The AI Boom And Its Funding Run On Different Clocks

The supply side (booming): Foxconn — assembler of ~40% of the world's AI server racks — posted record Q2 revenue of T$2.513 trillion ($78.7B), up 39.8% YoY and beating estimates. June was a record month (+52.1% YoY). Foxconn raised its full-year target to ~$350.5B (+36%) and said AI rack shipments will more than double this year, with Q3 up both sequentially and annually. The demand side (paying for it): The hyperscalers and neoclouds ordering that hardware are spending far faster than they generate cash. Oracle's June quarter closed down ~11% despite a beat, on a $40B capital raise against a ~$90B FY27 capex plan. Alphabet's free cash flow fell sharply on AI capex; Amazon's trailing free cash flow ran near flat. Four-hyperscaler free cash flow sits at a multi-year low on ~$725B of combined spending. Sources: Reuters/LSEG (Foxconn), Startup Fortune (pairing) — [confirm your links for the Oracle/Alphabet/Amazon FCF figures before publishing]

The framework's Layer 2 claim is that AI capex is growing faster than the revenue and cash flow meant to justify it. Foxconn and Oracle are the two ends of that gap in the same week. Foxconn's record is a coincident read on a buildout that's already been ordered and financed — it confirms the boom is live, not that it pays for itself. The strain shows up one link downstream, on the balance sheets of the buyers: raising debt and equity to fund capex while free cash flow deteriorates. A supplier's record quarter and a buyer's widening cash deficit aren't contradictory — they're the same event seen from opposite ends. The boom gets recorded as revenue at the supplier before it gets recorded as strain at the buyer. The gap between those two clocks is the exposure.