Kunlun Tech Co., Ltd. (300418.SZ) has drawn market attention with its recent announcement to pursue a Hong Kong listing. The company stated its intention to issue overseas-listed foreign shares (H-shares) and apply for a main board listing on the Stock Exchange of Hong Kong, aiming to further its internationalization strategy, establish a global capital operation platform, enhance its international brand image, and boost overall competitiveness.
Simultaneously with the Hong Kong listing announcement, the company also disclosed a plan to divest a gaming asset. Kunlun Tech Co., Ltd. intends to transfer its entire equity stake in Beijing Xianlai Huyu Network Technology Co., Ltd. (Xianlai Huyu), held through its wholly-owned subsidiary Beijing Kunnuo Yingzhan Information Consulting Co., Ltd., to Beijing Xinglan Huyu Network Technology Co., Ltd. for a consideration of RMB 750 million.
Notably, while Kunlun Tech Co., Ltd. is currently loss-making, Xianlai Huyu remains a profitable entity for the company. Why divest this gaming asset at this juncture? What strategic considerations lie behind this transaction?
Divesting a Gaming 'Non-Performing Asset'?
Public records show Xianlai Huyu was established in 2016, focusing on the development and operation of local-feature casual chess and card mobile games, creating localized online entertainment platforms tailored to regional gameplay preferences. Kunlun Tech Co., Ltd. has experienced sustained losses in recent years, accumulating over RMB 3 billion in losses across 2024 and 2025. In 2025, the company's revenue reached RMB 8.198 billion (up 44.78% year-on-year), but its net loss attributable to shareholders was RMB 1.593 billion. This loss trend continued into the first quarter, with revenue of RMB 2.57 billion (up 45.69% year-on-year) but a net loss attributable to shareholders of RMB 887 million, widening from the RMB 769 million loss in the same period last year.
In contrast, Xianlai Huyu is a profitable asset. In 2025, it generated revenue of RMB 316 million, operating profit of RMB 208 million, and net profit of RMB 72.2 million. In Q1 2026, it posted revenue of RMB 58.12 million, operating profit of RMB 39.38 million, and net profit of RMB 34 million.
Why would Kunlun Tech Co., Ltd. divest a profitable subsidiary? Xianlai Huyu once thrived by targeting users in third-to-sixth-tier towns through a model combining local chess and card games, room cards, and multi-tiered agent distribution. At its peak, it generated revenue exceeding RMB 1.6 billion and profits over RMB 1.1 billion. However, the chess and card gaming sector has faced stringent regulatory crackdowns. In 2018, the Ministry of Culture and Tourism launched strict investigations into online games promoting gambling content, and in 2019, the Publication Bureau of the Central Propaganda Department explicitly stated that chess and card games would no longer receive publishing licenses. The suspension of game licenses meant Xianlai Huyu could not launch new products or drive growth through fresh titles, forcing it to rely solely on its existing user base.
Indeed, Xianlai Huyu's revenue and profits have contracted severely in recent years. Its 2025 profit was down over 90% from its peak. Revenue declined by over 16% in 2024, with the contraction accelerating to over 20% in 2025. As the business continues to shrink, the corresponding goodwill of over RMB 900 million on the books faces impairment risk. By the end of the 2025 reporting period, the company had accumulated RMB 568 million in impairment charges, leaving nearly RMB 400 million in goodwill still unimpaired.
Is the 4.8 Billion Goodwill Hiding a Landmine?
As of the end of Q1 2026, the company's total goodwill stood at RMB 4.763 billion. With the divestiture of the gaming asset, does the remaining goodwill pose an impairment risk? Within the goodwill breakdown, the acquisition of Opera Norway AS accounts for the largest portion at over RMB 3 billion, followed by over RMB 1 billion related to the Tiangong large model business. No impairment has been recognized for these assets so far. But could these assets be exposed to impairment risks?
Firstly, Opera's global monthly active users (MAUs) have been continuously declining. Should the over RMB 3 billion goodwill raise red flags? In its 2022 annual report, the company stated that it had built a platform-level product matrix with nearly 400 million global MAUs. Opera, a globally recognized internet brand and a top traffic gateway in Europe, the US, and Africa, boasted an average of 324 million global MAUs at that time. By 2024, this figure had dropped to nearly 300 million, and in 2025, it further declined to 284 million. Given this sustained contraction in MAUs, shouldn't the over RMB 3 billion goodwill warrant careful scrutiny? Opera's main operating entity is Opera Software Ireland Limited. It is worth noting that while the company's revenue has maintained a relatively high growth rate, its net profit growth has been slowing, from over 36% in 2024 to 24% in 2025. Additionally, Opera's operating cash flow growth has shown a persistent downward trend. Wind data indicates that its operating net cash flow growth was over 113% in 2022 but slowed to just 12.15% by 2025.
Secondly, Skywork AI has fallen into losses after completing its performance commitments. In February 2021, Kunlun Tech Co., Ltd. announced that its wholly-owned subsidiary Kunlun Group Limited acquired a 60.65% equity stake in Star Group Interactive Inc. (later renamed Skywork AI Inc.) for RMB 1.395 billion in cash. Post-transaction, the company held a total of 80% through Kunlun Group and its subsidiary Opera Limited. Star Group was consolidated into the financial statements in April 2021, resulting in goodwill of nearly RMB 100 million. The transferor, Happy Entertainment Limited, provided performance commitments stipulating that Skywork AI's audited net profit after deducting non-recurring items for 2021, 2022, and 2023 would be no less than RMB 191.67 million, RMB 230 million, and RMB 276 million respectively, totaling no less than RMB 698 million over three years. Ultimately, Skywork AI exceeded its commitments, posting cumulative audited non-recurring net profit of RMB 807 million against the RMB 698 million target, an overachievement of RMB 111 million.
