#创作者冲榜 Lexeme, why is it the future of the crypto world?

A New Era in Crypto: From “Selling Traffic” to “Selling Lexemes,” China’s Power Grid and Telecom’s AI Token Revolution is quietly reshaping the global digital economy.
Imagine this: a late-night in March 2026, the wind farms in the Gobi Desert of Xinjiang are brightly lit, with tens of thousands of GPUs humming in low, steady buzz. At the same moment, a developer in San Francisco types on their keyboard, initiating an AI agent call to a Chinese large model—within seconds, a complete market analysis report is generated and returned.
Electricity has never left China’s power grid; but its value, in the form of “lexemes,” crosses borders instantly, fetching tens of times the premium. This is not science fiction but the reality happening right now.

The “Token” familiar to crypto enthusiasts is no longer just a term for crypto assets but is now the “digital crude oil” of the AI era—officially renamed “lexeme” by China’s National Data Bureau.
China’s daily lexeme call volume skyrocketed from 1 billion at the start of 2024 to 100 trillion by the end of 2025, surpassing 140 trillion in March 2026—over a thousandfold growth in two years. Chinese models account for 61% of token consumption on the world’s largest AI aggregation platform, OpenRouter, continuously overtaking the US for several weeks.
Why are China’s power grid and telecom no longer just “selling electricity” and “selling traffic,” but collectively shifting to “selling lexemes”? What does this transformation mean for the crypto world? How should ordinary people understand this wave of era dividends? Let’s analyze this wave of change like experts overlooking a chessboard—unraveling the logic, data, and opportunities behind it.

1. What exactly are lexemes? The “smallest settlement unit” in the AI world
Starting from the basics—don’t think of lexemes as the blockchain tokens you’re familiar with; they are the smallest “digital characters” for AI to process information.
Core definition: Lexemes are the smallest units that AI large models “understand” and “generate” text. They are not Chinese characters or English words but “semantic fragments” broken down by tokenization algorithms.
You can think of them as the ‘characters’ in the AI world.
How are they calculated?
Chinese: roughly 1 Chinese character ≈ 1 lexeme. For example, “我想吃火锅” → 6 lexemes (我、想、吃、火、锅 + punctuation handling).
English: an average of 4-5 letters ≈ 1 lexeme. “hello” = 1 lexeme; “beautiful” might be split into “beauti + ful” = 2 lexemes.
Why not use traditional “characters” or “words”?
Because using characters/words directly causes vocabulary explosion (tens of thousands of Chinese characters, hundreds of thousands of English words), making models unmanageable; single characters also break semantic integrity (e.g., “巧克力” split into “巧”“克”“力” loses meaning).
Lexemes strike a perfect balance between efficiency and semantics: manageable quantity, complete meaning, and cross-language unified processing. If you compare a large model to a master chef, lexemes are the “ingredients”—cut too coarse (using characters) wastes computing power; too fine (using letters) makes it hard to follow the recipe. Only with lexemes can efficient, intelligent service be cooked up.

2. The “stunning turnaround” of power grid and telecom: from resource selling to value selling
The end of AI is computing power; the end of computing power is electricity. But traditional electricity is hard to cross borders, and the revenue ceiling from traffic has appeared. Lexemes become the perfect “elevation.”
China’s “integrated power and computation” revolution: in western renewable-rich regions (like Guizhou, Yunnan, Xinjiang), wind and solar electricity prices are as low as 0.3 RMB per kWh. Generating 1 million lexemes consumes about 15-20 kWh, with electricity costs only a few RMB.
In international markets, the same quality lexemes are priced at $60–168 per million lexemes—more than 22 times the value! Electricity stays in the domestic grid, but the value is delivered globally—this is true “computing power going overseas.” China Telecom’s strategic shift from “traffic to lexemes”: announced at the March 2026 earnings call, with token-based operations as the main business line. Upgrading from traditional traffic management to a “five-in-one” intelligent cloud system (computing power + platform + data + models + applications). They have built over 3,000 edge computing nodes and over 25 million AI terminals, with AIDC revenue reaching 34.5 billion RMB in 2025.
Token services enable one-time monetization of computing power, networks, and data: for enterprise private deployment, public token packages, and overseas integrated plans. By 2030, China’s computing power electricity consumption is expected to reach 4.8 trillion kWh. Selling electricity is like “crude oil,” selling lexemes is like “aviation kerosene.”
Liu Leihong, director of the National Data Bureau, states plainly: “Lexemes are the value anchor and settlement unit of the intelligent era.” This is not hype but a reconfiguration of industry logic.
Worldwide, the “lobster farming” (OpenClaw agent boom) consumes massive lexemes, but China holds the cheapest “fuel”—green electricity + domestically produced computing power + efficient algorithms.

