Gartner 预测 AI 基础设施支出超 7 万亿美元，逾 3.5 万亿或减记
Gartner 算了笔账：7 万亿美元基建投入里可能有 3.5 万亿要打水漂，GPU 租赁商最危险，值得看看这组数字。
Gartner 预测 AI 提供商 2026 年将花费近 1 万亿美元建设基础设施，2027 至 2030 年再投入 6 万亿美元。该机构估计，到 2030 年底其中超过 3.5 万亿美元可能需要减记，因为收入无法覆盖成本。多数 AI 原生厂商采用轻资产模式，因此不到一半的 AI 收入会流向持有大量基础设施的公司。Gartner 还指出，GPU 租赁服务商风险最高，其 AI 优化 IaaS 支出将从 2027 年的 660 亿美元增长到 2030 年的 1740 亿美元，增速不足以让多数厂商在当前 AI 芯片被替换前收回投资。
Gartner forecasts that AI providers will spend nearly $1T on infrastructure in 2026 and another $6T from 2027 to 2030, and says more than $3.5T of that spending could need write-downs by the end of 2030 because revenue won’t cover its cost
Much of that revenue also flows elsewhere, because most AI-native providers run asset-light models, so less than half of AI revenue will reach companies that own significant infrastructure.
Neoclouds, the GPU rental specialists below the hyperscalers, sit most exposed because they lack the diversified revenue that could let hyperscalers fund AI investment for years.
Their market grows too slowly to help, since AI-optimized IaaS spending rises from $66 billion in 2027 to $174 billion in 2030, which Gartner says won’t let most providers recoup investments before current AI chips are replaced.
