By Dr. Atif Ansar

I was recently on the Data Center Go-to-Market Podcast with Joshua Feinberg, and our conversation turned to how more businesses die of gluttony than starvation. Or, as I put it, more die of suicide than murder.

In a boom, the instinct is to watch your competitors. To protect your pipeline. To worry about whether demand will hold. Those are the wrong things to worry about. The real danger is what your organization is doing to itself.

The Numbers Don’t Lie

Nine out of ten major data center projects are completed late. Not in struggling markets with inexperienced sponsors. In mature markets, with well-capitalized developers, backed by committed tenant demand.

Consider the scale of what’s at stake. Hyperscale capital spending on data centers has risen more than fourfold since 2020, reaching an estimated $328 billion in 2025. Demand is contracted before ground breaks. Vacancy rates in the U.S. sit below 2 percent. By almost every measure, this industry has never had more going for it.

In December’s newsletter, I wrote about what delay actually costs: for a 60-megawatt facility, each month of slippage destroys approximately $14.2 million in value through lost revenue, cost overruns, and penalties. IRR falls from around 17 percent if delivered on time to 8.8 percent after six months, often breaching the thresholds that justified the investment in the first place.

Oracle provides the freshest illustration of what this looks like in practice. When reports emerged late in 2025 that data center projects linked to OpenAI had slipped by roughly a year, Oracle’s share price dropped sharply in a single session, wiping several percentage points of market value before a partial recovery. Oracle disputed the reports. But the episode was instructive regardless of the specifics: even the perception of an execution slip was enough to move a large-cap stock materially. Demand was never in question. What unsettled investors was the possibility that timelines had drifted.

Markets are not being irrational. They have correctly identified that in a capital-intensive, high-velocity build environment, delivery performance is a first-order financial variable. When it slips, value is destroyed. Not deferred. Destroyed.

This is the gluttony problem. The industry has more capital, more demand, and more contracted revenue than at any previous point in its history. And it is still finding ways to leave that value on the table.

Three Ways the Industry Is Doing It to Itself

The waste is not random. It follows patterns. In a paper I’ve been working on, I’ve identified three persistent habits of thinking that recur across organizations and markets. They are not difficult to recognize; they are just difficult to admit.

The first is underestimating what you are building. There is a widely shared assumption, rarely stated aloud, that data centers are essentially warehouses with sophisticated fit-outs. They are not. A modern AI facility manages heat rejection at rack densities exceeding 100 kilowatts, operates tightly coupled electrical and mechanical systems, and requires commissioning sequences of considerable complexity. In terms of systems integration risk and execution discipline, these buildings are closer to oil refineries than to real estate assets. When you plan for a warehouse and build a refinery, you get delays. Every time.

The second is assuming delivery can be managed at arm’s length. In previous cycles, construction was an operational detail, delegated to heads of delivery who reported upward when things went wrong. That model no longer fits the environment. When a single powered shell delay can cut $200 million from a company’s full-year revenue guidance, as we saw with CoreWeave, schedule performance is not a site-level concern. It is a board-level financial variable. Organizations that have not restructured their governance to reflect this are running a known risk and calling it normal.

The third, and the most insidious, is the belief that experience is the same as accuracy. The data center leader who has delivered twenty projects is not more likely to forecast correctly than the one who has delivered five. In many cases they are less likely to, because accumulated experience can harden into exactly the optimism bias that causes the next overrun. What looks like pattern recognition is often pattern-matching to a past that no longer applies. The projects getting built today, at these densities, at this speed, in this regulatory environment, have no true historical precedent. Treating them as routine is itself a form of risk.

One Question Worth Sitting With

None of these three habits are the result of negligence. They are the product of success. Organizations that have built well, raised capital, and scaled fast develop assumptions that served them in an earlier phase of the market. The problem is that those assumptions have a shelf life, and in this environment, that shelf life is shorter than most people realize.

So here is the question I would encourage you to sit with honestly: which of these three habits is currently your most expensive blind spot?

Not in general. On your most critical active program, right now.

What’s Going On

I also contributed a chapter to Greener Data, published this month. The argument is one that I think will surprise people: predictive delivery is one of the most powerful and overlooked sustainability strategies available to this industry. Not because AI tools reduce emissions directly, but because construction delay is itself a significant source of waste, both financial and environmental. A data center built right the first time, on schedule, is a more sustainable one. I’m happy to share the chapter if it is of interest.

Next month, I want to go deeper on the question of why execution breaks down so consistently, even in organizations that know better and have the tools to do better. The answer involves something about how leadership capability is built and deployed in complex programs, and it is not what most people in this industry expect. I’m looking forward to sharing it.

In the meantime, I would genuinely welcome your perspective. Which of the three habits resonates most? Where are you seeing it play out? Leave me a comment or reach out directly.

– Atif