China Mobile’s AI Growth: From 46.1% of Incremental Compute Revenue to Group Earnings

The most striking AI number in China Mobile’s interim results was 486.1% growth in AIDC revenue. The more informative figure was 46.1%: intelligent-computing services supplied almost half the increase in computing-services revenue in the first half of 2026. AI is producing identifiable incremental sales. The next question is whether those sales can earn the margins and cash returns needed to support a large telecom operator’s transition.^[1]

Scale provides essential context. Group operating revenue was RMB538.0 billion and profit attributable to shareholders was RMB78.9 billion. Intelligent-computing revenue was RMB5.3 billion. The new business is growing quickly but remains small relative to the group, while communications revenue is under pressure. Its investment significance depends on absolute contribution and the performance of the established business as well as growth.

The contribution to growth matters more than the fastest percentage

The release reported RMB52.9 billion of computing-services revenue, up 14.0%. Within its discussion, data-center revenue was RMB19.0 billion, up 13.2%, with AIDC revenue growing 486.1%. Intelligent-computing services generated RMB5.3 billion, up 130.1%, and contributed 46.1% of the increase in computing-services revenue.^[1] These are not independent figures that can be added into a larger AI total. AIDC appears within the data-center disclosure, without a separate absolute amount.

The absolute growth can be reconstructed. Dividing current revenue by one plus its growth rate gives prior-period computing-services revenue of approximately RMB46.404 billion and an increase of RMB6.496 billion. Intelligent computing had a reconstructed base of approximately RMB2.303 billion and an increase of RMB2.997 billion. The latter represents about 46.1% of the former, consistent with the company’s disclosure. Rounded inputs make these scale estimates, not exact historical accounts.

Intelligent computing therefore matters more to growth than to the existing revenue base. It represents about 10.0% of computing-services revenue but almost half its increase. Continued expansion could change the business mix. One interim period does not establish the duration of that growth or show that the incremental business earns the same margin as established services.

AIDC growth of 486.1% means current revenue was approximately 5.861 times the prior-year level. Without the base, neither the absolute increase nor the profit contribution can be calculated. A small business growing very quickly and a larger business growing moderately can create similar incremental revenue. The base determines the financial significance of the percentage.

How much can RMB5.3 billion of intelligent-computing revenue change earnings?

Group size determines how large a new business must become to change the overall investment case. Intelligent-computing revenue was approximately 0.99% of group operating revenue; broader computing services represented approximately 9.83%. Those proportions establish scale but cannot allocate profit. The release does not separately disclose intelligent-computing or AIDC net margins, depreciation, financing costs or invested capital.^[1]

Consider a sensitivity rather than a forecast. Assume intelligent-computing revenue rises 50% from the current level in a future comparable half-year. Incremental sales would be RMB2.65 billion. Hypothetical after-tax incremental margins of 10%, 20% and 30% would produce RMB265 million, RMB530 million and RMB795 million of additional profit. Relative to current half-year group attributable profit of RMB78.9 billion, those amounts are about 0.34%, 0.67% and 1.01%. None of the assumed margins is a disclosed segment forecast.

Two implications follow. High intelligent-computing growth can still have a limited direct effect on near-term group earnings. At the same time, accumulated scale—or additional network, cloud and industry-service sales—could increase its significance. Cross-selling needs support from customer contracts and revenue evidence; a broad product portfolio does not establish it in advance.

Valuation should gradually move from recognizing that the company offers AI services to estimating the sustainable additional profit those services generate. For now, recognized revenue, plausible but unverified margin scenarios and prospective cross-selling should remain distinct. That preserves the growth opportunity without converting technical capabilities directly into assured earnings per share.

Networks and customers provide advantages—and complicate cost analysis

China Mobile can combine connectivity, cloud platforms, facilities and operations in one enterprise-service proposal. For a customer connecting multiple locations, managing data permissions and running applications continuously, that combination may reduce supplier-coordination costs. It may also make the service harder to replace than a stand-alone GPU rental. The commercial advantage depends on actual customer savings and should ultimately appear in retention, spending or service profit.

Integration also makes analysis harder. One contract may include equipment, dedicated links, cloud resources and software implementation. Recognition schedules, purchased inputs and continuing service obligations differ. Project deliveries can grow revenue quickly but require new wins in subsequent periods. Multi-year services can create a steadier revenue stream while often requiring capital well before revenue recognition.

