Tech Giants Deepen Nvidia Ties as AI Race Intensifies

The generative-AI boom has made computing capacity a strategic necessity for technology companies racing to train and operate increasingly complex models. Nvidia’s lead extends beyond graphics processors to its CUDA software, networking technology and integrated systems, creating an ecosystem that is difficult and costly to replace. Even companies developing chips internally continue to work with Nvidia because losing access to its platform could slow product development and weaken their competitive position.

At its GTC conference held March 18-21, 2024, Nvidia unveiled the Blackwell architecture, whose flagship GPU contains 208 billion transistors and delivers 20 petaflops of AI performance. Nvidia said Blackwell could provide as much as 30 times the inference performance of its H100 processor for some large-language-model workloads. Amazon Web Services, Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure were among the cloud providers named as early adopters, underscoring Chief Executive Jensen Huang’s central role in the AI industry.