Nvidia Recalibrates—but Reaffirms—OpenAI Funding After $100 B Megadeal Rumor

Big-ticket tech infrastructure promises have shrunk before: in 1999 Global Crossing announced $35 billion of undersea cables only to scale back sharply by 2002 when bandwidth demand proved less infinite than assumed. Huang’s climb-down echoes that era’s sober second thought—capital discipline catching up with an exponential narrative. The episode spotlights two long-running arcs. First, compute is displacing oil as the input firms hoard, driving ever-larger, vertically-tied financing loops (supplier buys its own chips). Second, the energy burden—10 GW for one customer—pushes AI toward the same regulatory and ecological scrutiny that electrification triggered after the Tennessee Valley Authority in the 1930s. Whether Nvidia pours $20 billion or $80 billion ultimately matters less than the signal that even the GPU king must diversify counterparties; hyperscale AI is leaving the honeymoon phase. A century from now this may be remembered as the moment the “AI railroad” financiers first balked at laying unlimited track, nudging the industry from exuberant land-grab to sustainable economics.