Nvidia Wants Customers to Buy New AI Chips — Without Giving Up Their Old Ones

Nvidia is repositioning its AI hardware strategy by emphasizing that older GPUs remain economically valuable even as it pushes customers toward newer systems like Vera Rubin. CEO Jensen Huang says Nvidia’s CUDA software stack can extend the productive life of Ampere, Hopper and Blackwell hardware, making compute more durable, rentable and financeable. The message reflects an AI market still constrained by strong demand and limited supply, where customers may continue using older chips for inference and enterprise workloads instead of immediately upgrading. The article suggests this approach could help Nvidia expand AI adoption without undermining demand for its premium newest chips, though it also raises the question of whether longer hardware lifecycles could slow replacement cycles over time.