傅里叶的猫·2025-09-10 12:29Core Viewpoint - Oracle has provided a strong revenue guidance for AI cloud services, projecting significant growth over the next five years, with expected revenues reaching $18 billion in 2026 and $1.14 trillion by 2029 [2][3]. Group 1: Oracle's Performance and Future Projections - Oracle's future AI cloud revenue guidance indicates a substantial increase, with projections of $18 billion in 2026, $32 billion in 2027, $73 billion in 2028, and $114 billion in 2029 [2]. - The report highlights a remarkable $455 billion in Remaining Performance Obligations (RPO), indicating a strong revenue assurance for the next 3-5 years [3]. Group 2: AI Infrastructure Contracts - The growth in RPO is primarily driven by AI-related cloud infrastructure contracts, with collaborations involving major companies such as OpenAI, xAI, and Meta [5]. Group 3: Capital Expenditure Trends - Recent earnings reports from major cloud service providers (CSPs) like Google, Meta, Microsoft, and Nvidia show significant revenue and net income growth, leading to increased capital expenditure guidance for AI infrastructure [7]. - Specific capital expenditure guidance includes $85 billion from Alphabet for 2025, $66-72 billion from Meta, and $80 billion from Microsoft, all aimed at enhancing AI capabilities [8]. Group 4: Google TPU Developments - Google is expected to ship 2.5 million TPU units in 2025, with a significant portion being the V5 series, which is popular due to its cost-effectiveness and compatibility [16]. - The average selling price (ASP) of Google TPU is projected to be around $4,500, with a slight increase expected in 2026 due to new product introductions [18][21]. - By 2026, Google anticipates shipping over 3 million TPUs, reflecting a 20% increase from 2025, driven by growing AI application demands [19]. Group 5: Supply Chain Innovations - Google is experimenting with supply chain strategies, involving MediaTek for backend production to reduce costs and mitigate risks, while Broadcom remains the primary partner for front-end design [22].