerable influence. This means that Cloud Leopard's related-party sales in 2024 and 2025 exceeded 90%, posing a major challenge to its operational independence. Cloud Leopard argued that it primarily engages in DPU R&D and sales, while Tencent is a leading domestic cloud service provider. It is reasonable and necessary for Tencent and its related parties to purchase DPU products and deploy them on a large scale in their data center servers.

Filing documents show that, based on 2025 revenue, Cloud Leopard ranks first among independent domestic manufacturers in China's full-function DPU market. However, the value of this "first-place" ranking is questionable given that 95% of its revenue depends on Tencent.

In summary, Cloud Leopard's independence faces three major questions: First, product independence or commercial independence. Some investors question whether the company's product definition, technical roadmap, and performance standards excessively cater to Tencent's needs. In its inquiry response, Cloud Leopard acknowledged a "deep alignment with Tencent's multi-scenario, multi-functional product requirements, performance standards, and technical specifications under the 800G network bandwidth cloud network and new-generation AI computing architecture." Whether this "deep alignment" is normal customer orientation or de facto "customized development" requires an answer from the company.

Second, pricing independence. The gross margin of the company's DPU boards plummeted from 46.29% to 24.38%. The company explained that this was because "Tencent, as a strategic seed customer, was given a slightly lower price considering its purchase volume and the company's own market expansion strategy." However, is this "strategic concession" fair? Can other customers enjoy the same price? If Tencent uses its position as a major shareholder to depress purchase prices, how are the rights of minority shareholders protected? These issues are equally concerning.

Third, governance independence. As mentioned, the director nominated by Tencent holds veto power. Although it has never been exercised during the reporting period, this "sword of Damocles" itself tests the company's independence. Tencent is simultaneously a nearly 20% shareholder, a major customer, and enjoys special governance rights.

Sharp Gross Margin Fluctuations

From 2023 to 2025, Cloud Leopard's comprehensive gross margins were 11.34%, 39.81%, and 27.62%. After a brief spike to nearly 40% in 2024, the gross margin fell back by over 12 percentage points in 2025. Cloud Leopard stated that the significant fluctuation in gross margin was mainly due to large differences in the company's product revenue mix during each reporting period. In 2023, only a small amount of other revenue was realized. DPU products began mass production and sales in 2024, with revenue primarily from DPU chips in 2024 and DPU boards in 2025.

More concerning is the performance of its core product, the DPU board. From 2024 to 2025, its gross margin fell from 46.29% to 24.38%. Cloud Leopard explained that the large change in DPU board gross margin was because, in 2024, the boards sold in the initial mass production phase were mainly small-volume samples to customers like China Mobile and Southern Power Grid, resulting in a higher per-card price and overall gross margin. In 2025, the business cooperation with its main customer, Tencent, shifted to the industry-standard model of board delivery. During this period, Cloud Leopard sold 43,100 boards to Tencent. As a strategic seed customer, Tencent received a slightly lower price due to its large purchase volume, bringing down the average card price. Additionally, the global memory component price cycle in 2025 increased the company's raw material costs. The company concluded that the decline in the 2025 DPU board gross margin compared to 2024 was reasonable.

Compared to its peers, Cloud Leopard's gross margin is significantly lower. From 2023 to 2025, the average gross margin of comparable companies was 48.11%, 55.06%, and 53.39%. Except for Suiyuan Technology, which had a similar gross margin, the other four peer companies far exceeded Cloud Leopard. Cloud Leopard noted that the comparable companies listed in the prospectus are all domestic computing chip companies that are either listed or planning to list, with different product types and target markets, making direct gross margin comparisons inappropriate.