人工智慧建设投资已达兆美元级别并呈现延续性，受数据中心扩张推动，多家公司开始在财报中呈现可量化成效。与网际网路泡沫相比，当前风险仍存在，但实际生产率改善正逐步支撑投资逻辑。近期案例显示生成式与代理型 AI 在营运中提高效率并扩大利润率，包括 Trane Technologies 的建筑系统管理、Nu Skin 的客制化产品销售，以及 Allstate 与 Colgate-Palmolive 的自动化流程与行销优化。
AI 在货运市场的数据最具统计显著性。C.H. Robinson 公布其卡车部门每名员工处理的货运量在三年间上升 40%，并在 2024 至 2026 年间将 AI 带来的调整后营业利益预期自 2.2 亿美元提高至 3.36 亿美元；其股价年内上升 45%，远高于同业与 S&P 500。XPO 在三年货运低迷期中以 AI 将第三季调整后营业利益提升 10%，并扩大利润率 150 个基点；其股价上升约 5%，而竞争者普遍下跌逾 20%。主要来源为成本下降、路线与空车里程最佳化，以及更高的生产率。
基础设施需求将因电力、资本与土地供应而受节奏限制，但高度持续性。JPMorgan 估算 AI 基础建设与电力投资可能超过 5 兆美元，燃气涡轮交期已达三至四年。与可使用二十年的光纤不同，晶片更新周期更短，使投资更趋反复。当前供应端受惠者包括 Vertiv、Amphenol、GE Vernova 与 Prologis，而企业生产率提升开始反映在利润率中，形成支撑 AI 扩张的主要统计基础。
The AI build-out has reached multi-trillion-dollar scale and continues to accelerate, driven by data-center expansion, with companies now reporting measurable earnings effects. Although parallels to the dot-com bubble create correction risk, productivity gains are emerging across industries. Recent examples include Trane Technologies’ building-system automation, Nu Skin’s personalized-product engine, and operational acceleration at Allstate and Colgate-Palmolive, indicating early but tangible return on investment.
Freight logistics provides the clearest numerical evidence. C.H. Robinson increased shipments handled per employee by 40% over three years and raised expected AI-driven adjusted operating-income gains from USD 220 million in 2024 to USD 336 million in 2026; its shares are up 45% this year, beating peers and the S&P 500. XPO lifted third-quarter adjusted operating income by 10% and expanded margins by 150 basis points during a multiyear freight slump; its stock rose about 5% while competitors fell more than 20%. Gains stem from cost reduction, optimized routing, fewer empty-trailer miles, and higher worker productivity.
Infrastructure demand will be constrained by power, capital, and land availability, supporting sustained investment. JPMorgan estimates AI infrastructure and power capacity may exceed USD 5 trillion, with gas-turbine backlogs of three to four years. Unlike fiber-optic cable with a two-decade life, faster chip cycles make investment recurrent. Beneficiaries include Vertiv, Amphenol, GE Vernova, and Prologis, while rising customer-side productivity is beginning to appear in margins, forming the statistical foundation for continued AI expansion.