沃爾瑪的AI成本失控，現開始對員工代幣實行配給
沃爾瑪在內部大型語言模型“code puppy”的成本超出預期後，對員工實施了代幣使用限制，這標誌著企業AI消費正從無限轉向計量模式。
walmart capped internal use of 'code puppy' after employees, previously encouraged to experiment freely, generated unexpectedly high costs through heavy usage and 'token maxxing'. the retailer now allocates fixed token amounts to staff, aiming to curb spending on expensive models for routine tasks like spreadsheet analysis. this reflects a wider trend as providers shift from flat subscriptions to pay-per-use billing, a financial shock that recently caused uber to exhaust its 2026 ai budget in just four months. the change forces large enterprises to treat ai inference as a variable cost rather than a fixed utility, prioritizing roi over unfettered access.