Meta, Amazon, Microsoft and Alphabet are expected to spend a combined $750 billion on capital expenditures in 2026, much of it to build data centers supporting artificial intelligence, Bloomberg reports.
The unprecedented investment is driving up costs across real estate, construction, energy and technology, as demand outpaces supplies of essential materials, equipment and skilled labor. Steel, copper and memory chips have become more expensive, while power equipment suppliers face backlogs extending beyond 2030. Land with existing electricity connections has also surged in value, and local opposition and regulatory delays are forcing some developers to revise plans or pursue costlier alternatives.
The expense of building AI infrastructure is illustrated by a hypothetical 1-gigawatt data center in Abilene, Texas, which could cost roughly $30 billion to develop and equip. Server racks and other computing hardware account for approximately half of total project costs, while electricity generation, cooling systems, fiber networks, construction and engineering add billions more.
Power availability remains a major obstacle, prompting developers to consider on-site gas-fired generation as utilities struggle to meet demand. Liquid-cooling technology is also becoming increasingly important because AI chips generate substantial heat.
The spending boom is creating opportunities across industries, including energy, construction, real estate and equipment manufacturing, while channeling billions into debt and equity markets. However, the scale of investment raises concerns about a potential AI bubble, particularly because some leading AI developers have yet to turn a profit.
More than 60 gigawatts of additional U.S. data center capacity is expected over the next four years, potentially tripling current capacity. With construction costs unlikely to ease soon, investors and developers are betting that demand for AI computing will grow enough to justify the enormous upfront expense, despite uncertainty about long-term returns.
