The Craze Goes Offline: From Shenzhen Sidewalks to National Tours
The scene was surreal for a tech capital. Outside Tencent’s (腾讯) Shenzhen headquarters, programmers sat at folding tables, offering passersby free installation of a piece of software symbolized by a red lobster. Within days, this grassroots phenomenon spread north to Beijing, replicating outside the offices of 360 (奇虎360), Baidu (百度), and JD.com (京东). 360’s Chairman Zhou Hongyi (周鸿祎) was even spotted personally helping users install the tool.
This software, OpenClaw, isn’t another large language model or a conversational AI app. It’s an operating system for AI Agents—intelligent systems that can autonomously execute complex tasks on a computer by orchestrating calls to underlying AI models. Released in late 2025 and open-sourced in January, its lobster icon earned it the nickname ‘The Claw’. What began as a niche developer tool has now exploded into mainstream consciousness, igniting what industry observers are calling a ‘Hundred Claws War’ among China’s internet giants.
In a whirlwind week in early March, Tencent, Alibaba (阿里巴巴), ByteDance (字节跳动), Baidu, JD.com, and NetEase (网易) all announced their own deployment solutions and localized versions of OpenClaw. Tencent and Baidu promoted comprehensive ‘Claw-raising family buckets’. Hardware players like Xiaomi (小米) and Huawei (华为) also began exploring mobile-integrated versions. The scramble for dominance in the AI Agent space is now the defining battleground in Chinese tech, shifting the industry’s focus from conversational bots to autonomous digital workers.
The Strategic Imperative: Seizing the ‘Entry Point’ for the Agent Era
The frenzy is not merely about OpenClaw itself. It represents a pivotal, high-stakes race to control what experts believe will be the primary user interface for the next generation of computing: the AI Agent.
Why the ‘Hundred Claws War’ is Happening Now
Liu Zhen (刘震), founder of AI research firm Xsignal (奇异因子), frames the current moment as the third wave of the AI revolution. The first was the ‘model era’, a race for superior foundational models. The second, starting around 2024, was the ‘application explosion’. ‘2025 onwards, the industry consensus is that AI Agents will become a crucial form—AI not just as a tool, but as a digital workforce,’ Liu told China News Weekly. ‘OpenClaw is the hallmark of this third stage.’
For internet giants, controlling this new entry point is existential. It’s about retaining user attention—and the immense advertising revenue that flows from it. Liu Sixing (刘思行), an AI product manager with eight years of experience at a major tech firm, explains the threat. Modern app interfaces are meticulously engineered to guide user behavior through designed layouts, ad placements, and content sequencing. ‘If users can now simply tell an AI, “Book me a flight” or “Buy milk tea,” and have it executed, they no longer need to open Ctrip, WeChat, or Meituan,’ Liu Sixing said. ‘Those former super-apps risk becoming mere “pipes,” providing data and services but losing their grip on user attention and the associated ad ecosystem.’
This anxiety over finding a new ‘super entry point’ has haunted big tech since the rise of large models. The recent Lunar New Year period saw fierce competition for conversational AI, with Tencent investing billions in marketing for its Yuanbao (元宝) assistant and ByteDance’s Doubao (豆包) securing the exclusive interactive slot on the CCTV Spring Festival Gala. However, as Liu Zhen notes, ‘Dialogue robots are more of a Q&A entry—users ask and leave. An Agent is a “doing entry,” requiring long-term operation and offering much higher user stickiness.’
The public declaration of this strategic shift came from the very top. In the early hours of March 12, Tencent’s founder and CEO, Ma Huateng (Pony Ma, 马化腾), took to his WeChat朋友圈 (Moments) to promote the company’s multi-pronged ‘Claw’ strategy: ‘Self-developed Claw, Local Claw, Cloud Claw, Enterprise Claw, Cloud Desktop Claw, secure isolated Claw rooms, cloud security, knowledge base… and a batch of products are on the way.’ This post from one of China’s most influential tech leaders signaled the profound strategic importance his company attaches to winning the ‘Hundred Claws War’.
