By Fang Yue

In recent weeks, the open‑source, self-hosted AI agent OpenClaw has created huge waves in China, with users around the country racing to install it on their laptops and figure how to turn its exceptional task-executing abilities into increased efficiency at work.

In just one month, OpenClaw reached 20 million monthly active users and 250,000 stars, a marker of user interest, on GitHub, a website and platform where people store, manage, and collaborate on code. It has grown faster than Linux or React and become perhaps the fastest growing phenomenon in the history of open source.

Functioning not as a chatbot but as something more like an automated junior employee, it also possesses the potential to break the cloud monopoly of big tech, allowing enterprises to own private, controllable armies of digital “teams”.

In this article, Prof. Fang Yue, CEIBS EVE Energy Chair in Economics and Decision Sciences and Director of the CEIBS Research Centre for AI and Management Innovation, offers an in-depth analysis of what this new technology, and the excitement it is generating, means for businesses.

The OpenClaw craze sweeping China points toward one key truth: in the future, AI competition will no longer depend on how much cloud capacity you can rent, but on how strong your digital workforce is. NVIDIA 2026 AI Report shows that 86% of enterprises worldwide have increased AI investment, and 44% have already deployed intelligent agents.[MR1.1][YF1.2] In terms of technology, commercial application, and ecosystems, OpenClaw is redefining the rules of the AI game. It has also become a cultural phenomenon in China, where “raising a lobster” (yǎng lóngxiā), a reference to OpenClaw’s name and logo, has become an online buzzword among those deploying their own AI agents.

OpenClaw is not just another large model. It is an AI agent framework whose application lies in autonomous execution. In moving from “AI that talks” to “AI that works”, it completes a critical leap in AI history—from passive response to proactive action—by directly addressing long standing industry pain points. It does so through three core breakthroughs.

Breaking down technological barriers with an ecosystem explosion

First, OpenClaw redefines AI’s core value proposition: no longer merely giving answers, but delivering results. OpenClaw abandons the shallow paradigm of chat, explanation, and content generation, and instead builds a full autonomous execution loop combining a decision core, tool invocation, and path planning. Without repetitive human instructions, it can independently complete end to end workflows such as market research, data analysis, and report generation, converting AI cognition directly into deployable, measurable, and reusable productivity.

Second is architectural innovation: sovereign local deployment that shatters cloud hegemony. OpenClaw fully supports localised operation, bypassing closed cloud ecosystems at the root and delivering three disruptive values:

1.    Full data sovereignty—core data remains on premises, eliminating leakage and compliance risks at the source.

2.    Deep system control—direct access to local files, hardware devices, and even industrial control systems, allowing AI to truly enter the physical world.

3.    Unconstrained autonomous evolution—agents can analyse unknown scenarios and independently use tools to solve problems.

Enterprises are no longer tenants renting cloud compute—they are primed to become the true owners of their AI systems.

Third is an interaction upgrade: extreme simplicity and human‑like agents that make AI accessible to everyone. OpenClaw discards bloated cloud development environments in favor of lightweight, high‑efficiency interactions. Through soul.md—the file in OpenClaw’s coding that defines the agent’s behaviour, rules, and decision-making philosophy— it gives agents “personality”, transforming cold algorithms into cooperative digital employees that offer something like warmth, boundaries, and social awareness.

Thanks to its open source nature and ultra low barrier to entry, small and medium sized enterprises no longer need additional manpower to automate core business processes. Through agents like OpenClaw, AI may finally be shifting from an expensive new toy to industry wide infrastructure.

Fueled by foundational technological breakthroughs, OpenClaw has rapidly formed three key product types in China: general‑purpose AI engines, vertical localized AI agents, and lightweight edge agents designed to run directly on local devices with minimal resources, while igniting a nationwide craze for “raising a lobster” across consumers, enterprises, policymakers, and capital markets.

Three product types, building a full‑stack ecosystem

General‑Purpose Foundational Engines

With OpenClaw at the core, these Operating System (OS)‑level intelligence hubs form the backbone of the “digital employee teams” that can now be generated for enterprises. OpenClaw’s 20 million monthly active users have ignited massive demand for local computing power, with cloud providers racing to launch one‑click deployment services. This is creating a virtuous cycle of deployment, to compute consumption, to ecosystem upgrading.

