AI investment is outpacing AI revenue by 46%. During the 2001 telecom bust, that gap was 32%. We're past telecom-bust territory and still accelerating toward $1T in annual capex by 2027.

Token prices tell the other side. The Silicon Data LLM Token Expenditure Index peaked at $2.06 per million tokens in May. It's at $1.62 now. Down 20% in six weeks. Down over 90% since 2023.
Total token spend has roughly doubled year over year. Cheaper tokens are expanding the market, not destroying it. The problem is the capex bill doesn't shrink when prices fall. The infrastructure costs the same whether you charge $2 or $0.50 per million tokens.
Training was a capital expense with a finite timeline. Inference is recurring. When token prices compress 90%, the margin on serving models compresses with it. The customer base grows but the unit economics thin out.
Hyperscaler compute pricing was built for a world where GPU capacity was scarce and buyers had nowhere else to go. Capacity is still scarce. The willingness to pay premium margins for it is eroding.