However, the company's performance deteriorated sharply after the commitment period. Skywork AI recorded a loss exceeding RMB 300 million in 2024, which ballooned to RMB 1.265 billion in 2025. According to regulations on goodwill impairment indicators, if cash flows or operating profits continuously deteriorate or fall significantly below expectations at the time of goodwill formation—especially if the acquiree fails to meet committed performance—this may signal an impairment. Does this imply an impairment risk for the company's related assets? It's worth adding that Skywork AI Pte. Ltd.'s losses primarily stem from R&D investment and marketing expenses. When assessing goodwill impairment for these assets, the company has consistently found the recoverable amount to exceed the carrying value, hence no impairment has been recorded. Furthermore, under the business segment breakdown, the Starmaker community business has already seen revenue decline in 2025.
The AI Narrative: Buying Traffic to Boost Revenue?
Both the gaming asset divestiture and the annual report underscore Kunlun Tech Co., Ltd.'s AI-centric narrative. Regarding the divestiture, the company stated that artificial intelligence, particularly AGI and AIGC technologies, is at a critical stage of rapid development and industrial application. The company has identified AGI and AIGC as its core strategic direction, continuously increasing investment in R&D, product innovation, industry solutions, and ecosystem development. It noted that Xianlai Huyu, after years of development, has formed a relatively independent operational system and model, but differs from the company's future-focused AGI and AIGC business in strategic positioning, technical pathways, and industrial synergy. Post-transaction, Kunlun Tech Co., Ltd. will no longer hold equity in Xianlai Huyu, which will be deconsolidated from the company's financial statements.
In its 2025 annual report, the company highlighted steady progress across the full AI industry chain, with significant achievements in R&D and product innovation. It touted that its four major AI models have been continuously iterated, with the video generation model SkyReels-V4 and music generation model Mureka V8 winning multiple global first-place rankings in the international authoritative third-party evaluation from Artificial Analysis. The short-drama and AI short-drama platform business achieved a monthly run-rate of nearly USD 36 million by the end of 2025, with an ARR exceeding USD 400 million. The AI music platform Mureka has achieved both technological innovation and commercial success, while Tiangong Super Agent has been fully upgraded to provide end-to-end AI productivity solutions.
But how substantive is Kunlun Tech Co., Ltd.'s AI business? Firstly, the company's incremental selling expenses appear to exceed its AI business revenue. Is this a case of buying traffic to generate AI growth? In 2025, the Tiangong AI business generated RMB 2.369 billion in revenue, a 251.74% year-on-year increase. However, selling expenses have surged alongside the AI business boom. Wind data shows the company's selling expenses jumped from RMB 1.5 billion in 2021 to RMB 4.2 billion in 2025, with the selling expense ratio climbing from 30% to 51% over the same period, and reaching 62% in Q1 2026. Breaking down the 2025 selling expenses of RMB 4.182 billion, Opera accounted for approximately RMB 1 billion. In other words, the company likely spent over RMB 3 billion on non-Opera operations. Given that the combined revenue of Tiangong AI and Starmaker community businesses was less than RMB 3.4 billion in 2025, is Kunlun Tech Co., Ltd. effectively buying revenue through user acquisition? The company itself acknowledged in its annual report that to support user acquisition and growth for short-drama platforms, AI music, AI SuperAgent, AI social, and other products, it intensified marketing and traffic investment, resulting in RMB 4.182 billion in selling expenses, an 81.53% year-on-year increase. Notably, market and promotion expenses alone reached RMB 3.2 billion in 2025.
This raises two significant concerns. First, with an incremental RMB 2 billion in selling expenses generating only RMB 1.7 billion in additional AI revenue, is the company essentially trading money for growth? Second, from a return-on-investment perspective, are these selling expenses effective, and is the high growth sustainable in the long term?
Secondly, how does the company's large model ranking actually stack up? On June 4, 2026, Arena.ai (formerly LMArena/ChatbotArena) launched Agent Mode, specifically designed to evaluate AI systems on real-world agent tasks, including tool orchestration, long-horizon task completion, steerability, Bash recovery, and tool hallucination rates. According to Arena's latest global large model strength ranking, focusing on AI Agent capabilities such as autonomous execution in coding, website building, and tool invocation, the top 10 were Anthropic's Claude Fable 5, OpenAI's GPT 5.6 Sol, Moonshot AI's Kimi K3, Zhipu's GLM 5.2, xAI's Grok 4.5, Meta's Muse Spark 1.1, Alibaba's Qwen3.7 Max, Google's Gemini 3.1 Pro Preview, DeepSeek's V4 Pro, and Tencent's Hy3. Notably, no Tiangong model appeared in the top 50 rankings. Furthermore, in February, iiMedia Ranking released its "2025 China Large Model Evaluation Strength Ranking TOP20," which factors in company fundamentals, monthly active users, product performance, and brand influence. In that ranking, Kunlun Tech Co., Ltd.'s Tiangong was positioned only seventeenth.
Comments