If Token is oil, WeChat's "ClawBot" is the Strait of Hormuz built by Tencent.

Today's global situation is extremely tense. Since Iran blocked the Strait of Hormuz on February 28, 2026, the global energy market has been in turmoil. This waterway, with a narrowest point of only 33 kilometers, accounts for about 20% of the world's oil transportation. Approximately 20 million barrels of crude oil and refined oil are transported through it every day, accounting for nearly one-third of the world's seaborne oil trade.

In the digital world, another "strait" is quietly taking shape.

Following the full public beta launch of Tencent's Q Claw on March 20, users can "raise lobsters" without an invitation code and chat with Yuanbao directly on WeChat. On March 22, the WeChat 「Claw Bot」 plug-in was officially launched, supporting the connection to Open Claw. Users can scan the code or copy the command to connect Open Claw to WeChat. After the connection, users can quickly interact with their "lobsters" through WeChat chats.

Liu Liehong, the director of the National Data Bureau, revealed a shocking set of data at the China Development Forum in March 2026: At the beginning of 2024, the daily average token calls in China were 100 billion; by the end of 2025, it soared to 100 trillion; in March this year, it exceeded 140 trillion, a growth of over a thousand times in two years. He clearly stated that "Tokens are not only the value anchor in the intelligent era but also the'settlement unit' connecting technological supply and business demand."

This means that WeChat has not only opened its own "lobster-raising" channel but also the gateway connecting its 1.418 billion user ecosystem with external agents. What has been built this time is not a simple bridge but an oil pipeline in the AI era. Whoever controls the token distribution channel will hold the pricing power of "oil" in the AI era.

In the AI era, the "token economy" is on the rise.

01

Why Did WeChat Open the Channel?

1. A Conservative for 15 Years

Let's look back at WeChat's development over the past 15 years.

Since its official launch in 2011, it took only 433 days to reach 100 million users by 2012. From the launch of WeChat Pay in 2014 to the official release of Mini Programs in 2017 and the launch of Video Accounts in 2020, the WeChat ecosystem has always been closed.

It was hard for outsiders to enter and insiders to leave.

Even in the wave of AI, WeChat has remained restrained for a long time. Tencent's AI technologies were mostly deployed outside of WeChat, in independent products like Yuanbao App. WeChat only introduced AI capabilities in a few areas such as the search box and video account comment section. It was once regarded as a conservative in the mobile Internet field.

Since March 20, WeChat already had Yuanbao as an AI entrance. The WeChat Claw Bot plug-in is like opening another AI entrance, but it runs external agents like Claude code, Open Code, and Kimi code.

2. Reasons for Opening at This Time

Why did it suddenly break through the gap at this time?

Firstly, wild solutions forced the official to step in.

WeChat acted very quickly this time because Open Claw was advancing rapidly, and various wild WeChat connection solutions were everywhere. Instead of letting these unregulated methods prevail, it's better to build an official pipeline so that at least the message chain remains in its own hands.

It's like in the "lobster war" among big tech companies, WeChat has built a "Strait of Hormuz." With 1.418 billion users, as long as WeChat controls this "Strait of Hormuz," it can adapt to all changes.

Secondly, the "lobster war" has reached a white-hot stage.

According to Tencent's 2025 financial report, in response to the explosive demand for AI agents represented by Open Claw, Tencent has quickly launched dozens of products such as WorkBuddy, QClaw, and Tencent Cloud Lighthouse, forming a complete "lobster-raising" matrix from the cloud to the end. Ma Huateng clearly pointed out at the earnings meeting that "Lobster applications can bring AI into various rich scenarios, instead of being concentrated in chatbots as before. This can better utilize Tencent's resources and enable all business lines to work together."

What WeChat has built this time is not a simple pipeline but an oil pipeline in the AI era. This also conforms to Tencent's consistent defensive and counterattack strategy. It doesn't rush to seize the opportunity but must take a key position at critical moments.

02

Users Are Assets

1. The Fierce Lobster War

We can see that big tech companies have launched a "lobster war," investing a large amount of manpower and resources in Open Claw.

According to a research report from BofA Securities and the latest financial report data of Tencent, from 2024 to 2026, the capital expenditures (including budgets) of Alibaba, ByteDance, and Tencent were approximately 365 billion, 390 billion, and 260 billion yuan respectively. In 2026, Tencent will allocate over 36 billion yuan in budget for new AI products, at least doubling the investment compared to 2025. More than half of it will be used for AI chip procurement and computing infrastructure, about 20 billion yuan will be used for iterating the Hunyuan large model and AI applications like Yuanbao, and it will significantly increase the recruitment of top AI talents and related R & D budgets.

2. What's the Most Valuable Thing in Lobsters?

What's truly valuable in Open Claw is its Skill.

The establishment of Skill requires continuous interaction between users and AI. When AI gradually understands the user's workflow and thinking mode and can replace 80% or more of the user's daily work, the user's work ability has completed the digital transfer.

Enterprises then have the possibility of cost reduction and efficiency improvement. Instead, one person can manage multiple agents to complete work, or even an agent with "management ability" can manage multiple agents with "practical ability."

