China's AI Divide: Why Smaller Internet Firms Are Falling Behind
China's smaller internet companies — armed with early ambition but starved of capital, talent, and scale — are being systematically shut out of the country's AI transformation, as first-quarter 2026 earnings reveal a widening chasm between tech giants and the rest of the market.
The latest earnings cycle laid bare a structural fault line in China's internet economy. While Alibaba, Tencent, and ByteDance have committed hundreds of billions of renminbi annually to AI infrastructure, a cohort of smaller platforms — Zhihu, iQIYI, Weibo and Zhiyun Group, operator of Momo and Tantan, and gaming-livestream platforms Huya and Douyu — reported Q1 2026 results that contained almost no quantifiable AI revenue, signaling that the sector's middle and lower tiers have effectively ceded the AI race before it reaches its decisive phase.
The timing is critical. As the industry pivots from chatbot-centric AI applications toward Agent-driven operating systems and hardware-embedded super-interfaces, the window for smaller players to carve out a defensible position is closing — and may already be shut.
Earnings Expose a Revenue Model Frozen in the App Era
The numbers tell the story with clinical precision.
Zhihu posted Q1 2026 revenue of RMB 650 million (approximately US$90.3 million), recovering sequentially and showing signs of stabilization after two consecutive years of revenue declines in 2024 and 2025. The company also returned to Non-GAAP profitability, reporting net profit of RMB 17.16 million (approximately US$2.4 million). Yet Zhihu disclosed no standalone AI revenue line, instead restructuring its reporting into three categories: paid content and IP operations, marketing services, and other income.
iQIYI, China's leading long-form video platform, reported Q1 2026 revenue of RMB 6.23 billion (US$865 million), with membership subscription fees accounting for close to 70% of the total and advertising contributing roughly 20%. Together, those two legacy streams generated approximately 90% of revenue. AI-related income was not separately disclosed.
Weibo generated RMB 2.9 billion (approximately US$403 million) in Q1 2026 revenue, of which advertising contributed more than RMB 2.5 billion (approximately US$347 million) — a concentration that underscores how little the platform's monetization model has evolved.
Zhiyun Group, operator of the social and dating apps Momo and Tantan, reported nearly RMB 2.4 billion (approximately US$333 million) in revenue, almost entirely from value-added services, including membership subscriptions and livestream gifting.
Huya and Douyu — whose proposed merger was blocked by regulators — continue to rely on the same livestream tipping and advertising businesses that have anchored their revenue models for years.
The pattern is consistent: these companies are generating cash, but from playbooks written in 2016, not 2026.
Early Movers Stumble: Three Years of AI Investment Yield Diminishing Returns
The irony is that most of these firms were not passive bystanders when the AI wave arrived. They moved quickly — and then stalled.
Zhihu launched its "Zhihai Tu AI" large language model in partnership with ModelBest just four months after ChatGPT's late-2022 debut, later evolving it into the "Zhihu Zhida" AI search product. When DeepSeek's R1 model went viral in early 2025, Zhihu Zhida integrated it within days. The platform held genuine structural advantages: a decade of structured, citation-rich user-generated content well-suited for both model training and retrieval-augmented generation. By investing in ModelBest, it also secured indirect access to algorithmic and compute resources.
Despite this, Zhihu's AI division has not produced measurable commercial traction. The company's most recent disclosure suggests a strategic retreat: it is now building out data-platform infrastructure to sell curated content via MCP, Skills, and API interfaces to third parties — effectively monetizing its data asset externally rather than using it to power proprietary AI products.
iQIYI followed a comparable arc. It established an AIGC Content Technology Innovation Center in mid-2023, developed internal tools such as "Script Workshop" and "Image Workshop", completed regulatory registration of its proprietary "Qizhi" large model in 2024, and launched the consumer-facing "Taodou" AI assistant in April 2025. The product's feature set — content search, broadcast reminders, conversational interaction — remains narrowly scoped and has not been cited as a revenue driver.
Weibo deployed an AI comment bot in July 2023, rebranded it as "Comment Robert" by year-end, registered its "Zhiwei" large model in 2024, and launched "openclaw-weibo" in early 2026 to ride the OpenClaw trend. Huya unveiled a desktop-form-factor AI robot at ChinaJoy 2025, designed to provide in-game coaching and companionship powered by a large model.
Each of these initiatives delivered internal productivity gains. None produced a market-facing AI business of scale.
Capital Asymmetry Defines the Battlefield
The structural explanation is straightforward: the cost of staying competitive in frontier AI has escalated beyond the reach of smaller operators.
Tencent Holdings spent RMB 31.9 billion (approximately US$4.4 billion) on capital expenditures in Q1 2026 alone — equivalent to roughly 40% of its total capital spending in 2025 (RMB 79.2 billion, or approximately US$11.0 billion).
