- There's a flood of data center projects on the way in 2025
- But large scale projects are often delayed, leading to huge cost penalties
- Less optimism in forecasting construciton timelines could help keep projects on track, Foresight has found
497. That’s the number of known data center projects in the global pipeline, according to Synergy Research Group. And there could be plenty more on the way given hyperscalers have announced plans to spend upwards of $315 billion expanding their cloud capacity in 2025 alone. More often than not, though, these projects are delayed, which means sorely needed compute power might not become available as fast as expected. But it doesn’t have to be this way.
“It’s very hard to deliver projects on time and on budget. For data centers, 9 out of 10 data centers have a delay and on average the delay is 34%,” Foresight Works co-founder and Oxford University teaching fellow Atif Ansar told Fierce. Practically speaking, that means a two-year data center project “flirts with becoming a three-year” project, he added.
This isn’t a new problem. Ansar studied mega project dynamics for his Ph.D and founded Foresight with Igor Shifrin back in 2018 to address project management issues that were found to be pervasive across a wide range of verticals. Data centers currently account for around 40% of Foresight’s business, with customers including the likes of Iron Mountain, Compass Datacenters and Airtrunk.
Ansar told Fierce that while Foresight’s software can be used across industries, the company was able to hone its models on privileged information about data center construction projects by partnering with the private equity companies financing the builds. The biggest takeaway? Optimism is a schedule killer.
“At a macro level, the key risk is not power or supply chain,” he explained. “The root cause, according to our research, are our human biases toward poor estimation.”
That’s not to say power constraints are not an issue – they are and can take anywhere from two to six years to resolve, according to CBRE. But, Ansar said, power and supply chain lead times are known quantities. Ignoring these to forecast a faster build time is folly.
Ansar noted that given the interconnected nature of data center projects, delays in one area can have a compounding domino effect on the overall project timeline. And that’s a big problem.
“Delays are very bad,” Ansar said, noting tenants can walk out and developers can be slapped with SLA penalties and face both inflationary and standing army costs. “A three-month delay can basically wipe out your net present value on a data center. So, you can be operating the data center for 10 years, but from an economic and return perspective you may never meet your investors’ initial forecast,” he added.
What, then, is the answer? Better planning. No, really.
Ansar said the industry is in such a rush to action, that companies oftentimes don’t spend enough time on upfront planning. The fix is to “think slow, act fast,” he said. The best part of this approach is that since data center projects tend to be similar, builders can do the hard thinking once and create a prototype schedule that can be replicated on future projects.
But that’s just phase one. Phase two is executing on those well-thought out plans, something that isn’t exactly a strength for the industry today because companies tend to delegate too much responsibility to the construction teams on-site. Corporate teams, Ansar said, should stay closely connected to projects and review progress and timelines on a weekly basis so there are no last minute schedule surprises. And lower level employees should be trained to think about projects like the C-suite would.
“It’s not about trust, it’s about accountability and making sure the project people have the support they need,” he said.
Those data center builders who don’t clean up their act risk dire consequences, Ansar warned. “Those who keep with smoke and mirrors will flirt with bankruptcy and that will hurt the numbers people are expecting.”