How much capital expenditure will Alibaba add? Morgan Stanley says "3GW increase per year," while UBS says "1-2GW," with each 1GW approximately equal to 100 billion RMB in capital investment

Alibaba Cloud revealed its capacity roadmap at the Cloud Habitat Conference, signaling a significant change in the industry scale.

Wallstreetcn previously mentioned that Alibaba CEO Eddie Wu's statements at the conference sent very positive signals. Eddie Wu disclosed that the usage of model tokens is doubling every two to three months, indicating explosive growth in AI demand.

He also stated that the company plans to make additional investments on top of the 380 billion RMB investment over the next three years to further strengthen its computing power infrastructure. Eddie Wu also expects that by 2032, the power consumption of Alibaba Cloud's global data centers will increase tenfold compared to 2022.

On September 24, Morgan Stanley projected in its latest research report that Alibaba Cloud will add over 3 gigawatts of new data center capacity annually from 2026 to 2032, nearly equivalent to the total new capacity of the entire Chinese market in 2025. This will provide sustained and strong growth momentum for infrastructure suppliers such as data center operators.

UBS Group believes that the intensity and sustainability of Alibaba Cloud's investments "exceed market expectations." The bank estimates that Alibaba Cloud may expand its data center capacity at a rate of 1-2 gigawatts (GW) per year in the future, which could translate into incremental capital investments of up to 100 billion to 200 billion RMB annually.

Ten-Year Roadmap: Adding One "Chinese Market" Each Year

Morgan Stanley stated that global data center power consumption is expected to reach ten times that of 2022 by 2033, providing a backdrop for the large-scale expansion of cloud vendors.

The Morgan Stanley report detailed Alibaba Cloud's expansion path: Using the 2022 capacity of approximately 2.5 gigawatts as a baseline, reaching the target of 25 gigawatts in ten years means that its investment intensity will remain high for the foreseeable future. The bank expects that by 2026, Alibaba Cloud's capacity will reach approximately 5 gigawatts, and from 2026 to 2032, the annual new capacity will exceed 3 gigawatts.

Morgan Stanley indicated that this means Alibaba Cloud will need to add more than 3 gigawatts of capacity each year during the seven years from 2026 to 2032, a scale "almost equivalent to our forecast of 3-4 gigawatts of new capacity for the entire Chinese market in 2025."

This indicates that the capital expenditure from Alibaba Cloud alone is sufficient to support the growth of the data center industry for many years to come.

Based on the above analysis, Morgan Stanley holds a positive view on the data center industry, believing that strong growth in token usage, increased capital expenditure from cloud vendors, and soaring power demand all constitute direct benefits.

The report reiterated its "overweight" rating on data center operators GDS and VNET, viewing them as "key beneficiaries" of Alibaba Cloud's continued high-intensity investments.

On the other hand, the report also highlighted the disruptive risks brought by AI. As AI-native applications gradually become mainstream, the trend of "AI eats software" may pose negative impacts on traditional enterprise software companies

Investment intensity exceeds expectations, with annual capital expenditures potentially reaching hundreds of billions

UBS has translated Alibaba Cloud's targets into specific capital expenditure estimates in its report. The bank's analysts have made the following calculations:

First, UBS estimates that Alibaba Cloud's total data center capacity (IT load) in 2022 was approximately 1.5GW. If this serves as a baseline, a tenfold increase over the next decade would mean total capacity reaching 15GW, representing a net increase of about 13GW. This equates to an average annual net increase of at least 1-2GW of capacity over the next ten years. The report also notes that if improvements in Power Usage Effectiveness (PUE) are considered, the actual demand for new capacity may be even higher.

Second, UBS has linked its capacity growth forecasts to capital investment. According to the bank's estimates, each 1GW of data center demand, if fully allocated to AI, would require approximately RMB 100 billion in IT equipment capital expenditure. Based on this, an annual increase of 1-2GW in capacity directly corresponds to at least RMB 100 billion to RMB 200 billion in new capital expenditures each year.

UBS believes that this calculation significantly exceeds current market expectations, dispelling doubts about the sustainability of the AI capital expenditure cycle and the intensity of capital expenditures by large-scale vendors