文章指出，虽然科技乐观者将人工智慧（AI）视为“极度去通膨”（massively disinflationary）的契机，但作者认为以今日资讯直接以 AI 作为调降利率依据为时过早。文中提到，Northern Trust 资产管理主管主张美国联准会（Federal Reserve）应保持利率稳定，而 Kevin Warsh 也认为 AI 可缓解物价压力，然而美国人口普查局（Census Bureau）资料显示，到2025年底仅不到五分之一（<20%）的美国企业已采用生成式 AI。
从实务面看，AI 对通膨的影响仍不确定。部分企业已把自动化作为裁员原因，但一项全球研究显示，超过80%的高阶主管至今尚未观察到 AI 对就业或生产力有明显影响。随著技术普及加深，效率提升若伴随劳工转向新部门并推升收入，需求端可能上升；国际清算银行（Bank for International Settlements）指出，AI 对整体通膨的最终作用，取决于供给端效率提升与需求端扩张之相对规模与速度。
在短期，AI 的基础设施与资安投入可能对通膨形成反向压力。于超大型云端供应商（hyperscalers）中，资料中心资本支出今年预估约为 7000亿美元（$700 billion），再加上 AI 相关科技股估值与模型训练、推理所需电力需求，可能推高价格。另有地缘政治冲击、老龄化支出与国防开支等需求面因素，均可能抬高「中性利率」（neutral interest rate）；因此美联储应持续监测 AI 的采用与生产率变化后再纳入决策，而非以尚未验证的预期立即制定利率政策。
The article argues that although AI is being portrayed as a potentially strongly disinflationary force, including by Northern Trust’s head of asset management and Kevin Warsh, it is still too early to use AI as the basis for immediate rate cuts. As of the end of 2025, Census Bureau data show fewer than one-fifth (<20%) of U.S. companies had adopted generative AI, so broad macro effects are not yet established.
Practical evidence remains mixed. Some firms have linked automation to redundancies, but a global study reports that more than 80% of senior executives have so far seen no clear impact on employment or productivity. As adoption expands, if efficiency gains are substantial and workers are reallocated into new sectors, rising incomes could increase aggregate demand. The Bank for International Settlements (BIS) notes that net inflation effects depend on the relative size and speed of supply gains versus demand pressures.
In the near term, AI can be inflationary through supporting expenditures: among hyperscalers, data-centre capital spending is projected at around $700 billion this year, while AI valuations and rising electricity demand from model training and inference can add price pressure. Geopolitical shocks, ageing-related spending, and defence demand add further upward risks and could lift the neutral interest rate. Central banks should therefore monitor adoption and productivity outcomes carefully and treat AI as a long-run variable, not yet a reliable basis for current interest-rate decisions.