3. The “hardcore truth” crypto insiders must know: this is an infrastructure-level token economy
Veteran crypto players understand the “narrative + liquidity” of tokens, but this time, lexemes are the underlying currency driven by real demand:
China has become a global token factory: OpenRouter data shows China’s models consume over 7 trillion lexemes weekly, continuously surpassing the US. Electricity costs are only 40% of those in the US, and algorithmic competition further lowers prices (DeepSeek models cost only 2 RMB per million lexemes vs. OpenAI’s 70 RMB).
Reshaping the value chain: upstream (green electricity + GPUs), midstream (computing power scheduling), downstream (AI agent applications)—a full chain. Lexemes = “water, electricity, coal” of the AI era, billed by volume and settled in real time. Invisible trade going overseas: no customs, no tariffs, electricity stays domestically, but lexemes are monetized globally. Following the “new three” (new infrastructure, new energy, new economy), intangible computing power becomes China’s new export engine.
Huge dividends for the crypto world:
DePIN + AI integration: decentralized computing projects (like Render, Akash’s Chinese mirror) will explode. Explosive demand for lexemes = hardware tokenization, computing tokenization.
AI agent economy: platforms like OpenClaw burn millions of lexemes daily, fueling the “token economy” narrative—from Polymarket’s automated trading bots to industry-specific agents. Opportunities in vertical sectors: healthcare, legal, coding, etc., will see “2 RMB dedicated lexemes” rules. Crypto can invest in related infrastructure tokens, data tokens, or directly participate in Token Hubs (Alibaba has already established a Token Business Group).
Risks and rationality: don’t blindly chase high prices; focus on real computing power deployment and electricity cost advantages. Following trends = rational allocation of AI infrastructure-related assets, not just hype.

4. How can ordinary people simply understand this transformation?
Imagine lexemes as the “electricity bill” of the AI world: using ChatGPT once costs a few hundred lexemes’ “electricity fee.” Previously, the power grid/telecom sold “meters” (traffic/electricity), now they sell “smart services” (lexeme packages). Your home solar panels selling power to the country = traditional model; now, the country turns electricity into AI answers and sells globally—your home indirectly participates in the global AI dividend.
How can ordinary people get involved? Play with OpenClaw to “raise lobsters” → experience lexeme consumption firsthand. Develop small tools/agents using domestically produced low-cost models. Follow related listed companies (telecom, power grid partners) and potential Web3 projects.
Most importantly: understanding the underlying change allows you to proactively position yourself for the era’s dividends.

5. Future trends: lexemes will be “used on demand” like water and electricity
Looking ahead to 2030:
More efficient: tokenization + compression tech, handling more content with fewer lexemes.
Multimodal integration: text, images, videos, audio all converted into lexemes, unified processing.
Longer context: models remember longer conversations, process entire documents.
More rational billing: from pure token count → effective information + service quality-based billing.
Vertical customization: specialized lexeme rules for healthcare, legal, and other sectors. This wave of change is not just technological iteration but a triple fusion of energy, computing power, and intelligence.
China, with the world’s largest green power capacity, advanced UHV grid, and engineering talent dividend, holds the “pricing power” in the AI era. The transformation of power grids and telecoms marks China’s leap from a “resource giant” to a “value giant.”

Crypto friends, this is not just another narrative bubble but a real trillion-level infrastructure dividend. Based on your risk appetite, rationally follow: invest in computing power, focus on agent ecosystems, embrace the token economy. Ordinary people need not worry—by making good use of AI, we are all participants and beneficiaries of this transformation.
The era is here. With lexemes in hand, the future is promising.