AI data centers require more than chips: steel, equipment and power become critical

AI data-center expansion is increasingly viewed as a full-stack buildout spanning servers and chips, industrial inputs such as steel and equipment, and electricity. The analysis points to constraints in the production of data-center components and in available power; one assessment says additional power may not be available over the next 24 months.

Large capital projects must be planned two, three or four years ahead, creating uncertainty about demand and economic conditions when they are completed. The market is described as relatively well educated on compute, while significant innovation is expected across the infrastructure stack and new companies and technologies may emerge in the coming years.