By Dr. Atif Ansar

The AI infrastructure boom is the largest coordinated capital buildout in modern history. And most of it is being governed with a delivery model that was never designed for this scale.

$562 billion is being committed to AI infrastructure this year. Nine out of ten projects will be late.

Sit with that for a moment. Not nine out of ten underfunded projects in emerging markets. Nine out of ten large data center builds — in mature markets, with experienced developers, with more committed capital than this industry has ever seen. Amazon alone is spending $200 billion. Vacancy rates are below two percent. Demand has never been stronger.

This is not a capital problem. It is not a demand problem. It is a visibility problem. And visibility problems only get solved when the right people decide they are worth solving.

A delivery model built for a different era

The lesson going into 2026 is stark: AI data centers are not missing delivery targets because developers lack talent, capital, or capability. They are missing because the scale and velocity of the AI buildout have outpaced the industry's tools for managing time.

As I wrote in an article for The Fast Mode last month, every data center company is trying to scale at unprecedented speed with a delivery model inherited from an era when the stakes were lower, the equipment was simpler, and the market could absorb a slip. That era is over. Time-to-market is a competitive differentiator and a foundational condition. Firms that cannot convert capital into operational capacity on schedule will not compound scale, regardless of how much they raise or how strong their tenant demand is.

Every month of delay erases revenue that will never be recovered. It does not shift to next quarter. It vanishes.

Why delivery risk stays hidden — until it's too late

Most organizations are still governing delivery with models built for a fraction of today's scale. Scheduling data sits siloed from procurement and supply chain ad there is no single view of what is actually happening across the program. Scope, time, and cost run out of sync, each discipline reporting independently without a single integrated picture. Optimism bias keeps dashboards green until it is too late to act. Contractor and owner disconnects mask emerging risk. Static tools buckle under the complexity of today's program scale. And by the time variance surfaces upward through lagging indicators, the intervention window has already narrowed.

The result is a system that suppresses early warning signals until they become unavoidable. By the time a delay surfaces in a leadership report, recovery options are near to none. What looked like a manageable two-week variance in week four, has compounded into a quarter's worth of missed revenue by week twelve.

What surprises people is that this pattern does not only apply to inexperienced teams. I was recently asked on the Data Center Go-To-Market Podcast why even the most experienced operators keep getting this wrong. I answered with a common illusion well researched in psychology.

The Müller-Lyer illusion presents two lines of identical length. One has arrows pointing inward at each end; the other has arrows pointing outward. Despite being the same length, they look different — one appears longer than the other. Every person who has seen this illusion before knows the lines are identical. And yet they still look different. Even trained experts, even people who have encountered the illusion hundreds of times, cannot override the perception through knowledge alone. The only way to know the truth is to take out a ruler and measure.

Optimism bias in project delivery works exactly the same way. Working in this industry for 10, 20, or 30 years does not exempt a leadership team from it — in some cases, experience reinforces it. The confidence earned from past builds becomes the very thing that prevents a clear-eyed read of the current one. The schedule, filtered through multiple layers of summary, tells leadership what they expect to hear. Until suddenly, it does not.

Experience building data centers tells you what usually happens, the data tells you what is happening now.

The CEO's Blind Spot

Too often when I start speaking to a CEO about their project schedule, I get a version of the same response: 'We have a team of planners and schedulers who manage that.' Or: 'Our project controls team handles delivery.' It is a confident answer, and it comes from a genuine belief that delivery can be safely delegated to the frontline — with minimal accountability required from C-level leadership or investors.

In the current era of AI infrastructure projects, the organizations pushing ahead have rejected that belief. They do not see the schedule as a project document. They see it as a means to impact the balance sheet. Because when teams see time in dollars; IRR, valuation, lease commitments — behavior changes, and delivery governance becomes a board-level and investor-level responsibility.

Last month I wrote about the Three Lines of Defense model for project delivery. The third line is senior leadership — approving critical milestone shifts, knowing the baseline, and holding the schedule accountable to business goals. The leadership teams embracing that role are the ones who will win this AI race. Because predictable project delivery starts with one belief: that time is money, and managing time, and thus the delivery schedule, is a C-suite conversation.

The $562 billion being committed this year assumes the industry can deliver at this scale. Most of the systems governing that delivery were not built for it. Changing that is a technology decision, but most importantly it is a leadership one.

I'd welcome your perspective: how does your organization currently translate schedule performance into financial terms — and who owns that conversation at the top? Leave a comment or reach out directly.

— Atif