Nvidia's $100B deal with OpenAI: a hilarious FT Alphaville FAQ (www.ft.com)

The Financial Times’ Alphaville reports that Nvidia and OpenAI have struck a multiyear, blockbuster agreement rumored to be worth around $100 billion, centered on supplying the GPU capacity and datacenter hardware OpenAI needs to train and run large models. Details in the piece frame the deal as a deep strategic tie — going beyond routine purchases to long-term capacity guarantees, prioritized access to next‑generation accelerators and possibly bespoke hardware or software optimizations — though precise terms and formal confirmations remain limited. For the AI/ML community this matters because compute is the single biggest bottleneck and cost driver for state‑of‑the‑art training and inference. A sustained, preferential supply line from Nvidia would accelerate model scale and iteration at OpenAI, amplify Nvidia’s pricing and market power in AI infrastructure, and could widen the gap between well‑funded labs and smaller research groups. Technically, the implications include faster turnaround for large‑scale training runs, earlier access to new accelerator features (memory, interconnect, tensor cores and software stack tuning), and stronger pressure on cloud and chip competitors to expand capacity or pursue different architectures. It raises competitive and policy questions about concentrated access to scarce AI compute resources and how that shapes research direction and deployment.