China Mobile, one of China's largest telecom operators, has said it will invest more than 100 billion yuan in computing infrastructure in 2027 and double its AI computing capacity each year for the next two years. The remarks were made by Chairman Chen Zhongyue at the 2026 Mobile Network Conference, held in Guangzhou on November 6, 2026, and were reported by National Business Daily, which covered the event on site, and by a detailed account of the speech from Shanghai Securities News and China Securities Network (republished by 10jqka).

However, more computing capacity does not automatically translate into a larger business. China Mobile's AI computing service revenue in the first half of 2026 jumped 130.1% year on year, but it still accounts for only about one-tenth of total computing service revenue. To assess the new investment plan, it is necessary to separate the expansion of equipment and computing capacity from the revenue that actual customer usage generates.

AD

Over 100 billion yuan for computing infrastructure in 2027, with total procurement above 500 billion yuan

China Mobile's computing infrastructure investment is intended to expand equipment and computing resources to meet demand for AI and other processing needs. According to the detailed account from Shanghai Securities News and China Securities Network, the company plans to invest more than 100 billion yuan in computing infrastructure in 2027, and to bring total company-wide procurement to more than 500 billion yuan. Both figures are in renminbi.

It is important to note that the two figures cover different things. The 500 billion yuan-plus total procurement figure includes non-AI businesses, so not all of it will go to AI equipment. The same speech also set out plans for more than 100 billion yuan in terminal purchasing and sales, and more than 100 billion yuan...

Correction for precision: the speech set out plans for more than 100 billion yuan in terminal procurement and sales, and more than 10 billion yuan in resources for sales channels. The announcement can be read as placing the computing buildout within a broad, company-wide procurement and investment policy spanning the whole telecom business.

Note that National Business Daily reports the computing infrastructure investment as 100 billion yuan and total procurement as 500 billion yuan, while the detailed account of the speech attaches "or more" to both. The reports are broadly consistent on scale, but these are only plans for 2027. They do not represent actual spending or accounting-approved capital expenditure budgets. The reports also do not confirm how the investment will be classified for accounting purposes or who the specific suppliers will be.

In addition, terminal purchasing and sales differ from computing infrastructure investment in what they count and in their nature. It would be inappropriate to simply add the items together as a breakdown of business spending, or to estimate the allocation of AI-related investment from the ratio of 100 billion to 500 billion yuan. What can be confirmed at this point is that, while presenting procurement plans across a wide range of businesses, China Mobile positioned the expansion of computing infrastructure as a key investment target.

Seven gigawatt-class data center sites planned, with liquid cooling and dedicated networks

Supporting the capacity expansion is a plan to build gigawatt-class AI data center sites. According to the detailed account from Shanghai Securities News and China Securities Network, China Mobile began construction on three new sites in 2026 and plans to build two more in 2027, for a total of seven sites. Each site is to adopt a standardized modular design and use liquid cooling to handle high-density computing equipment.

The plan goes beyond building data centers. The speech also mentioned developing solutions covering AI data centers as a whole, building dedicated networks, and strengthening the company's cloud service, Mobile Cloud (移動雲). The aim is to create a foundation that keeps computing equipment running reliably, including coordination of power, water and computing resources, and delivers services to customers over the network. Introducing liquid cooling for high-density equipment is positioned as part of that groundwork.

That said, "gigawatt-class" refers to power scale and does not directly indicate AI computing performance. No details were provided that would allow conversion into GPU counts or computing performance, and the figure of seven sites does not mean all of them have been completed or are operating. There is also no basis for asserting that the seven sites will actually consume a combined 7 GW. Going forward, it will be necessary to distinguish between groundbreaking, completion and start of operations.

For current computing capacity, China Mobile's official interim report gives specific figures. The company disclosed AI computing capacity of 112,700 PFLOPS (FP16) for the first half of 2026, covering both self-built and leased equipment. PFLOPS is a unit expressing floating-point operations per second; 1 PFLOPS equals (10^{15}) operations per second. FP16 refers to 16-bit half-precision floating-point arithmetic, and this performance figure is a different metric from the power scale of a data center.

Likewise, the policy of doubling computing capacity every year for the next two years cannot be projected by simply taking 112,700 PFLOPS as a starting point. The detailed account of the speech does not make clear the reference date for the doubling or the precise definition of the computing capacity covered. Judging the actual scale of the buildout will require continued disclosure based on the same computing precision and the same scope of aggregation.

