Microsoft–Nvidia $15 B Cash-for-Compute Deal Catapults Anthropic’s Valuation

Circular, supplier-customer financing like this recalls the 1980s Japanese keiretsu cross-holdings and the 1999 dot-com vendor-financing craze (e.g., Cisco lending to web-hosting startups) that collapsed when revenue lagged capital outlays. Here, cloud landlords (Microsoft) and chip arms-dealers (Nvidia) pre-sell future capacity by injecting capital into the very AI tenants obliged to buy it—an arrangement that inflates paper valuations while masking real cash flows. The episode underscores a structural trend: AI scaling economics are shifting from software iteration to heavy-industry style cap-ex races measured in gigawatts and exaflops, concentrating power in a few hyperscalers and silicon suppliers. Whether this moment marks a durable platform shift or a bubble depends on if AI demand grows fast enough to fill those data centers over the next decade; on a 100-year horizon, it echoes previous technology manias where infrastructure built during booms (railroads 1870s, fiber optics 2000s) later became cheap, ubiquitous foundations for subsequent innovation.