According to TechRepublic, Turner & Townsend’s 2025-2026 Data Center Construction Cost Index reveals the AI construction boom is hitting serious obstacles. The report surveyed 280 industry experts across 52 global markets and found that 48% cite power availability as the main obstacle to meeting delivery schedules. Even more striking, 83% of experts believe supply chains aren’t equipped to deliver advanced cooling technologies for AI data centers. Global construction cost inflation for traditional data centers is projected at 5.5% in 2025, while AI-ready facilities in the U.S. carry a 7-10% cost premium. Tokyo leads as the world’s most expensive market at $15.2 per watt, followed by Singapore at $14.5 per watt and Zurich at $14.2 per watt.
Sponsored content — provided for informational and promotional purposes.
The Power Problem
Here’s the thing that really jumps out from this research: power has become “the biggest single challenge” for data center delivery. Nearly half of the experts surveyed pointed to energy grid access as their primary bottleneck. And we’re not talking about small power requirements here – AI data centers consume dramatically more electricity than traditional facilities. Basically, the utilities that were built to serve regular data centers are now being asked to support facilities that might use 5-10 times more power. That’s creating a massive infrastructure gap that can’t be solved overnight.
Supply Chain Failures
But wait, it gets worse. The fact that 83% of experts think supply chains can’t handle advanced cooling tech is absolutely staggering. We’re talking about liquid cooling systems, specialized heat exchangers, and other complex equipment that most suppliers simply aren’t set up to manufacture at scale. These aren’t your grandfather’s server racks with basic air conditioning. AI chips generate insane amounts of heat that require sophisticated cooling solutions. And the supply chain for that technology? It’s basically nonexistent at the scale needed for this AI gold rush.
The Cost Explosion
So what does all this mean for the bottom line? Well, construction costs are going through the roof. That 7-10% premium for AI-ready facilities in the U.S. is just the beginning. When you look at markets like Tokyo at $15.2 per watt, you start to understand why some companies are getting priced out of the AI race. These aren’t incremental increases – we’re talking about fundamental shifts in what it costs to build computing infrastructure. And the full report suggests this is only going to get worse as demand continues to outstrip supply chain capacity.
Adapt or Fail
Paul Barry from Turner & Townsend put it bluntly: developers need to embrace off-grid solutions and strengthen supply chains. But here’s the million-dollar question: how quickly can the industry actually adapt? We’re talking about redesigning entire facilities, rethinking energy sourcing, and building new manufacturing capacity for cooling technology. That doesn’t happen in a quarter or two. Meanwhile, the AI train keeps rolling forward, demanding more compute power every day. Something’s got to give – either the AI revolution slows down, or we see some serious innovation in data center design and construction. My bet? We’ll see both happening simultaneously.
