A recent analysis estimates that building and equipping a single 1-gigawatt AI data center costs about $30 billion.
Semiconductor chips are the largest expense, at roughly $17 billion. Power infrastructure adds $3 billion. The facility’s physical construction costs $2.4 billion.
The price tag highlights the capital-intensive AI race among Microsoft, Google, Amazon and Meta. Such costs create a significant barrier to entry and a major revenue opportunity for suppliers, particularly chipmakers like NVIDIA, which dominate the AI processor market.