AI Data Center Construction Faces Massive Delays Despite $750B Investment

The artificial intelligence industry's ambitious data center expansion is running into harsh physical realities. While AI companies have committed $750 billion to infrastructure investments this year alone, according to JPMorgan, only about half of the computing capacity scheduled to come online between now and 2028 is expected to meet its target date, Goldman Sachs reports.

The gap between ambition and execution is stark. AI companies have announced plans for 3,969 new US data centers, according to research firm Aterio, yet just 802 are currently under construction. The United States ended last year with 5,427 data centers, meaning the planned expansion would nearly double that figure—if it actually materializes.

Why it matters

Data center delays directly constrain AI development timelines and competitive positioning. Companies racing to deploy large language models and other AI applications depend on massive computing infrastructure that simply isn't arriving on schedule. The bottlenecks also signal that AI's infrastructure costs may be higher and take longer to realize than many investors and executives have modeled into their plans.

The supply chain squeeze

Multiple constraints are compounding to slow construction. Building materials have become difficult to source due to surging demand across the industry. More critically, the specialized chips that power AI workloads remain in short supply. Taiwan's TSMC fabricates virtually every leading AI chip, including Nvidia's Blackwell and AMD's MI300X processors, creating what Stanford University's AI Index Report calls "a single point of dependency in the global AI supply chain."

About 60% of data center capacity planned for 2027 completion hasn't broken ground yet, JPMorgan found. Another 7% of projects that started construction have since been delayed. Historically, roughly 72% of scheduled data center capacity comes online on time, but AI facilities are falling well short of that benchmark.

Power and workforce gaps

Electrical grid constraints present another fundamental challenge. Data centers already consume approximately 8% of US electricity, a figure the American Edge Project predicts could reach 12% by 2028. Many AI companies are building their own power generation plants, but wait times for generation step-up transformers have tripled, according to JPMorgan. GE Vernova reported that bookings for its power generators have doubled to $200 billion over five years.

The labor shortage is equally acute. Meeting proposed buildout timelines would require adding 500,000 electricians, 300,000 welders, and 550,000 plumbers to the US workforce, the American Edge Project estimates. Recent immigration policy changes have further tightened labor availability.

The reality check

Columbia Business School real estate professor Stijn Van Nieuwerburgh expects only 180 gigawatts of the 565 gigawatts currently planned to actually get built over the next decade—calling two-thirds of the pipeline "implausible." Many developers submit multiple simultaneous applications across regions and select only the most viable projects, Goldman Sachs noted.

Despite the delays, spending continues to surge. Data center construction jumped 7% in June to $68.3 billion, up 46% year-over-year, according to Census Bureau data. A single state-of-the-art AI campus can cost around $8 billion. The spending has grown so large that Minneapolis Federal Reserve President Neel Kashkari cited data centers as a factor fueling inflation.

These details were first reported by CNN.