1. The four major US cloud providers—Google, Microsoft, Amazon, and Meta—continued to significantly increase their AI capital expenditure forecasts. Combined capital expenditures are projected to exceed $860 billion in 2026, representing a year-on-year increase of approximately 80%; and are expected to further rise to approximately $1.2 trillion in 2027, a year-on-year increase of approximately 38%.
2. This expansion is not solely based on optimistic expectations, but is supported by a backlog of over $2.3 trillion in long-term customer contracts and orders, representing a quarterly increase of approximately 16%, indicating significantly improved demand visibility.
3. AI sales at each company continue to maintain high growth rates of 80%-100%, and profit margins for both cloud and AI businesses are improving. The short-term cost is pressure on free cash flow. As capital expenditures rise to over 100%-115% of operating cash flow, the industry's free cash flow margin is projected to fall to approximately -1% in 2026, and further to -5% to -6% in 2027-2028. However, it is expected to recover thereafter as returns on AI investments materialize.
4. The five largest hyperscale manufacturers have raised approximately $270 billion this year, primarily through long-term debt and equity financing. This is largely to optimize balance sheets and maintain liquidity, and does not necessarily indicate a tight cash flow.
5. The main beneficiaries in the industry chain include computing chips, storage, semiconductor equipment, power semiconductors, and optical communications. (Source: Bank of America)
