2026年将是英伟达股价的盘整之年。 2026 年，人工智能行业将从前期的爆发式增长阶段逐 步过渡到基础设施成熟期，而受全球贸易不确定性以及超大规模云服务商竞争加剧的影响，英 伟达将面临机遇与挑战并存的局面。 【如需和我们交流可扫码添加进社群】 值得关注的是，就在分析师上一次发布英伟达分析报告后不久，谷歌推出了全新的 Gemini 3 大模型。无论是机构分析师还是终端用户，对该模型的市场反馈都颇为积极。在分析师看来， Gemini 3 的问世，标志着谷歌在定制化芯片研发领域达成了一个重要的硬件里程碑 —— 该 模型是基于谷歌自研的张量处理器（TPU）完成训练的。经过数代迭代，这些自研芯片的性能 已发展到足以对英伟达人工智能显卡构成实质性威胁的水平。Gemini 3 大模型的训练同时采 用了第五代和第六代张量处理器（v5e 与 v6 版本）。 分析师毫不意外，下一代张量处理器（TPUv7）也将很快实现大规模应用。有报道称，脸书母 公司Meta（Meta Platforms, META）、人工智能公司 Anthropic 等核心客户，正与谷歌洽 谈采购这类张量处理器用于自身人工智能业务，这也印证了谷歌自研芯片的竞 ...

Core Insights - A significant power construction plan is underway in the U.S., with major tech companies advancing towards a scale of 10GW and millions of AI chips [1] - Capital expenditure (CapEx) is a key indicator for observing power investments, with over 80% of CapEx typically allocated to power procurement [1] - By 2026, the combined capital expenditure of the five major tech companies is projected to exceed $470 billion, indicating a conservative growth rate of at least 26% [1][2] Group 1: Capital Expenditure Trends - The total capital expenditure of the five major tech companies is expected to surpass $370 billion in 2025, reflecting a growth rate of approximately 64% compared to $220.8 billion in 2024 [2] - Microsoft anticipates a capital expenditure of $937 billion for the fiscal year 2026, following a 45.1% increase in 2025 [10] - Amazon's capital expenditure is projected to be around $1.25 billion in 2025, with significant growth expected in 2026 [10][11] Group 2: Demand and Supply Dynamics - The backlog of orders (remaining performance obligation, RPO) for the five major tech companies is growing at a rate that exceeds capital expenditure growth, indicating a supply-demand imbalance [5][16] - As of Q2 2025, the total RPO for these companies is estimated at $846 billion, with significant year-on-year growth [17] - Microsoft reported nearly $400 billion in backlog orders, with demand exceeding expectations [22] Group 3: AI Chip Demand - The projected capital expenditure will fund at least 16GW of power and over 3 million AI chips, with total new investment exceeding $160 billion [12][10] - Nvidia is a direct beneficiary of this trend, with 88% of its revenue coming from data center AI chips, primarily from these major tech companies [6][8] - The demand for AI chips is expected to continue growing, driven by the increasing need for computational power in AI applications [28] Group 4: Future Outlook - The overall power market is forming a positive cycle from upstream chip supply to downstream AI application demand [9] - Concerns about potential AI bubbles persist, but the consensus is that the current investments are necessary for future growth [24][31] - Major tech companies are prioritizing capital investments to ensure they meet the growing demand for AI capabilities, with a focus on long-term resource availability [32]