However, large language models need a vast amount of learning materials to continuously improve. Currently, the most commonly used training data sources like C4 and Refined Web have started to restrict AI companies from using their data. The non - profit research institution Epoch AI predicts that by 2026, AI companies will exhaust the fresh text resources available for training and will face a data famine.

Those implicit knowledge and Skills that only exist in human bodies and daily habits and have not been coded become the data that AI craves.

Therefore, users and their uncoded Skills are the greatest assets.

So we can see that there are already recruitment activities in the market. Enterprises provide you with Open Claw to operate agents and build Skills. When users think they are renting out the right to use their skills out of curiosity, in fact, they are providing the "fuel" to deskill themselves.

Once these Skills are learned and replicated by AI, your personal core competitiveness will cease to exist.

What's more, some of these recruitments even charge fees. They not only want your "oil" but also your money.

03

Tokens Are Like Oil

After all is said and done, how far is the implementation of agents from us?

The answer is still far away. Even if we divide the implementation into the commercial and civilian stages, it's still difficult to achieve commercial implementation first.

Why? Because the most crucial Tokens are still as expensive as oil in the early days of the gas era in the 19th century due to the lag in extraction technology and pipeline infrastructure.

However, AI and agents cannot function without it. It is the fuel for large - model reasoning, the blood for agent operation, and the industrial energy in the AI era.

Let's first look at the cost. Take AI video generation as an example. The charging standard of the AI video generation model Seedance 2.0 is: 28 yuan per million tokens for video input (video editing) and 46 yuan per million tokens without video input (pure text - to - video). According to the measured data, generating a 15 - second video consumes about 308,880 Tokens. In the pure generation mode, the cost of a single video is about 15 yuan, which is about 1 yuan per second. The price of OpenAI's Sora 2 API is even higher, costing 0.1 - 0.3 dollars per second for generating a 720p - resolution video and up to 0.5 dollars per second for a 1024p - resolution video.

This means that if it costs one yuan to generate one second of video content, based on an eight - hour workday, the daily cost will reach 480 yuan, and this is without considering any post - generation modifications. 480 yuan is equivalent to a "monthly salary of over 10,000 yuan." The high cost means that agents are still in the initial stage. According to a calculation by CITIC Construction, if the Sora APP reaches the user volume of ChatGPT, it will incur a daily reasoning cost of 14 million dollars, amounting to a staggering 5.12 billion dollars annually.

Now let's look at the prices of large - model APIs. According to the measured data in March 2026, the input/output price of GPT - 4.1 is 2.00/8.00 dollars per million Tokens, that of the domestic Deep Seek V3 is 0.27/1.10 dollars, that of Doubao Seed 2.0 Pro is 0.47/2.37 dollars, and after the price increase, Tencent's Hunyuan HY2.0 Instruct is 0.63/1.55 dollars, while Hunyuan - TurboS is 0.11/0.28 dollars.

Data from Volcengine shows that the daily average Token usage of the Doubao large model has exceeded 50 trillion, more than a ten - fold increase compared to the same period last year. On the Chinese New Year's Eve in 2026 alone, the TPM (Tokens per minute) of the Doubao large model reached a peak, with a reasoning throughput of 63.3 billion Tokens.

Looking back at human history, every time there is a productivity revolution, there must be a major infrastructure upgrade first. The advent of the mobile Internet era was underpinned by the emergence of fiber - optic cables and 5G mobile networks. Therefore, the implementation of agents must mean that the infrastructure for large models has been laid, and computing power and storage have been further developed. Only then will there be more and faster Tokens for agent operation.

Therefore, Tokens are like oil. When more oil is discovered worldwide and production capacity increases, the implementation of agents will be closer to us.

However, note that this is only commercial implementation, and there is still a long way to go before civilian use.

Take short videos as an example. In the 4G era, the cheapest telecom plan cost 100 yuan for 2GB of data. Would you be willing to use your data to watch Douyin at that time? Even in the 5G era, watching short videos on mobile data still poses a financial burden on individuals.

Therefore, even after the commercial implementation of agents, there is still a long way to go before civilian implementation.

04

Conclusion

So, even though the whole internet is telling you that if you don't use AI or know how to build agents, you will be left behind by the times. However, we still need to make rational judgments.

Even though new - energy vehicles have become mainstream, gasoline - powered cars, electric bicycles, and bicycles still have their own user groups and market positions.

Currently, AI is in a stage where it's neither very useful nor a pity to abandon. The so - called AI anxiety is as common as countless other anxieties in the world.

Of course, we also need to maintain a certain amount of curiosity and eagerness for AI.

So, instead of being anxious about catching up with AI, let AI develop for a while. First, like People's Daily, think about giving AI a good name. After all, in this era of "lobsters" running wild, the most important thing is not how many "lobsters" you raise but to figure out whether you're using an "electric car" or a "gasoline car."

This article is from the WeChat official account “Shitianhao Observation” (ID: shitianhao01). The author is Chen Cong and it is published by 36Kr with authorization.