In 2027, AI capital expenditure is expected to reach $1 trillion.

The AI arms race is accelerating, and the bill is much larger than market expectations. According to Wind Trading Platform, on April 29, Bank of America Securities released a semiconductor industry research report. Vivek Arya's team of analysts raised the forecast for global ultra-large-scale cloud computing enterprise capital expenditure in 2026 to over $800 billion (a year-on-year increase of 67%) after major U.S. tech giants Alphabet Inc. Class C (GOOGL.US) (Alphabet), Microsoft Corporation (MFST.US), Amazon.com, Inc. (AMZN.US) (AWS), and Meta (META.US) successively announced their first quarter 2026 financial results and updated their capital expenditure outlooks. They also expect it to exceed $1 trillion in 2027 (a year-on-year growth of approximately 25%). Four giants collectively increase investment Alphabet Inc. Class C, Microsoft Corporation, Amazon.com, Inc., and Meta have collectively raised their 2026 capital expenditure guidance. Microsoft Corporation had the largest increase, exceeding market expectations by approximately $24 billion. Alphabet Inc. Class C: Capital expenditure in 2026 is approximately $185 billion, higher than the previous forecast of $178.3 billion; and foreshadowed a "substantial increase" in 2027. Microsoft Corporation: Capital expenditure in 2026 is $190 billion, a 61% increase from the previous year, well above market expectations of around $154 billion. Amazon.com, Inc.: Maintains a $200 billion capital expenditure guidance for 2026, with a growth rate of over 50%. Meta: Raised the median capital expenditure for 2026 to $135 billion, higher than the previous $125 billion plus Oracle Corporation ($59.6 billion). The report estimates that the combined capital expenditure of the top five technology companies in the U.S. will be approximately $769.6 billion in 2026. Including China's top cloud companies (Alibaba Group Holding Limited Sponsored ADR, Tencent, Baidu Inc Sponsored ADR Class A totaling approximately $35 billion), the total capital expenditure for global ultra-large-scale cloud companies in 2026 will exceed $800 billion, a year-on-year increase of 67%. AI sales explosion, investment returns are being realized The core logic behind this round of capital expenditure is that the commercialization of AI is rapidly being realized. Alphabet Inc. Class C: Gemini's large models process over 16 billion tokens per minute via API, a 60% increase from the previous quarter; Gemini Enterprise Edition's paid monthly active users increased by 40% month-on-month, becoming the main DRIVE for Alphabet Inc. Class C's cloud growth in the first quarter; Alphabet Inc. Class C search revenue grew by 19%, the fastest growth in recent years, driven by the rapid increase in AI search queries. Microsoft Corporation: AI business annualized revenue has exceeded $37 billion, a 123% year-on-year increase; management expects Azure cloud services to accelerate moderately in the second half of 2026 despite ongoing supply constraints. Amazon.com, Inc.: AWS first quarter revenue grew by 28% year-on-year to $37.6 billion, the fastest growth in over three years, driven by AWS's own workloads and partnerships with OpenAI, Anthropic, and other Bedrock partners. Supply constraints, pricing power tilts towards sellers Bank of America especially points out a key detail: computational power remained in short supply throughout 2026, and the expansion willingness of ultra-large-scale cloud companies is supported by customer commitments rather than purely stockpiling inventory. A more noteworthy aspect is the cost transfer logic. Microsoft Corporation specifically stated that out of its $190 billion capital expenditure budget, approximately $25 billion comes from rising prices of components such as memory, wafers, and substrates. The bank believes that this statement is a positive signal for AI semiconductor suppliers: major computing and network equipment manufacturers have the ability to pass on cost increases to customers, and pricing power and profit margins are expected to remain stable. The AI semiconductor beneficiaries named in the report include: computing power (NVIDIA Corporation), storage, semiconductor equipment, power semiconductors, and optoelectronic modules. The bank also noted that most ultra-large-scale cloud companies are equally advancing "heterogeneous deployment" - commercial GPUs and custom chips are used in parallel, rather than single-path betting. This trend implies that the market space is not a zero-sum game, and the entire AI infrastructure industry chain is expected to benefit. At the current pace, the bank predicts that global ultra-large-scale cloud company capital expenditures in 2027 will exceed $1 trillion, a year-on-year growth of approximately 25%. Alphabet Inc. Class C has already indicated a "substantial increase" in capital expenditure in 2027, and there are no signs of convergence in the expansion willingness of the other giants.