AI data-center buildout drives trillions in capex and supply pressure

Tech companies and cloud providers are racing to build vast new data center capacity to support artificial intelligence, committing billions of dollars to computing, networking and storage projects across the United States and beyond. The push, concentrated over the past several years, is reshaping supply chains for chips, memory and disk drives and prompting startups and incumbents alike to redesign how data is stored and moved to meet the low-latency, high-throughput demands of modern AI workloads.

Market research firms forecast that capital spending on AI-capable data centers will run into the trillions in the coming years. One analyst group expects global capital spending on AI data centers to top $1 trillion by 2029, while other consulting estimates put total data center investment through 2030 at several trillion dollars, with the bulk earmarked for AI-capable facilities. The expansion is visible in large, remote buildouts and in a surge of orders and backlogs at infrastructure suppliers.

Nvidia’s GPUs supply most of the high-performance compute power for training large models, and several analysts put its share of the data center GPU market at roughly the low 90s percent....