Author: Black Lobster, Deep Tide TechFlow

In the summer of 1858, a copper-core cable spanned the Atlantic seabed, connecting London and New York.

The significance of this event was never about transmission speed, but about power structures; whoever laid the submarine cable could siphon off control over the flow of information. The British Empire relied on this global telegraph network to grasp intelligence from its colonies, cotton prices, and news of wars.

The strength of the empire was not just in its fleet, but also in that cable.

More than one hundred sixty years later, this logic is replaying itself in an unexpected way.

By 2026, China's large models are quietly consuming the global developer market. The latest data from OpenRouter shows that China's models accounted for 61% of the token consumption among the top ten models on the platform, with the top three all coming from China. API requests made by developers in San Francisco, Berlin, and Singapore are crossing the Pacific seabed fiber-optic cable to reach data centers in China, where computing power is consumed, electricity flows, and results are sent back.

Electricity has never left the Chinese power grid, but its value has been delivered across borders through tokens.

Mass Migration of AI Models

On February 24, 2026, OpenRouter released weekly data: the total token consumption for the top ten models on the platform was about 8.7 trillion, with Chinese models alone accounting for 5.3 trillion, representing 61%. MiniMax M2.5 topped the list with 2.45 trillion tokens, followed closely by Kimi K2.5 and Zhizhu GLM-5, the top three all coming from China.

Latest data as of February 26

This is not a coincidence; a fuse has ignited everything.

At the beginning of this year, OpenClaw emerged, an open-source tool that truly allows AI to start "working," enabling direct control of computers, executing commands, and parallel processing of complex workflows, with GitHub stars surpassing 210,000 within weeks.

Financial professional John installed OpenClaw immediately, integrated it with the Anthropic API, and started automatically monitoring stock market information, reporting transaction signals promptly. A few hours later, he stared at his account balance, frozen for a few seconds: several dozen dollars, gone.

This is the new reality brought by OpenClaw. In the past, chatting with AI consumed thousands of tokens per conversation, an insignificant expense. After integrating OpenClaw, AI runs a dozen sub-tasks simultaneously in the background, repeatedly calls contexts, and loops iterate; token consumption is not linear, but exponential. The bill accelerates like a car with its hood open, the fuel gauge dropping, unstoppable.

A "trick" began to circulate in the developer community: use OAuth tokens to directly connect Anthropic or Google's subscription accounts to OpenClaw, transforming the "unlimited" quota of monthly fee subscriptions into free fuel for AI Agents, which many developers adopted.

The official countermeasures followed swiftly.

On February 19, Anthropic updated its policy, explicitly prohibiting the use of Claude subscription credentials for third-party tools like OpenClaw. To access Claude’s functions, one must use the API billing channel. Google extensively banned subscription accounts that integrated Antigravity and Gemini AI Ultra through OpenClaw.

"The world has long suffered from Qin," John quickly turned to domestic large models.

On OpenRouter, the domestic large model MiniMax M2.5 scored 80.2% on software engineering tasks, while Claude Opus 4.6 scored 80.8%, with the gap nearly negligible. But the price difference is astronomical; the former charges $0.3 per million tokens, while the latter charges $5, a difference of about 17 times.

John made the switch, the workflow continued to operate, and the bill shrank by an order of magnitude; this migration is synchronously occurring worldwide.

OpenRouter's COO Chris Clark stated directly, the reason why Chinese open-source models can capture a large market share is that they occupy an extraordinarily high proportion in the workflow of American developers.

Electricity Export

To understand the essence of token export, one must first clarify the cost structure of a token.

It seems light, with one token approximately equal to 0.75 English words; a typical conversation with AI consumes only a few thousand tokens. But when these tokens stack up in trillions, the physical reality behind them becomes weighed down.

Breaking down the cost of a token, there are only two core components: computing power and electricity.

Computing power is the depreciation of GPUs; if you buy an Nvidia H100 for approximately thirty thousand dollars, its lifespan translates into a depreciation cost for each inference. Electricity is the fuel for continuously operating data centers, with each GPU consuming about 700 watts when fully loaded, and adding the costs of cooling systems, a large AI data center's electricity bill can easily exceed hundreds of millions annually.

Now, visualize this physical process on a map.

An American developer makes an API request from San Francisco. The data travels from California over the Pacific seabed fiber-optic cable to a data center somewhere in China, where GPU clusters start working, electricity flows from the Chinese grid to those chips, inferences are completed, and results are sent back. The entire process may only take one to two seconds.

Electricity has never left China's power grid, but the value of electricity has been delivered across borders through tokens.

There is a magical aspect here that ordinary trade cannot reach: tokens have no physical form, do not need to pass through customs, will not be impacted by tariffs, and are not counted in any current trade statistics. China has exported a vast amount of computing power and electricity services, but in official merchandise trade data, it remains almost invisible.

Tokens have become derivatives of electricity; the essence of token export is the export of electricity.

