According to a research report released by China Securities Co., Ltd. (601066), overseas computing power investment continues to expand, with 2027 demand sustainability and supply delivery being key focal points. The report estimates that the capital expenditure of the five major overseas CSPs in 2027 will reach USD 1.12 trillion under a conservative scenario and USD 1.38 trillion under an optimistic scenario, corresponding to year-on-year growth of 38% and 70% respectively; leading cloud providers have relatively ample existing resources, while emerging cloud providers face greater delivery and financing pressure. Leading CSP revenue continues to grow rapidly, but order growth at Alphabet Inc. (GOOGL.US) and Microsoft Corp. (MSFT.US) has decelerated, requiring tracking of new contract signings and actual usage. Combined purchase commitments from Alphabet Inc., Amazon.com Inc. (AMZN.US), and Meta Platforms Inc. (META.US) total USD 1.63 trillion, indicating that medium-to-long-term construction willingness remains strong.
AI investment continues to weigh on cash flow, with free cash flow at Alphabet Inc., Amazon.com Inc., and Meta Platforms Inc. all having turned negative. The continued expansion of AI investment relies more heavily on cash generation from traditional core businesses; Microsoft's office segment and Meta's social business remain relatively resilient. Going forward, attention should be paid to external sales of proprietary chips, MaaS volume-price and cost dynamics, as well as the monetization value of AI search and Agent entry points.
Key takeaways from China Securities Co., Ltd.
Overseas computing power investment remains in expansion mode, with 2027 demand sustainability and supply delivery in focus. In 2025, combined capital expenditure of the five major CSPs (Microsoft Corp., Amazon.com Inc., Alphabet Inc., Meta Platforms Inc., and Oracle Corp.) totaled approximately USD 412.3 billion, up 75% year-on-year; in the first three quarters of 2026, combined capital expenditure of the five major CSPs reached approximately USD 560 billion, up 98% year-on-year. According to estimates, under conservative and optimistic scenarios for 2027, combined investment by the five companies will be approximately USD 1.12 trillion and USD 1.38 trillion respectively, corresponding to year-on-year growth of approximately 38% and 70%. Leading CSPs still have order backlog support, but the growth slope has become the core observation point for demand; on the expenditure side, purchase commitments continue to expand, and leading CSPs' medium-to-long-term computing power construction willingness remains strong.
In Q3 FY26, revenue from Google Cloud, Microsoft Azure, and Amazon Web Services grew 103%, 48%, and 39% year-on-year respectively, with new computing power coming online and efficiency improvements from existing resources jointly supporting growth. On the profitability side, margin expansion at Google Cloud and Amazon Web Services is moderating, and incremental depreciation and operating costs will be gradually reflected. However, year-on-year growth in order backlog at Alphabet Inc. and Microsoft Corp. has slowed, and large long-term contracts are relatively concentrated among leading model developers, necessitating attention to new contract signings and actual customer usage.
Regarding future expenditures, major cloud providers continue to lock in equipment, energy, and related services in advance. In Q3 FY26, purchase commitments from Alphabet Inc., Amazon.com Inc., and Meta Platforms Inc. reached USD 924.4 billion, USD 254.4 billion, and USD 454.5 billion respectively, up approximately 887%, 249%, and 460% year-on-year, totaling approximately USD 1.63 trillion.
Free cash flow has weakened across the board, and AI investment pressure has raised the bar for traditional core businesses. As infrastructure investment expands, free cash flow margins at Amazon.com Inc., Alphabet Inc., and Meta Platforms Inc. have successively turned negative, with only Microsoft Corp. still maintaining a positive figure as of Q3 FY26. Mature core businesses are an important source of funding for major cloud providers' continued AI investment, and the impact on each company's core business varies. Alphabet Inc. and Amazon.com Inc. are more directly exposed to user search and shopping decision entry points: for Google, attention should be paid to the dilution effect of AI Mode's lower monetization level on search advertising revenue; for Amazon, observation is needed on whether advertising growth can recover after the cross-quarter impact of promotions fades. Microsoft Corp.'s M365 commercial cloud revenue grew 18% with seat growth of 5%, as premium packages and AI commercialization support higher revenue per customer, demonstrating resilience in rigid office scenarios; Meta's social relationships and content consumption are difficult to directly replace with general-purpose models, and AI recommendations and advertising optimization also help with ad monetization.
Subsequent industry variables: focus on CSP proprietary chip external sales ramp-up, MaaS volume-price changes, and long-term entry point value. Track external customer expansion for TPU and Trainium as well as internal computing cost reduction for Maia and MTIA; whether AI search, Copilot, and Muse can leverage entry point advantages to generate new revenue; monitor demand and commercialization space brought by Agents; and observe whether model price reductions can be absorbed by usage growth and cost declines.
Risk warnings: Risk of AI computing power demand growth falling short of expectations, risk of customer concentration among leading clients and order fulfillment, risk of chip and server supply constraints, risk of power access and data center delivery delays, risk of new computing power utilization falling short of expectations, risk of intensifying cloud service competition and price declines, risk of rising equipment depreciation and operating costs, risk of advertising demand slowdown and user traffic shifting to third-party AI platforms, risk of AI product commercialization falling short of expectations, risk of capital expenditure returns falling short of expectations, risk of delayed customer payments and concentrated payment of procurement and leasing obligations, risk of funding gaps exceeding expectations and rising financing costs.
Comments