By Dr. Atif Ansar
Earlier this month, I had some of the most candid project delivery conversations that I’ve had in a while. The Foresight team spent the Monaco Grand Prix weekend with customers, partners, and industry peers, and people were talking openly about what is working and what is not. And a pattern kept surfacing.
Nearly every conversation circled back to the same challenge: this industry is running late. Projects that looked fine six months ago are now slipping. Milestones are moving. Explanations are being issued. And the explanations, almost universally, point outward: to power constraints, transformer lead times, interconnection queues, permitting delays, and more.
These are, of course, all real constraints, and all genuinely difficult. But I kept asking a different question: are these actually the causes? Or are they the places where problems become visible?
That distinction is worth sitting with.
The Wrong Diagnosis
The numbers tell an interesting story.
JLL's 2026 Global Data Center Outlook forecasts an infrastructure investment supercycle requiring up to $3 trillion by 2030, with nearly 100 GW of new capacity expected to come online between 2026 and 2030, effectively doubling global capacity. Demand is not the problem. Capital is not the problem. The industry has never had more opportunity in front of it.
Yet the same report notes that more than half of projects experienced delays in 2025, to the point where tenants can no longer assume delivery dates without meaningful schedule buffers.
CBRE's latest North America Data Center Trends report reveals another telling signal. Despite unprecedented demand, capacity under construction fell from 6,350 MW at the end of 2024 to 5,994 MW at the end of 2025, the first contraction since 2020. The visible explanation is familiar: permitting challenges, zoning hurdles, labor shortages, and power procurement delays.
But visible constraints are not necessarily root causes. Indeed, if we treat symptoms as causes, we inevitably end up solving the wrong problem.
The common thread running through all of these explanations is that they describe constraints. What they do not explain is why some teams navigate those same constraints successfully while others fall months behind.
Every major project today faces power uncertainty, long equipment lead times, permitting complexity, and labor pressure. Yet outcomes vary dramatically. That suggests the decisive factor is not the existence of constraints themselves, but how early teams understand their impact and how effectively they adapt their plans around them.
In other words, the question is not whether risk exists. The question is whether the schedule reflects reality before that reality arrives.
The Gap That Decides Outcomes
There is a distinction worth making carefully here, one that sits at the heart of what Foresight was built to address: the difference between schedule confidence and schedule plausibility.
Schedule confidence is the team's belief that the date will be met. It is what gets communicated in status updates, reported to boards, and shared with lenders. Most programs track it carefully and report it consistently.
Schedule plausibility is a different question entirely: given the dependency structure, the float distribution, and the actual track record of the counterparties involved, do the underlying conditions support the stated date? A schedule can carry high confidence and low plausibility simultaneously. When it does, the team is committed to a date that the conditions around it cannot yet support.
The gap between confidence and plausibility is where late-stage failures are born. Closing that gap requires four organizational disciplines: a live dependency map across organizational boundaries, not just internal milestones; an evidence base on how every material counterparty has performed against prior commitments; probabilistic forecasting that surfaces the range of outcomes rather than a single point estimate; and board-level reporting that presents plausibility alongside confidence, so that leadership is seeing the same picture the schedule is actually telling.
These are organizational habits before they are software features. The firms that are building them now will deliver materially more capacity, at materially better economics, than those that continue to manage by reported confidence alone.
The encouraging part is that this is not a theoretical problem. We now have the data, computational power, and organizational experience to measure plausibility directly rather than relying on confidence as a proxy.
That shift is exactly what Foresight was designed to enable.
Foresight Wins Best AI Innovation at the 2026 Datacloud Global Awards
That innovation was also recently recognized by Datacloud: a different way of seeing what is actually happening on a program. At Datacloud Global Congress in Cannes this month, Foresight was named the winner of the Best AI Innovation award. The Datacloud Awards bring together the global digital infrastructure community to recognize organizations delivering outstanding achievements across the sector, and this year's judging panel included some of the most senior figures in the industry.
The recognition matters to me for a specific reason. It is a validation of the problem Foresight was built to solve. As I said: for decades, organizations have accepted delays and overruns as inevitable. We believe they are predictable and preventable. That belief is what the platform is built on, and it is good to see the industry beginning to recognize it. Read more about it on our website.
Alongside the award, I joined fellow industry leaders on a panel exploring what it takes to deliver the next generation of digital infrastructure. The themes will be familiar to anyone reading this: AI demand is reshaping what project delivery requires, and the industry is only beginning to reckon with what that means for transparency, collaboration and risk management at scale.
What's Going On?
Foresight is sponsoring Advancing Data Center Construction Texas in July. It is one of the sector’s most focused forums for the people who are actually building these programs, and the conversations tend to be direct and substantive. If you are attending, I hope you will connect with the team.
We are also speaking at the event. Our session will address what the data actually shows on AI data center construction timelines and budgets, and which bottlenecks matter most as build complexity accelerates. We will cover where cost, scope, and schedule breakdowns concentrate in modern data center programs, and where execution risk originates. If those are questions on your mind, I think you will find it worth attending. More details are available here.
Closing Reflection
The F1 teams that won in Monaco did not react to the race. They had already modeled every scenario, mapped every contingency, and distributed every decision long before race day. Their advantage was built in the months before the cars arrived on the street circuit.
The delivery teams getting this right in data centers are beginning to operate the same way. They are building the systems that surface risk months before it becomes a problem, tracking what counterparties are actually doing rather than what they have announced, and asking whether their schedules are plausible rather than simply whether their teams believe in them.
The rest of the industry is still writing post-mortems.
As always, I would genuinely welcome your perspective on this. Leave a comment, push back on the argument, or reach out directly. The conversations that come out of these newsletters are often the most useful ones I have.
Atif