This is also beneficial due to China's relatively low electricity prices, approximately 40% lower than in the United States, reflecting a cost advantage at the physical level that competitors can easily replicate.

Additionally, China’s AI large models possess algorithmic and "involution" advantages.

The MoE architecture of DeepSeek V3 activates only part of the parameters during inference, with independent tests showing its inference cost is about 36 times lower than GPT-4o, and MiniMax M2.5 also activates only 10B of its total 229B parameters.

At the top level is involution, with companies like Alibaba, ByteDance, Baidu, Tencent, the Dark Side of the Moon, Zhizhu, MiniMax, and others competing fiercely in the same field, prices have long dropped below reasonable profit margins, and losing money while making noise has become industry norm.

Looking closely, this is similar to the export of Chinese manufacturing, leveraging supply chain advantages and industry involution to drive down token prices severely.

From Bitcoin to Tokens

Before tokens, there was a wave of electricity export.

Around 2015, power plant managers in Sichuan, Yunnan, and Xinjiang began to welcome a group of strange guests.

These individuals rented abandoned factories, filling them with densely packed machines that ran on electricity 24 hours a day. The machines produced nothing but constantly solved mathematical problems, occasionally deriving a Bitcoin from these infinite calculations.

This was the first generation of electricity export: transforming cheap hydropower and wind power into globally circulating digital assets through mining machines’ hashing operations, then cashing out as dollars on exchanges.

Electricity did not cross any borders, but its value flowed into the global market through Bitcoin.

In those years, China accounted for over 70% of the global Bitcoin mining computing power. China’s hydropower and coal power participated in a global capital redistribution through this indirect route.

In 2021, all of this came to a sudden stop. Regulatory crackdowns ensued, miners scattered, and computing power migrated to Kazakhstan, Texas, and Canada.

However, this logic never disappeared, just waiting for a new shell, until the emergence of ChatGPT and the competitive landscape of large models, former Bitcoin mining sites transformed into AI data centers, mining machines became computing GPUs, and the previously produced Bitcoins were converted into tokens, with only electricity remaining constant.

The logic of Bitcoin export and token export are structurally similar, but tokens have greater commercial value today.

Mining with mining machines is purely a mathematical calculation, and the Bitcoins produced are financial assets whose value comes from scarcity and market consensus, unrelated to “what was solved.” The computing power itself is not productive, more like a byproduct of a trust mechanism.

Large model inference is different. GPUs consume electricity, and the output is real cognitive services: code, analysis, translation, creativity. The value of tokens directly derives from their utility to users. This is a deeper embedding; once a developer's workflow relies on a specific model, the switching cost will accumulate over time and become higher.

Moreover, there is a key difference: Bitcoin mining was driven out of China, while token export is a choice actively made by global developers.

Token Wars

The submarine cable laid in 1858 represented the sovereignty of the British Empire over the information highway; whoever owns the infrastructure can define the rules of the game.

Token export is also a war without declaration, fraught with obstacles.

Data sovereignty is the first wall; an API request from an American developer is processed through a Chinese data center, and the data physically flows through China. This is not an issue for individual developers and small applications, but when it involves sensitive corporate data, financial information, and government compliance, it becomes a significant barrier. This is why the penetration rate of Chinese models is highest in development tools and personal applications, yet nearly nonexistent in core enterprise systems.

The chip ban presents a second wall; China’s AI development faces export controls on Nvidia’s high-end GPUs. MoE architecture and algorithm optimizations can only partially offset this disadvantage; a ceiling still exists.

However, the current resistance is just a prologue, with a larger battlefield taking shape.

Tokens and AI models have become a new strategic game dimension between China and the United States, comparable to semiconductors and the internet in the 20th century, and even closer to an older metaphor: the space race.

In 1957, when the Soviet Union launched Sputnik 1, the United States was shocked and immediately initiated the Apollo program, pouring in resources equivalent to thousands of billions today, ensuring not to fall behind in the space race.

The logic of AI competition is astonishingly similar, but the intensity will far exceed that of the space race. After all, space is a physical domain, undetectable to ordinary people, while AI permeates the economy's capillaries; behind every line of code, every contract, and every government decision-making system, there may be a large model from a nation. Whoever's model becomes the default option for global developers essentially gains structural influence over the global digital economy.

This is precisely why China's token export causes genuine unease in Washington.

When a developer's codebase, agent workflow, and product logic are all built around an API from a specific Chinese model, the migration costs will rise exponentially over time. By then, even if the U.S. legislates restrictions, developers will resist with their feet, just as no programmer today can abandon GitHub.

Today's token export may just be the opening chapter of this long game. Chinese large models have not claimed to overthrow anything; they merely deliver services to every developer with an API key at a lower price.

This time, the ones laying the cable are teams of engineers writing code in Hangzhou, Beijing, and Shanghai, and GPU clusters operating day and night in some province in the south.

This struggle has no countdown; it is ongoing 24 hours a day, measured by tokens, with the battlefield at every developer’s terminal.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。