Nvidia-OpenAI $100 B Data-Center Deal Stalls, Talks Pivot to Smaller Equity Stake

Capital-heavy tech love affairs often cool once spreadsheets arrive: in 1985, IBM’s plan to pour $1 B into the up-start Microelectronics & Computer Technology Corp collapsed within months when internal bean-counters balked at unclear returns— eerily similar to Nvidia’s jetzige volte-face. The stall exposes two structural currents: (1) the AI arms race is morphing into a raw infrastructure contest where chipmakers, cloud titans and model labs must lock in multi-decade power and fabrication capacity, and (2) hardware suppliers are no longer passive vendors but quasi-investment banks shaping their customers’ balance sheets. Whether this deal revives or not, the moment signals that even trillion-dollar optimism has a cost-of-capital governor; the railroads of the 1870s likewise over-promised track mileage before a financing crunch forced consolidation. On a 100-year arc, the episode will matter only if it marks the point when AI compute ceased to be a limitless credit fountain and began following the boom-bust cadence of every previous General-Purpose Technology build-out—from electrification (1880s) to fiber optics (1990s).