American technology companies have announced plans for nearly 4,000 new data centers across the country. Fewer than a quarter of them have a single shovel in the dirt. That gap between what has been announced and what is actually being built is the real story of the AI construction boom, and it is widening.

The United States ended last year with 5,427 data centers, according to Stanford University’s AI Index Report. AI companies have since announced plans for 3,969 more — a figure that would nearly double the national count, according to Aterio, a data center research firm. Of those, just 802 are currently under construction.

The reason is not public opposition, though there is plenty of it. A recent Gallup poll found 71% of Americans oppose data centers being built in their area, politicians are campaigning against them, and roughly a dozen states have floated construction moratoriums — with New York and Texas recently putting temporary bans into effect. But Goldman Sachs points to permit approvals, not bans, as the bigger obstacle standing between a developer and a finished building.

Announcements That Were Never Real

Part of the shortfall is baked into how the industry works. Developers routinely file multiple applications across several regions at once, then advance only the site that pencils out best, Goldman Sachs noted. The other applications were never firm projects; they were options.

That practice inflates the headline numbers considerably. Of the 565 gigawatts of computing power AI companies currently have on the drawing board — more than ten times what is running today — Columbia Business School real estate professor Stijn Van Nieuwerburgh expects roughly 180 gigawatts to actually get built over the next decade. He calls two-thirds of the pipeline implausible.

Even the credible third is enormous. Van Nieuwerburgh puts that buildout at about $10 trillion — 50% larger than the 19th century railroad expansion, the previous record holder for American capital spending booms. A single state-of-the-art AI campus runs around $8 billion.

Four Bottlenecks

The projects that do move forward are moving slower than planned. Historically about 72% of scheduled data center capacity comes online on time, according to Goldman Sachs. For capacity scheduled to activate between now and 2028, only about half is expected to hit its target date. Data centers typically take 18 to 24 months to build, and those timelines are stretching.JPMorgan counts $750 billion in AI infrastructure investment this year alone, yet finds that roughly 60% of capacity slated for completion in 2027 has not begun construction, with another 7% of started projects already delayed.

Four constraints explain most of it. Building materials have grown hard to source as demand surges. The chips going inside are scarcer still, concentrated at Taiwan’s TSMC, which fabricates virtually every leading AI processor including Nvidia’s Blackwell and AMD’s MI300X — described in Stanford’s report as a single point of dependency for the entire global supply chain.

Power is the second. Data centers already consume roughly 8% of US electricity, a share the American Edge Project projects will reach 12% by 2028. Companies building their own generation to compensate are hitting their own wall: wait times for generation step-up transformers have tripled, according to JPMorgan, and GE Vernova, the largest natural gas turbine maker, has seen bookings for its power generators double to $200 billion over a five-year span. Since 2020, transformers and power regulators have posted the second-steepest inflation of the 47 categories tracked in the Bureau of Labor Statistics wholesale price index.

Labor is the third and hardest to fix quickly. Meeting the announced construction schedules would require the country to add 500,000 electricians, 300,000 welders and 550,000 plumbers, per the American Edge Project — and recent immigration policy changes have not helped. “Some of our clients are developing 24/7/365, and contractors are moving around all day, but there’s nothing they can do if all the labor is tied up in existing projects,” said Joe Macejak, who heads Marsh Risk’s US property digital infrastructure business.

What Is Getting Built

The money is still flowing at record pace. Census Bureau figures show data center construction spending rose 7% in June to $68.3 billion, a 46% jump from a year earlier. There are now 438 separate developers with active US projects, according to energy data firm Cleanview. The scale has grown large enough that Minneapolis Federal Reserve President Neel Kashkari cited data centers as a contributor to inflation last week.

For contractors, electrical suppliers, and building trades, the shortage of capacity is a seller’s market. For investors, it is a warning about timing. “It’s very hard to get the timing right with these big buildouts, and often what ends up happening is we get overexcited and accrue too much debt and then a bunch of these investments go bust,” Van Nieuwerburgh said.

JBizNews Desk | New York

© JBizNews.com All Rights Reserved. Reproduction or distribution without written permission is prohibited.