Beyond Hype: The Emerging Business Model for AI
The ‘Hundred Claws War’ is fueled by more than just strategic FOMO (Fear Of Missing Out). It presents the first plausible path to monetizing generative AI for the consumer market in China—a problem that has plagued the industry since its inception.
Transforming Compute from a Cost Center to a Revenue Stream
Unlike in Western markets where products like ChatGPT and Gemini can charge subscription fees, Chinese AI applications have largely been free for end-users. Tech giants have shouldered the colossal costs of GPU clusters, data centers, and model training with unclear paths to recouping these investments from consumers. OpenClaw changes this calculus dramatically.
‘The clear change brought by OpenClaw’s popularity in China is that the Token fees generated by calling large models are no longer entirely borne by the manufacturers. Users pay directly, and a significant portion of people accept this rule,’ observed Liu Zhen. This reveals a new possibility: computing power is no longer just a cost; it has the opportunity to transform into a source of recurring revenue. For China’s cloud service providers, whose primary clients have traditionally been enterprises with long sales cycles, the ‘Claw’ phenomenon demonstrates that individual users can become sustained consumers of compute and model APIs.
Consequently, the driving force behind this wave of OpenClaw promotion in China is not the AI research teams, but the cloud service divisions of these companies. Products like Alibaba Cloud’s CoPaw, Tencent Cloud’s QClaw, Baidu AI Cloud’s DuClaw, and ByteDance’s Volcano Engine ArkClaw are all tightly bundled with their respective cloud platforms, offering tiered subscription packages.
The ‘Compute Black Hole’ and the Social Platform Nexus
The business case is strengthened by OpenClaw’s insatiable appetite for compute, jokingly called a ‘compute black hole’ by netizens. Unlike a single question-and-answer interaction, an Agent executing a task engages in continuous planning, model calls, web searches, and tool use, with automatic retries on failure. A complex task can involve hundreds of model calls, consuming tens or even hundreds of times the Tokens of a simple chat.
Liu Sixing illustrates the complexity: ‘You might see one “Claw” leading an army of “Claws” to complete a task.’ For an instruction like ‘Buy me effective whitening products,’ the Agent would spawn sub-agents to research ingredients, compare brands and prices across platforms, and finally execute the purchase. This level of orchestration guzzles compute, making engaged users highly valuable to cloud providers.
Another sensitive battleground is the interaction platform. Users need a place to send instructions and receive feedback from their Agent. While international users often connect OpenClaw to Telegram, in China, ByteDance’s Lark (飞书) was an early, natural fit. This potential shift in user engagement alarmed Tencent, for whom social platforms WeChat and QQ are foundational. In response, Tencent rapidly announced QQ integration with OpenClaw, allowing users to create bots and interact with their ‘Claw’ directly within the messenger. ‘If a new usage pattern slowly shifts users to other platforms for communication and instruction, this is certainly something Tencent needs to be vigilant about,’ noted Zhou Ting, a front-end developer interviewed for the original report.
Assessing the Reality: What Can a Domestic ‘Claw’ Actually Do?
Amid the marketing blitz and executive cheerleading, the practical utility of these AI Agents for the average user remains a critical question. Early adopters are finding value, but within specific, often technical, domains.
Early Use Cases and Personalization
For developers like Zhou Ting, the ‘Claw’ has become a productivity booster for information management. She has configured an Agent to automatically read blogs from her email subscriptions, filter for valuable information, summarize it, and send her a digest. ‘It’s also starting to show clear personalization characteristics,’ she noted. Because the system has memory, it learns her habits and preferences, and its output increasingly mirrors her style. She recently published a graphic technical tutorial on Xiaohongshu (小红书) that was almost entirely completed by her ‘Claw’, which even auto-filled her account name as the author.
