AI Data Centers Now Cost Tens of Billions Each as Big Tech Capex Nears $750B This Year
Summary
- • Individual AI data centers now reportedly cost tens of billions of dollars each to build
- • Meta, Amazon, Microsoft and Alphabet are projected to spend about $750B combined on capex this year
- • Memory prices are reportedly more than 5x higher than a year ago, a major cost driver
- • A theoretical 1 GW Abilene, Texas site is estimated at roughly $30B at the high end
Details
Each AI data center now costs tens of billions of dollars to build
Per-site costs have reached a scale once associated with national infrastructure projects. The figure covers construction, power, and the computing hardware inside.
Meta, Amazon, Microsoft and Alphabet projected to spend about $750B combined in capex this year
That is reportedly more than half of the total capital expenditure of the entire S&P 500 in the prior year, showing how concentrated AI infrastructure spending is among four companies.
Memory prices are more than 5x higher than a year ago
Analyst estimates cite memory as a leading driver of rising build costs, because AI servers need large amounts of high-performance memory per system.
Backlogs for power generation and transmission equipment extend beyond 2030
Electricity supply is a hard constraint on new capacity. Long lead times for generation and grid equipment can delay projects regardless of how much money is available.
Prices of steel, cement, chips and copper are elevated
Demand for physical construction materials adds to costs on top of hardware, compounding the cost increases from memory and power equipment.
A theoretical 1 GW AI data center in Abilene, Texas could cost about $30B at the high end
The breakdown comes from Cushman & Wakefield, combined with analyst hardware estimates. It is a theoretical model, not a specific announced project, and hardware makes up much of the total.
More than 60 GW of capacity is slated for the US over four years, tripling current levels
The pipeline implies very large cumulative spending, though actual costs will vary by site and configuration.
Financials = costs/spending, Stat = headline figure, Market Impact = price and supply pressures, Infrastructure = physical buildout and power
What This Means
Building AI capacity has become one of the largest capital spending efforts in the US economy, concentrated in a handful of companies. Rising prices for memory, materials and especially power equipment mean that money alone does not guarantee faster expansion. Reports of equipment backlogs stretching past 2030 suggest supply constraints could shape who can scale AI and how quickly. Figures are estimates.