Localisation and trusted‑computing pioneers

Similar AI agent projects such as KimiClaw and MiniMaxClaw, launched by China’s AI players in response to OpenClaw’s rapid rise, deeply integrate large domestic models, focusing on secure, controllable private deployments. KimiClaw reached 1.6184 million users in its first month, while MiniMax’s M2.5 model topped usage charts for three consecutive weeks. Their quick success demonstrates enterprises’ needs for data security and technological sovereignty.

Lightweight Edge Agents

Projects like PicoClaw and NanoClaw target vertical niches and embed into edge devices like laptops and phones as well as IoT hardware like cameras and sensors. This marks AI’s transition from big‑tech labs into the hands of small businesses and individual users, achieving full‑scenario, full‑level industry penetration.

A technology‑driven mass movement

On the consumer side, entirely new service tracks have emerged around the installation, local operations, and customisation of AI agents like OpenClaw. Tech giants and startups alike are offering on‑site deployment and hands‑on guidance for users keen to explore the uses of this new technology.

At one recent OpenClaw event in Shenzhen, one of China’s tech hubs, over 1,000 attendees attended to install it, a vivid symbol of the technology’s mass adoption.

Accordingly, local government is quickly stepping in. Shenzhen Longgang district’s so-called “Ten Lobster Policies” and Wuxi High‑Tech Area’s “Twelve Lobster Policies” offer up to RMB 4 million and 5 million in compute subsidies respectively. As with many emerging technologies in China, the goal is to form a dual momentum of policy leadership and industrial implementation and demonstrate state support for key sectors.

Capital markets are also responding. MiniMax and Zhipu Ai, for example, have both surpassed HKD 300 billion in Hong Kong market capitalisation, as what is known as the “lobster concept” becomes one of the hottest investment themes in Chinese tech.

Yet, we must note that for most individuals, “raising lobsters” is still mostly about entertainment and experimentation, not real value conversion. For now, it is largely limited to a desire to participate in a new technological wave.

Commercial restructuring: Rewriting the AI value matrix

OpenClaw represents not incremental change, but a fundamental disruption of AI’s commercial logic. Across applications, compute structures, and profit models, it rebuilds the industry’s value system, turning AI from an auxiliary tool into a core growth infrastructure.

Applications: Traditional Software as a Service (SaaS) subscriptions are collapsing as software delivery shifts from cloud subscriptions to local skill‑file transactions. Developers sell packaged business templates directly and enterprises connect private data and run them instantly—cutting costs in half while doubling security.

Compute structures: Compute focus moves from centralised large‑scale training to distributed, high‑frequency inference. Local deployment ignites demand for servers and GPUs, while continuous agent operation drives massive hidden token consumption, unlocking long‑term dividends. Nvidia CEO Jensen Huang has already predicted that chip architectures in this space could be a trillion‑dollar market by 2027.

Profit models: Billing evolves from per‑API calls to “digital employee” pricing. Enterprises no longer buy software accounts but purchase a set number and workload of digital employees. AI completes its leap from selling tools to selling labor.

The ultimate significance of OpenClaw is its full realisation of the six breakthrough business models previously identified by McKinsey: emotion‑first products, network‑driven commerce, micro‑segments and producers, the knowledge economy, conglomerates 3.0, and AI‑native consumer platforms. It does so with near‑zero marginal cost, comprehensively reshaping industry competition.

Its explosive rise has made one thing clear in China’s tech community: next‑generation AI systems have now arrived. The domestic market now sees four competing forces trying to leverage these developments through full-scale competition: tech giants, startups, traditional enterprises, and high-power individuals.

Tech giants: Full stack positioning and closed ecosystems. ByteDance has launched organisational products ArkClaw and internal ByteClaw; Tencent released QClaw and two others, deeply integrated with WeChat, deployable in three minutes with over 5,000 prebuilt skills; Alibaba Cloud launched an exclusive pre-installed image for OpenClaw, strengthening storage and security; Baidu has focused on mobile aggregation of mainstream models.

Startups: Vertical breakthroughs and overseas expansion. Avoiding head‑on competition, they target niche markets. Over 500 compatible projects have appeared on Product Hunt, an online platform where people discover, share, and discuss new products, in a single month; Moltbook builds agent social networks; IronClaw focuses on enterprise security; Moonshot AI has surged via overseas markets, raising USD 1.2 billion in two months.