The company reported its “4+N+31+X” computing-infrastructure footprint, 112,700 PFLOPS of FP16 intelligent-computing capacity and in-network resource utilization above 90%.^[1] PFLOPS measures computing capability at a specified precision; it does not directly identify a GPU count. The release does not separately define that utilization figure as paid GPU occupancy, customer workload occupancy or effective computing efficiency. It cannot simply be inserted into a revenue model.

Price and cost still stand between resource utilization and profitability. The same operating time can produce different revenue under promotional pricing, internal use or different workloads. Conversely, improved scheduling can reduce idle time and increase billable output without a proportionate hardware addition. Returns on deployed capital are the economically relevant measure.

Stronger cash flow cannot all be attributed to AI

Group operating cash inflow was RMB114.9 billion, up 37.0%, and free cash flow was RMB53.9 billion, up 111.6%. These are important supports for shareholder distributions. They are group figures, however, and the release does not separately attribute cash generation to intelligent computing or AIDC.^[1] Explaining the entire improvement as AI profitability would exceed the disclosure.

Owned computing equipment creates an initial cash outflow, followed by service revenue, while depreciation is recognized over several accounting periods. Slower construction can improve free cash flow before operating profit changes substantially. An investment surge can weaken cash flow despite strong orders. Capital spending, payment dates and incremental revenue need to be placed on the same timeline.

Equipment also has a different economic life from facilities and networks. A data center may support several server generations, while a particular accelerator can face pricing pressure sooner. Sustainable returns should allow for the replacement investment needed to stay competitive. They cannot assume that the initial purchase is followed indefinitely by electricity and maintenance expenses alone. The release does not supply enough information to calculate a stand-alone intelligent-computing return on capital.

This limits attribution without diminishing the value of the existing cash-generating base. Group cash can support new investment. Whether the new business earns more than its cost of capital still requires separate observation. Expansion that continually adds depreciation and replacement costs without improving customer payments and margins can gradually consume that cash advantage.

Traditional telecom services still set the pace required of the transition

Communications-services revenue was RMB350.4 billion, down 5.7%, while reported attributable profit declined 6.3%. The company also said group revenue and profit increased on a comparable basis and cited factors including changes in VAT classification.^[1] Reported comparisons and management’s comparable-basis description should be read together. The release does not provide a complete reconciliation, so neither attributing all the decline to operating deterioration nor ignoring the reported change is justified.

That background changes what investors need from intelligent computing. If established services stabilize, incremental profit can more readily become group growth. If they remain under pressure, AI must first offset losses elsewhere. The same amount of additional revenue can therefore produce very different group earnings outcomes.

Currency also matters for shareholder distributions. The interim dividend was declared at RMB2.51 per share, up 0.3%. The announcement’s conversion produced HK$2.9003, up 5.5%.^[1] The two growth rates should not be used interchangeably as measures of an operating improvement in dividends: exchange rates and denomination arrangements affect the comparison. Analysis of A-share and Hong Kong shareholder returns must retain the relevant denomination and payment currency.

For an investor considering China Mobile as both a cash-distribution business and a growth opportunity, the question is whether those roles remain compatible. Does computing investment constrain distributable cash? Does traditional network investment stabilize? Do new services cover their incremental depreciation? Those questions explain total shareholder economics more directly than assigning a high sales multiple to an AI label.

The next evidence should connect scale, earnings and investment

The positive case already has substance. Intelligent-computing services contribute materially to computing-services growth, and an existing network and customer base may reduce the friction of selling and delivering additional services. Standardized products and better scheduling could improve returns as well as revenue by increasing reuse and billable output.

Competition can absorb those advantages. Customers can negotiate lower prices; customized projects can increase receivables and delivery costs; rapid hardware replacement can leave profit growth behind revenue growth. High utilization does not rule out low-price contracts. Strong group liquidity does not establish that every new project deserves investment.

The most useful additional disclosures would include AIDC’s absolute revenue and scope, intelligent-computing customers and contract duration, segment profitability, and the relationship between equipment spending and facilities investment. They would help distinguish expansion of low-return assets from a reusable service business. The existing release does not support a stand-alone AI valuation or price target.

The current investment conclusion is that intelligent computing has become an important source of computing-services growth, but is not yet large enough to dominate group profit. A durable change in the investment case requires incremental revenue to leave more earnings and cash, alongside stabilization in the established business. That proposition is more informative than the highest growth percentage—and can be tested against subsequent results.