Nearly half of the AI data centers planned for the U.S. in 2026 won’t make it. The bottlenecks aren’t temporary — they’re structural.

📊 The Numbers

Analysts at Sightline Climate estimate that 30% to 50% of AI data centers planned for U.S. deployment this year will face delays or outright cancellations. Across 140 construction projects, data centers representing at least 16 gigawatts of capacity are slated to come online before the end of 2026.

The problem? Only about 5 GW is currently under construction. Typical build times run 12 to 18 months. The remaining 16 GW sits in the “announced” stage with no clear signs of progress.

This isn’t a new pattern. Last year, manufacturers delayed 26% of announced capacity and pushed back commercial operations for another 10%. And the 2027 pipeline is even more precarious — over 25 GW announced, but less than 10 GW being built.

⚡ The Power Problem

Insufficient power grids remain the central bottleneck. AI workloads demand staggering amounts of electricity, and the existing infrastructure simply wasn’t built for it.

“If one piece of your supply chain is delayed, then your whole project can’t deliver,” said Andrew Likens, energy and infrastructure lead at Crusoe Energy Systems.

Communities near proposed sites are seeing their energy bills climb as tech giants compete for grid capacity. A Pew Research survey found Americans are increasingly wary of AI data centers over environmental and energy concerns — while still cautiously optimistic about potential job creation.

🔧 The Supply Chain Squeeze

Hardware supply is the other half of the problem. AI data centers have already disrupted consumer tech markets by diverting production capacity for memory, storage, and CPUs — driving up prices for PCs, game consoles, and everyday devices.

But the deeper issue is more fundamental. Critical electrical components like batteries and transformers are in short supply. Despite the Trump administration’s efforts to reshore manufacturing through steep tariffs on China and other exporters, U.S. production capacity still falls far short of what the AI sector needs.

Canada, Mexico, and South Korea have become the biggest suppliers of high-power transformers for AI data centers. But even those supply chains are strained. The result: American AI companies are still relying on Chinese components to fill the gap — tariffs or not.

🏘️ Communities Are Fighting Back

It’s not just logistics. People don’t want these things in their backyard.

• New Brunswick, New Jersey scrapped a planned data center after resident backlash
• Massachusetts cities are moving to block data center development within their borders
• Rockland County, New York residents are protesting expansion over rising utility bills
• An Indianapolis councilman’s home was shot at after backing a $500 million data center project

The sentiment is growing. As one commenter put it: “Less data, more clean water, peaceful neighborhoods, fresh air. We do not need these data centers.”

🏗️ The Big Projects Stalling

It’s not just small projects feeling the squeeze:

• Oracle and OpenAI dropped plans to expand a Texas data center site (Reuters)
• OpenAI’s Stargate UK has been put on hold amid cost concerns and regulatory issues
• Multiple mega-projects announced with fanfare are quietly missing construction milestones

🔮 What Happens Next

The AI industry has been operating on the assumption that infrastructure will simply materialize to support its growth. It won’t — not at this pace, and not without tradeoffs.

Power generation takes years to scale. Transformer manufacturing can’t be reshored overnight. Communities are organizing faster than data center companies can permit. And the gap between what’s announced and what’s being built keeps widening.

This isn’t a temporary supply chain hiccup. It’s a structural reality check on the AI boom. The data centers that do get built will command premium pricing. The ones that don’t will leave big tech scrambling for alternatives — cloud partnerships, distributed computing, and more efficient architectures.

The AI revolution is still coming. It’s just going to have to squeeze through a much smaller pipe than anyone planned for.

---

🔍 THE BOTTOM LINE

The AI industry has been building like the infrastructure would just show up. It won’t. Here’s what this means in practice:

• Your cloud bills aren’t going down. Fewer data centers means less capacity, which means compute stays expensive. If you’re paying for AI API calls, budget accordingly.
• Big tech gets more concentrated, not less. The companies that already have data center capacity (Amazon, Microsoft, Google) just saw their moat get deeper. Startups building on their clouds are paying the toll.
• Community opposition is the new NIMBY frontier. If you live near cheap power and open land, expect a data center proposal — and expect your neighbors to fight it.
• The “AI will be too cheap to meter” crowd needs a reality check. Infrastructure is the bottleneck, not algorithms. The smartest model in the world is useless if there’s nowhere to run it at scale.

Think of it like highways: everyone wants to drive, nobody wants a motorway through their suburb, and the construction keeps getting delayed. AI’s rush hour is coming — but half the roads haven’t been built yet.