JPMorgan Raises AI Data Center Capital Expenditure Forecast
JPMorgan has significantly raised its forecast for capital expenditures at the four major U.S. cloud service providers for data centers in 2026, increasing the expected growth from +52% to +63%.
The additional spending in 2026 is expected to exceed $200 billion, marking the largest annual increase; in 2027, a strong growth rate of 40% is anticipated, with additional spending surpassing $210 billion, setting a new record.
Market mechanisms indicate that hyperscaler funds are shifting significantly from traditional infrastructure to AI training and inference clusters, benefiting AI supply chain companies like Nvidia and Broadcom, as well as data center operators, while non-AI related tech expenditures face pressure, further increasing overall AI capital concentration.
Source: Public Information
ABAB AI Insight
摩根大通此前已多次上调云资本开支预测，此次从+52%直接跳升至+63%，反映四大云商（Microsoft、Amazon、Google、Meta）在2026年最新指引中持续加速AI基础设施部署，延续2023年以来AI驱动的指数级资本支出曲线。
资本路径上，四大云商将现金流与融资资源重点锁定数据中心、GPU集群及定制ASIC，2026年新增超2000亿美元主要用于H100/H200/B200等新一代算力扩容，战略目标是通过规模领先锁定AI模型训练主导权，同时为未来Agent与企业级部署储备容量。
类似案例包括2024-2025年云资本开支已创历史纪录，此次2026-2027年连续创下2000亿+年度增幅，远超以往任何技术周期；当前AI基础设施处于从爆发式扩张向持续高位平台期过渡的阶段。
本质属于资本集中：AI算力成为新一代基础设施，机制在于前沿模型训练与推理对高性能GPU/集群的极端依赖下，少数 hyperscaler 通过巨额预付合同锁定供应链，导致定价权从传统IT支出向AI专用数据中心及芯片供应商高度集中，同时推高全行业资本门槛与进入壁垒.