Alibaba has announced a RMB 380 billion (approximately US$52.8 billion) AI infrastructure investment plan, with CEO Wu Yongming indicating that the company's actual AI spending over the next five years will exceed that figure. Meanwhile, market speculation has placed ByteDance's long-term AI investment ambitions as high as RMB 500 billion (approximately US$69.4 billion).
Against these figures, the smaller platforms collectively generate annual revenues in the low tens of billions of renminbi. Matching even a fraction of big-tech AI spending would consume their entire operating budgets. Declining share prices further constrain their ability to raise equity capital at viable terms.
Even DeepSeek, widely regarded as the most capital-efficient frontier AI lab globally after its R1 breakthrough in early 2025, abandoned its self-imposed restrictions on external financing after releasing its V4 model in 2026, acknowledging that scale economics ultimately cannot be engineered away.
The implication for smaller firms is stark: the foundational layers of the AI stack — base models, AI applications, coding assistants, and Agent frameworks — are structurally inaccessible to capital-constrained operators.
Mid-Tier Playbook Offers Only Partial Relief
China's mid-tier internet companies have demonstrated two viable AI strategies that smaller firms have attempted to replicate, with limited success.
Kuaishou focused its AI investment on a single vertical — AI video generation — and incubated Kling, which has entered the first tier of global AI video tools. Kling is now reportedly in the process of spinning out as an independent entity at a reported valuation of US$18 billion, a figure that would make it one of the most valuable AI video assets globally.
Bilibili adopted a different approach, positioning itself as an advertising conduit for AI companies rather than an AI product builder. With ByteDance, Baidu, and the cohort of Chinese AI startups known as the "AI Six Dragons" competing aggressively for user acquisition, Bilibili has captured a disproportionate share of AI advertising spend, driving a meaningful improvement in its profitability metrics over recent quarters.
Smaller platforms have attempted to emulate both strategies. Zhihu and Weibo have packaged AI-related content as premium advertising inventory. Several firms have explored vertical AI products. But their traffic bases are insufficient to build the high-engagement, high-conversion AI content ecosystems that make the Bilibili model work, and their vertical AI experiments have largely been wound down after one to two years without achieving commercial viability.
Agent Era Threatens to Eliminate the Small-Company Niche Entirely
The competitive dynamics are set to intensify. The industry is transitioning from a model in which AI capabilities were delivered through discrete applications — a structure that still left room for niche operators — to one in which AI Agents function as operating-system-level interfaces, aggregating demand across categories that previously supported independent businesses.
ByteDance has launched the Doubao smartphone, embedding its AI assistant at the system level. Tencent's WeChat is in discussions with handset manufacturers to deploy itself as an Agent-native platform. Internationally, OpenAI is developing an Agent-optimized smartphone, while Microsoft is building Project Solara, an Agent-oriented computing platform.
In the mobile internet era, the app store model created distribution infrastructure that was structurally agnostic to content — smartphones could not themselves provide every service, leaving extensive white space for specialized operators. The emerging Agent paradigm inverts this logic: AI giants function simultaneously as distributors, service providers, and demand generators, enabling a degree of winner-take-all concentration that has no precedent in the platform economy.
The early evidence is already visible in user behavior. Conversational AI platforms are absorbing demand for information retrieval, content discovery, and entertainment recommendation — use cases that previously sustained independent platforms. As voice and natural-language interfaces reduce the friction of switching, the structural case for maintaining separate specialized applications weakens.
For companies like Zhihu, iQIYI, and Weibo, the question is no longer whether they can build competitive AI products. It is whether the user behaviors that underpin their core businesses will survive the Agent transition intact.
Impact Assessment: What Investors Should Watch
Revenue model durability: Platforms with advertising revenue concentration above 80% — Weibo being the clearest example — face the highest structural risk as AI-driven interfaces divert search and discovery behavior away from traditional social and content platforms.
Data monetization as fallback: Zhihu's pivot toward selling data via APIs represents the most coherent near-term strategy available to content-rich smaller platforms, but it converts a potential competitive moat into a commodity input for larger rivals.
Livestream platform exposure: Huya and Douyu, whose revenue models have been static for years, face a dual threat: AI-generated content eroding the scarcity value of human streamers, and Agent interfaces reducing platform-specific engagement.
Valuation re-rating risk: As the gap between AI-native and AI-adjacent business models widens, investors are likely to apply increasingly differentiated multiples. Companies unable to demonstrate a credible path to AI revenue contribution may face sustained valuation compression regardless of near-term profitability.
The Q1 2026 earnings season has made one thing clear: in China's AI economy, the distance between the front of the pack and the rest is no longer measured in quarters. It is measured in architecture.
Related Coverage:
Tencent Cloud Turns QQ Browser into a WeChat-Controlled AI Agent with QBotClaw
Kuaishou's Kling AI Targets $20B Valuation in Planned Spinoff
ByteDance's Four AI Priorities in 2026: World Models, Coding, Video, and Monetization