However, the most common uses currently involve automating repetitive digital tasks. The viral concept of ‘several Claws holding a meeting to work for you’ is still more gimmick than practical reality for most. ‘Different conversation contexts aren’t interconnected by default, so true collaboration is hard without manual configuration,’ Zhou Ting explained. ‘Most ordinary users also don’t have that kind of 24/7 Token-consuming demand.’
The Gap Between Hype and Daily Utility
Data from Xsignal shows meteoric growth, with weekly active users of OpenClaw surging from 1.9 million in late February to over 5.2 million by early March. However, Liu Zhen cautions that ‘the current value of OpenClaw on the consumer side may be overestimated.’ Much of the heat stems from curiosity-driven downloads. Whether these users retain their ‘Claws’ long-term is uncertain.
Many touted applications remain distant from the average user’s daily needs. Using it for stock analysis yields unreliable advice. Automating complex work requires API tuning and coding skills beyond most people’s reach. Having it draft a weekly report necessitates teaching it your specific format and data sources—a process that can be more tedious than writing it yourself. For tasks like booking flights or checking the weather, existing mobile apps are often more convenient. ‘The so-called “digital workforce” for the consumer end still has a considerable distance to go in real-world usage and is a way off from becoming a daily necessity for the masses,’ Liu Zhen concluded.
The Road Ahead: Productization and the Search for Killer Scenarios
The ‘Hundred Claws War’ has successfully lowered the initial technical barrier. The next, more difficult phase is transforming a powerful but raw technology into indispensable, user-friendly products that solve clear problems.
The Packaging Problem and the ‘What For?’ Question
Liu Sixing, the AI product manager, spent over a week experimenting with OpenClaw before figuring out how to use it effectively. ‘For most ordinary users, faced with the blank input box of a “Claw,” they may have no idea where to start,’ he said. ‘Because it looks more like a piece of technology than a mature product.’
Tech giants have completed the first layer of productization, turning a command-line deployment process into a ‘click-next’ installation wizard. But this is insufficient. The next step involves deeper design. ‘Usually when we use a “Claw,” we have to write a long background prompt first, like “You are a security expert, you are good at…” before the specific task. Big manufacturers next need to encapsulate this layer of background for users,’ Liu Sixing explained. The ideal state is where users don’t need to articulate their needs in perfect prose but can accomplish goals through intuitive buttons, drag-and-drop actions, or simple commands—akin to how Meitu (美图) simplified Photoshop’s complexity into one-click beauty filters.
The fundamental challenge, however, is identifying clear, high-frequency application scenarios. ‘Why did Yuanbao and Doubao catch on during the Spring Festival? Because they offered very specific usage scenarios, like ordering milk tea or writing New Year’s greetings,’ Liu Sixing pointed out. ‘The problem now is, what exactly can the “Claw” do? Even the big manufacturers haven’t fully figured it out. This is an issue that must be resolved before a true consumer-side explosion can happen.’
Navigating the ‘Hundred Claws War’: A Market in Flux
The frenzy surrounding OpenClaw and the ensuing ‘Hundred Claws War’ marks a significant inflection point for China’s tech industry. It signals a collective strategic pivot from conversational AI to autonomous, task-oriented AI Agents as the presumed next platform. The battle is being fought on three fronts: cloud infrastructure, user interaction platforms, and, most crucially, the search for a killer application.
While the short-term momentum is undeniable, driven by a mix of strategic imperative and newfound monetization potential, the long-term winners are far from decided. Success will not belong to those who merely deploy a ‘Claw’ the fastest, but to those who can most effectively productize the technology, integrate it seamlessly into users’ digital lives, and—above all—answer the simple user question: ‘What can this actually do for me, today?’
The coming months will be telling. Watch for which company moves beyond generic installation campaigns to demonstrate tangible, everyday utility. Monitor user retention metrics beyond the initial hype cycle. And observe how the integration deepens between AI Agents, cloud services, and super-app ecosystems. For global investors and observers, China’s ‘Hundred Claws War’ offers a real-time case study in how a hyper-competitive market is attempting to navigate the uncertain but transformative shift from AI as a tool to AI as an active, digital participant in work and life.