Meanwhile, the form of service China Mobile is aiming for is expanding from the traditional business of renting out computing equipment to supplying the tokens needed for tasks such as AI model inference. The official interim report also describes trials of token-related businesses, so the new plan can be seen as a further extension of existing efforts. The key question is how much of the expanded computing capacity can be tied to actual revenue.

AD

AI computing services revenue up 130.1%, about one-tenth of total computing services

According to China Mobile's 2026 interim report summary (printed pages 6–7), computing service revenue in the first half of 2026 was 52.9 billion yuan, up 14.0% year on year. The computing services as defined by the company include data centers, cloud computing services and cloud applications. Of this, AI computing service revenue was 5.3 billion yuan, a high growth rate of 130.1% year on year.

First-half 2026 revenue metric Revenue (100 million yuan, RMB) Year on year
Total computing services 529 +14.0%
AI computing services 53 +130.1%
Data centers (reference) 190 +13.2%

The 5.3 billion yuan for AI computing services and the 19.0 billion yuan for data centers are indicators for understanding the performance of computing services as a whole, not figures to be added on top of the 52.9 billion yuan. The table shows that while AI computing services is growing at an outstanding rate, it remains only a part of total computing services in revenue terms.

The share of AI computing service revenue of 5.3 billion yuan in total computing service revenue of 52.9 billion yuan can be calculated as follows.

[ \frac{53}{529}\times100 \approx 10.0% ]

This calculation uses revenue figures for the same period and in the same currency as stated in the official interim report. However, because the disclosed amounts are themselves rounded, the share of about 10.0% is also an approximation. In addition, revenue share indicates business scale; it does not represent the payback rate or profit margin on the investment planned for 2027.

That said, the fact that the revenue share is about one-tenth does not mean AI computing services contribute little to growth. According to China Mobile's interim report, AI computing services accounted for 46.1% of the increase in total computing service revenue. In other words, while its share of overall existing revenue is still small, AI computing services generated roughly half of the new revenue growth. Because revenue share and contribution to incremental revenue use different denominators, the two need to be evaluated separately.

China Mobile also disclosed that utilization of resources on its existing network exceeded 90%. However, this reflects the usage of the existing resources covered by the disclosure and does not imply future utilization of the data center sites now under construction. Building out computing capacity and having customers use that capacity to generate revenue are separate stages.

The 130.1% growth rate, the roughly 10.0% revenue share and the 46.1% contribution to revenue growth therefore each show a different aspect. The growth rate shows how quickly AI computing services are expanding, the share shows the current business scale, and the contribution to incremental revenue shows how much it has contributed to the growth of computing services overall. Combining them gives a picture of the reality of China Mobile's AI computing business before it embarks on large-scale capital spending.

What matters more than the investment amount is whether added capacity turns into revenue

According to the detailed account from Shanghai Securities News and China Securities Network, Chairman Chen Zhongyue indicated that, in building out computing infrastructure, results need to be compared and not just the amount invested. The scale of investment, over 100 billion yuan, shows that China Mobile places weight on building AI computing infrastructure, but it alone does not guarantee success as a business. To judge the results of the investment, it will be necessary to keep checking both the state of the buildout and its actual contribution to revenue.

On the equipment side, what matters is whether the planned investment is carried out and whether sites already under construction progress to completion and operation. When comparing the buildout of computing capacity, measurement conditions such as computing precision and the scope of self-built versus leased equipment need to be aligned. A rise in the PFLOPS figure alone does not allow an accurate judgment of how much capacity has actually been added.

On the revenue side, in addition to the amount of AI computing service revenue, its contribution to the increase in total computing service revenue is an important indicator. However, even if revenue grows, profitability or margins have not necessarily improved. Information including equipment operating costs is needed, and simply comparing first-half 2026 revenue with the full-year 2027 investment plan does not allow calculation of the period needed to recoup the investment.

China Mobile's goal of doubling computing capacity every year is a plan to greatly expand its supply capacity for AI infrastructure. At the same time, judging from first-half 2026 results, AI computing services are growing rapidly but still account for only about one-tenth of total computing service revenue. The focus going forward will be whether the planned sites are completed, actual usage expands and AI computing service revenue keeps rising. The value of the investment plan of over 100 billion yuan will be judged not only by the scale of the equipment but by how far it can be developed into a profitable business.