Foreword:

A good lobster needs the seasoning of ginger, garlic, and scallions, as well as the right cooking timing, to become a delicious dish.

Today's OpenClaw is still a fresh and lively lobster with unlimited potential. However, only by adding the five essential [ginger, garlic, and scallions]—security, deployment, cost, commercialization, and ecosystem—and through time-tested cooking and refinement, can it truly be served on the productivity table.

The Lobster That Tops the Chart in 100 Days Hits the Most Precise Pain Points of the AI Industry

In March 2026, OpenClaw officially surpassed industry infrastructure projects like React and Linux, which have been refined for over a decade, with over 250,000 GitHub stars, becoming the most-starred runnable software project in GitHub history.

React took 13 years to accumulate 243,000 stars, while Linux earned 218,000 stars through 30 years of ecosystem development. OpenClaw achieved this feat in just 100 days.

OpenClaw's explosive popularity precisely addressed the core pain points that have plagued the AI industry for years. Combined with the flywheel effect of the open-source community and resonance with the zeitgeist, it created this mythical ascent to the top in 100 days.

This fundamental leap from [providing answers] to [delivering results] is the core strength that resonates with users.

The flywheel effect of the open-source community further fueled its explosive growth.

The final piece of the puzzle for this popular (explosive popularity) was the emotional resonance amid widespread AI anxiety.

However, as the hype fades, we must soberly recognize: a lively lobster without the seasoning of ginger, garlic, and scallions can never become a dish worthy of the table.

And currently, OpenClaw lacks several essential ingredients.

Companies Start [Lobster] Processing Business: Winners Behind the Hype

In a gold rush, those who profit the most are always the ones selling shovels.

The biggest winners are domestic large model providers. OpenClaw's token-hungry nature gives domestic models a chance to overtake competitors.

Over the past two years, domestic large models have been chasing the SOTA (State-of-the-Art) levels of overseas top models but struggled to find stable C-end token consumption scenarios.

OpenClaw's emergence made extreme cost-effectiveness more critical than absolute performance.

Anthropic's Claude Sonnet 4.6 charges $15 per million tokens for output, while MiniMax M2.5's overseas pricing is just $2.4—a more than sixfold difference.

Moonshot AI's Kimi K2.5, with its deep adaptation to agent scenarios and low pricing, became the top choice in the OpenClaw community.

The data doesn't lie: In February 2026, the top 10 models on the OpenRouter platform consumed about 8.7 trillion tokens, with Chinese models accounting for 61%, and the top three all being domestic models.

Within 20 days of Kimi K2.5's release, its cumulative revenue exceeded the total for 2025, with overseas revenue surpassing domestic for the first time. MiniMax's M2.5 model saw its daily token consumption grow over sixfold in February compared to December 2025. Its Hong Kong stock price surged over 480% since listing, with a market cap once exceeding HK$300 billion.

Even Zhipu's GLM-5, which once suffered service lag due to insufficient computing power, temporarily topped OpenRouter's invocation rankings thanks to OpenClaw's dividend (dividends).

OpenClaw, this overseas lobster, ultimately became the [God of Wealth] for domestic large models.

It not only brought explosive token demand for domestic models but also served as a springboard for their global expansion, breaking the monopoly of overseas large models in the high-end API market.

Cloud providers followed closely. OpenClaw's high local deployment threshold deterred ordinary users, but cloud providers' one-click deployment services solved this pain point.

Baidu Intelligent Cloud, Alibaba Cloud, and Tencent Cloud successively launched OpenClaw-exclusive images and one-click deployment solutions, allowing users to have a runnable instance in minutes with just a few clicks.

For cloud providers, OpenClaw became an excellent tool to digest inventory and lock in users.

Over the past two years, large model training demand has been cloud-centric, but inference-side computing consumption has consistently fallen short of expectations, leaving many lightweight application servers languishing in warehouses.

OpenClaw, with its 7×24 online presence and continuous consumption of computing power and memory, perfectly matches the product positioning of lightweight cloud servers.

More importantly, users drawn by OpenClaw today will stay on the platform tomorrow for other agent products. Cloud providers aren't just selling a one-time deployment service but long-term cloud resource subscriptions.

Domestic chip manufacturers also joined in. Loongson's 3B6000M chip completed local deployment of OpenClaw, while Lenovo launched notebooks and NUC mini-boxes pre-installed with OpenClaw, giving end-side AI hardware new growth opportunities amid this wave.

