China's Three Telecom Giants Selling 'Shovels': Final Monetization of Pipe Value, Not Transformation

My judgment is clear: The essence of the three major telecom operators collectively entering the Token business isn't a business model innovation, but rather a repricing of their own network infrastructure value. The core logic of this track isn't AI applications, but the underlying support for the Token economic model.

Let's start with the conclusion. Operators selling Tokens is equivalent to upgrading their past "selling connectivity" business model into a composite model of "selling compute + selling traffic." But this is not a disruptive transformation; it is the monetization of existing monopolistic resources.

From a valuation perspective, operators have historically been positioned by the market as "utility stocks," with a valuation center around 10-15x PE. If the Token business accounts for more than 20% of revenue, the market will reprice them. The key question is: How high is the ceiling for this "selling shovels" business?

Core Question 1: What are the operators actually selling?

The Tokens mentioned in the news are essentially units of measurement for AI model invocation and computing power. Operators aren't building large AI models themselves; they are providing computing infrastructure and network channels. This is like the Gold Rush era—operators aren't digging for gold; they're selling shovels and jeans.

But there are tiers even in selling shovels. China Mobile, China Telecom, and China Unicom are selling a complete computing power network from data centers to edge nodes. The core barriers to entry for this network lie in: nationwide physical infrastructure coverage, government and enterprise customer relationships, and policy licenses.

From this angle, operators possess natural monopoly characteristics in this track. Any AI company, whether doing model training or inference, needs to use operator networks and computing power. The ceiling of this track depends on the total market size of AI Tokens.

Core Question 2: Is the business model sustainable?

Currently, the Token pricing model is clear: pay-as-you-go, similar to utilities like water, electricity, and gas. But the problem is that the marginal cost decline of AI Tokens might be faster than expected. Computing costs for large model training and inference could drop by over 30% annually due to technological progress and economies of scale. If operators cannot establish pricing power, profit margins will be compressed.

More critically, do operators possess the capability for full competition? Alibaba Cloud, Huawei Cloud, and Tencent Cloud are all doing similar things. While the three major operators have advantages in government/enterprise clients and infrastructure, they lag behind internet cloud vendors in terms of technical iteration, ecosystem building, and service capabilities.

I judge that the competitive moat for operators in this track comes more from policy resources and network coverage than from technical capability. This means they are better suited to be "infrastructure providers" rather than "AI service providers."

Core Question 3: Where is the investment value?

From an investment perspective, the value of operators' Token business lies in stable cash flow and high certainty. Demand for AI Tokens is rigid and will continue to grow as AI applications become widespread. This gives operators a stable growth engine.

However, the ceiling is also obvious. Operators are not AI application companies and cannot enjoy the high-growth premium brought by the explosion of AI apps. They are more like the "railroads and highways of the AI era," with valuation logic still leaning towards utilities, just shifting from "low-growth utilities" to "medium-growth utilities."

In terms of valuation, if the Token business can account for 20% of operator revenue, the PE ratio could rise from 10x to 15-20x. This is a reasonable range for valuation re-rating. But reaching this proportion requires the AI Tokens market to achieve exponential growth within 2-3 years, with daily Token consumption reaching at least 10 times the current level.

My Judgment:

1. Operators selling Tokens is a rational strategic choice, but not a disruptive transformation. Market hype around AI concepts will bring short-term valuation boosts, but in the long run, operators remain "shovel sellers," not "gold diggers."

2. In terms of investment logic, the core value of operators' Token business lies in certainty. Unlike the high-risk, high-reward nature of AI startups, operators offer low-risk, steadily growing cash cows. This suits investors seeking steady returns but not those chasing high growth.

3. The risk point is: Is the Token pricing model sustainable? If the marginal cost of AI Tokens drops too quickly, or if internet cloud vendors steal market share through technical advantages, operators' profit margins may be compressed.

Finally, leaving an open question: When AI applications truly explode and Token demand reaches the trillion-level, will operators continue to do the "shovel-selling" business, or will they have the opportunity to become "platform players"? The answer depends on whether operators can build AI application ecosystem capabilities beyond just computing networks. If it's the latter, the valuation logic will be completely different.

Original Link: https://www.tmtpost.com/8074796.html