The construction of a massive wall along the border of the United States and Mexico is one of President Donald Trump’s central campaign promises. And it’s a promise he intends to keep.

Within days of taking the oath of office in January, Trump began laying the groundwork for the construction of a series of walls and fences that would span some 1,250 miles along the border. On Monday, the Department of Homeland Security issued a memo outlining its commitment to “begin planning, design, construction and maintenance of a wall” to deter and prevent illegal entry into the United States. The memo follows an executive order in which Trump called for the wall’s “immediate construction.”

But how immediately can Trump’s wall be built?

One of the latest estimates, from an internal Department of Homeland Security report obtained by Reuters, is that the wall will take three-and-a-half years to build. The agency is aiming to seal the border in three phases of construction of fences and walls, completing its work by the end of 2020, Reuters reported.

But that estimate is almost certainly too ambitious, and for a few reasons. First and foremost, there’s the fact that Congress still has to approve the bulk of the money for a project that is likely to cost tens of billions of dollars, according to several estimates.

Even if lawmakers approved that kind of cash this week, the wall almost certainly wouldn’t be complete by the end of Trump’s first term—or even a potential second term. The “iron law” of infrastructural megaprojects, according to a paper by Bent Flyvbjerg published in the Project Management Journal in 2014, is that they will go “over budget, over time, over and over again.”

This is obviously the case in projects where everything that can go wrong seemingly does (think: Boston’s infamous Big Dig highway project). But even for well-managed megaprojects, building major infrastructure always seems to take longer than estimates suggest. Sometimes that’s because a time estimate only pertains to the actual construction—not the time leading up to it, says Andrew Natsios, the director of the Scowcroft Institute of International Affairs at Texas A&M University.

Natsios is a public-finance expert who was hired to salvage the Big Dig project after catastrophic cost overruns. He also has a reputation for stinginess. (Evidence for his frugality, The New York Times reported in 2000, was displayed on a conference room in his office at the time: “a red sign with the simple word ‘No.’”) Natsios told me that he hasn’t studied the wall, specifically, but that it is likely to face the same sorts of challenges that any megaproject bumps up against.

“Some of the wall will have to be built on private property and the owners may challenge the federal government’s taking of the land,” he wrote in an email. “Politics extended the construction of the Big Dig in Boston for years because of innumerable disputes. That will likely be the case in the construction of the wall on the U.S. border with Mexico.”

That said, there are some aspects of the project that make it easier to stick to a timetable. The fact that the wall is to be built on “open land,” rather than in an urban area, is hugely helpful, Natsios told me. “Building on land where there are existing buildings, in urban areas, that takes much longer,” he said.

That’s not to say it will be easy. “The border presents huge topographical challenges to construction,” analysts at the research firm AllianceBernstein wrote in a report last year. “It runs through remote desert in Arizona, over rugged mountains in New Mexico and, for two thirds of its length, along rivers... These difficulties were illustrated during the construction of the existing fence which was beset by delays, surging costs and disputes with private land owners.”

Overrun models demonstrate that delays to major infrastructure projects often beget more delays. This is an easy concept to understand, but a difficult one to quantify. “The potential to lose productivity as work slips from its initial plan is generally understood but rarely considered explicitly in planning and estimating,” wrote Stephen Grey, an expert in uncertainty models, in a 2010 paper on risk management. “It raises the hypothesis that some projects lose control over their schedule because insufficient effort was devoted to understanding the effect of small deviations from the plan—and to preparing to prevent it from snowballing into a progressive decline.”

The bigger the project, the worse this snowball effect can be. And all over the world, infrastructure projects are ballooning in size. “Not only are megaprojects large,” wrote Flyvbjerg, who is the founding chair of Major Program Management at Oxford University’s Saïd Business School, in his 2014 paper, “they are constantly growing ever larger in a long historical trend with no end in sight.”

All the while, project managers are generally terrible at developing accurate scheduling estimates—whether because they’re deceptive or just irrationally optimistic. “People have a tendency to think what they want to think is true,” Erik Angner, a philosopher who studies rationality told The Wall Street Journal in 2010. “It’s conceivable that bidders are lying. But it’s also conceivable they managed to convince themselves they can do it as cheaply as they say they can.”

What this means for Trump’s border wall is uncertain. But it does underscore the likelihood that one of the president’s central campaign promises won’t be fulfilled by the end of his presidency.