Technicals — Down 0.8% to $140.49, RSI 28.7 deeply oversold, 3% off the 52-week low. Down 35.5% in a month and halved from the peak, with late June marking the worst week since the 2001 dot-com bust. An oversold bounce could fire anytime, but every moving average fans downward — there is no right-side structure at all.
Fundamentals — Bull and bear cases share one source: $638B of RPO (+363% YoY) proves demand, but ~$300B of it hangs on OpenAI — $25B of annualized revenue and loss-making. Delivering the backlog has loaded $122B of long-term debt against -$24B of free cash flow, with $40B more to raise. This is credit pricing of 'can the customer pay,' not growth-stock pricing.
News — Down 25% in H1 and 50%+ from last year's high, with Motley Fool and others running 'dire AI warning' pieces on its debt model; the market has made it exhibit A in the AI capex-bubble trial.
RSI under 30 with zero stabilization signals — credit-fear declines don't obey oversold mean-reversion; in 2001 RSI camped below 20 for a month. Wait for structure, not for cheapness.
The database franchise and OCI demand are real — if AI customers pay, today's price is a deep discount; but the debt leverage amplifies any default scenario. The payoff is bimodal, unsuited to a middling position.
- OpenAI funding trouble craters RPO conversion
- Dilution and rate costs of the $40B raise
- An AI compute-leasing price war erodes returns
- StockTwits · Dip buyers have bought three floors down, each one convinced it was the basement — there may be a sub-basement belowOriginal ↗
- 华尔街共识 · 45 analysts average $255 (+81% vs. spot) — but targets are being cut in batches, unable to keep pace with the decline
- CNBC · Late June logged the worst week since 2001 as its AI financing model became the market's obsessionSource ↗