Traditional enterprises: Accelerated, but cautious, transformation. While AI tools have delivered early cost and efficiency gains, limitations in talent, technical adaptation, and digital foundations keep most firms in a ”wait-and-see” stage.

High-power individuals: The rise of the one‑person company. 36% of OpenClaw users are developing monetisable applications; a single person can complete an entire commercial loop; financing participation is only 14.7%; survival logic is clearly focused on amplifying strengths with AI, not patching up weaknesses.

Hidden risks, and the endgame

Amid mass enthusiasm, four major risks loom in relation to OpenClaw:

Severe security vulnerabilities. While the executional ability of OpenClaw has reached a new peak for AI, safety judgment lags far behind. Simple prompts can trigger data deletion, leaks, or privilege escalation. Stanford identified 11 high‑risk vulnerabilities; ClawHub, the official public skill registry for OpenClaw agents, has logged as many as 230 malicious scripts in one week, with 48,000 exposed nodes and 35.4% susceptibility to remote code execution.

Bubble accumulation. Low entry barriers can lead to rampant cloning and homogeneity; many projects are heavily hyped yet remain in testing; OpenClaw continues to have limited practical value for ordinary users, and has arguably been overly hyped up in first-tier cities.

Critical talent shortages. 38% of enterprises cite AI talent scarcity as the biggest obstacle to adoption, especially SMEs; hybrid talents who understand both business and agent orchestration are extremely rare, and as a result, no matter how advanced the technology becomes, it remains difficult to truly transform it into the productivity of enterprises

Compliance and experience gaps. In high-value fields such as finance and healthcare, compliance requirements are extremely strict, and the autonomous execution features of OpenClaw significantly increase the complexity of compliance control; this makes compliance a tough threshold for entering core markets.

In the face of this disruptive transformation, regulators have responded swiftly, providing risk warnings and safety regulations to guide the industry forward. Enterprises, for their part, must proactively build defensive walls, using security sandboxes and minimum access rights to reinforce the bottom line of AI governance.

Despite these risks, the transformation being driven by products like OpenClaw is unstoppable. As ecosystems evolve toward smarter collaboration and AI becomes accessible on billions of devices, three irreversible trends will redefine business and human value.

The rise of high-power individuals. One person plus an AI toolchain can now command a 24/7 digital army at near zero marginal cost. For the first time, individuals now have the competence to confront large enterprises head-on. The organisational boundaries of traditional enterprises have been completely shattered; flexible and efficient "one-person enterprises" will become the dominant business model.

AI orchestration, not coding, becomes the scarcest asset. As AI tools become ubiquitous, technical development alone no longer constitutes a competitive advantage. Instead, the ability to more effectively design business rules and orchestrate automated, multi‑level AI workflows defines true competitive advantage. Leaders will shift from managing people to designing human‑AI systems.

Productivity democratization as open source breaks big‑tech monopolies. OpenClaw’s open‑source DNA enables small and medium‑sized enterprises and individual entrepreneurs to truly own the digital means of production.

Data shows that 58% of small businesses now view open source as central to their AI strategy, turning it into a universally accessible competitive weapon with the power to drive innovation from the top down.

Large models showed us that AI can think. OpenClaw shows that AI can work. The OpenClaw revolution in China is fundamentally a restructuring of production relations: AI is becoming less a tool and more of an autonomous, value‑creating, deeply collaborative digital addition to the workforce.

The ultimate question this poses to individuals is: Will you control AI, evolve alongside it, or be replaced by it?

In the AI era, human value is being redefined. Standardised execution will be replaced; creativity, empathy, judgment, and mission will remain the core advantages of human employees. The future belongs to individuals and organisations that deeply integrate human value with AI capability.

The real challenge facing the AI industry is not merely whether AI can work, but whether it can be harnessed to enable more efficient collaboration and cooperation, both with humans and with other AI agents,

The future belongs neither to speculators chasing hype nor to conservatives clinging to the past, but to those who see the essence of this technology, can rebuild their business strategy, and actively embrace evolution.

Fang Yue is EVE Energy Chair in Economics and Decision Sciences and Director of CEIBS Research Centre for AI and Management Innovation at CEIBS. His teaching and research focus on artificial intelligence and management innovation, big data and decision analysis, digital economy and corporate digital transformation, and fintech and financial engineering.