Behind the Shiny Star Count: The Five Missing [Ginger, Garlic, and Scallions] for the Lobster

Beneath the 250,000-star halo lies a harsh reality: GitHub data shows OpenClaw has fewer than 50,000 forks, and truly active users who completed local deployment account for less than 10% of the star count.

This stems from OpenClaw having the [body] of a lobster but lacking the five core ingredients to truly land and become a productivity tool.

① First Ingredient: The [Safety Valve] of Security and Compliance

This tool, which requires system root privileges to unleash its full capabilities, has inherent security contradictions from birth: the fewer privileges granted, the less it can do; the higher the privileges, the greater the risk of out of control (loss of control).

Security researchers found that 93% of the 42,000 OpenClaw instances exposed online have authentication bypass vulnerabilities, allowing hackers to easily take remote control.

In its skill marketplace, 12% of skills contain malicious code, with the most downloaded skill hiding a Trojan that steals user credentials.

As a result, Google, Anthropic, Meta, and South Korea's top three tech companies have issued bans restricting employees from using OpenClaw. Google even directly blocked hundreds of accounts accessing services through OpenClaw.

Gartner bluntly stated in its report that behind OpenClaw's productivity lies unacceptable cybersecurity risks for enterprises.

② Second Ingredient: The Usage Threshold from [Viral Stars] to [Real-World Deployment]

OpenClaw appears omnipotent in demo videos but often falls short in real business scenarios.

For example, when processing travel reimbursements, it can smoothly query flight tickets but gets stuck in a company's non-standardized financial system.

It understands written expense rules but cannot judge whether a 10-hour layover with a low-cost flight meets user travel needs.

It can generate standardized reimbursement forms but doesn't know which approver to send them to, completely halting when the approver is on leave or the process changes.

It can complete standardized [actions] but cannot handle the [frictions] of the real world.

In the highly structured digital world, AI seems omnipotent, but real business scenarios are filled with non-standardized processes, ambiguous rules, and complex organizational relationships—all gaps OpenClaw struggles to cross.

Many ultimately find that to complete a 10-minute task, they spend hours writing configuration files and manually debugging, effectively becoming AI's [assistant].

③ Third Ingredient: Cost Control to Escape the [Token Burner] Label

OpenClaw itself is open-source and free, but every thought, execution, and self-correction it performs consumes tokens from underlying large models.

Unlike traditional conversational AI's [no consumption unless asked] logic, OpenClaw's heartbeat mechanism, multi-round task execution, and long context carrying turn it into a never-stopping token consumption engine.

Users complain on social media that 57 messages exhausted their $40 monthly Google One quota.

According to community tests, normal heavy usage costs about 7,500 yuan monthly in tokens, with extreme scenarios exceeding 26,000 yuan.

Many excitedly deploy OpenClaw only to find that [open-source and free] is just an entry ticket, and the subsequent [shrimp-raising] costs far exceed expectations for individuals and small businesses.

④ Fourth Ingredient: A Sustainable Commercialization Path for Open-Source Projects

OpenClaw's biggest hidden worry is its lack of a clear commercialization path. Purely community-driven open-source projects cannot last long.

Linux has sustained investment from commercial companies like Red Hat, while React enjoys full resource support from Meta. These infrastructure projects that have endured for over a decade all have stable commercial revenues backing their R&D.

OpenClaw, under the MIT open-source license, allows anyone to use, modify, and even develop commercial products from it for free, meaning it can hardly generate revenue through software licensing.

More critically, any attempt at commercialization immediately faces community backlash for [betraying open-source ideals].

Today, OpenClaw stands at the same crossroads as AutoGPT once did.

Conclusion:

Open-source projects often need a commercial fulcrum to truly succeed, and OpenClaw hasn't found this fulcrum yet.

But today's OpenClaw still resembles a freshly served lobster: the main ingredient is large, and the aroma is strong.

What truly determines the flavor are the missing ginger, garlic, and scallions: industry applications, security governance, and commercial ecosystem.

References: Lijin: 'This Wave of OpenClaw—I Advise Against Jumping on the Bandwagon', Shanshang: 'The Redder OpenClaw Gets, the Happier MiniMax and Others Are', Geek Park: 'After OpenClaw, the Logic of AI Entrepreneurship Changes', China Entrepreneur Magazine: 'A Lobster Becomes the God of Wealth for MiniMax, Moonshot, and Zhipu', Zimu AI: 'Google Bans It Overnight, the Entire Industry Blocks It—What Cake Is OpenClaw Cutting?', Geek Park: 'The Explosive OpenClaw Is Repricing All AI Entrepreneurship Tracks'