Real Estate Poised to Profit in the AI Era

Forecasting who will prosper and who will not becomes a daunting task when a new industrial era is taking shape. We may be at such a pivotal economic moment today with AI, which naturally prompts the question of where the winners will emerge. 

Will the current AI frontrunners—OpenAI and Anthropic—be the leaders? Will incumbent tech giants like Meta and Amazon, who have put substantial bets on AI, come out on top? Or will software firms such as Microsoft, whose momentum seems especially AI-boosted, claim the crown? Beyond firms, what kinds of workers will gain the most?

History shows there is enormous uncertainty surrounding these questions. Yet I cling to one overarching insight with relative confidence: land will be a winner. And so will the construction workers who transform land into homes.

But before we turn to land, let’s begin with two historical illustrations that underscore how difficult this exercise can be. 

Picture yourself in 1900 with the foresight to recognize not merely the promise of the early automobile sector but to foresee Henry Ford as the innovator who would introduce mass production, catapulting us from thousands of vehicles a year to millions.

Yet, you would still end up on the losing side financially. 

In 1900, Henry Ford held roles as superintendent, inventor, and shareholder at the Detroit Automobile Company. To an investor with impeccable foresight, everything might have appeared perfectly aligned.

Nevertheless, the company ceased operations in January 1901. Ford Motor Company—the true breakout star of the early auto era—was founded two years later. Choosing the actual winners is difficult even when you know the right innovator and the right technology, and you’re merely two years from the opportune moment.

The Detroit Automobile Company never gained substantial momentum, producing only a handful of vehicles. Yet even ventures that take an early lead can collapse when technological upheaval arrives. The arc of the computer era makes this point clear. 

In the 1960s, mainframe computers occupied vast rooms and dominated the landscape. The sector was steered by IBM and, later, a cadre of smaller makers colloquially labeled “IBM and the seven dwarfs.” 

As Clay Christensen argued in his foundational management treatise The Innovator’s Dilemma, the ascent of so-called minicomputers disrupted these incumbents. Neither IBM nor any of the seven dwarfs became leaders in the minicomputer market. Instead, insurgents like Digital Equipment Corporation (DEC) displaced massive batch-processing mainframes with smaller, more affordable, interactive general-purpose machines such as the PDP-1. The machine shown below exemplified the kind that seeded hacker culture and hosted some of the earliest video games. The mainframes and their makers were left behind.

Yet DEC and other minicomputer makers would themselves be overtaken by the personal computer era, driven by Apple, Tandy, and the revived heavyweight IBM. Later, the center of gravity shifted to Hewlett-Packard and Compaq, the latter absorbing a diminished DEC in 1998. 

You could have profited at multiple points along the way from these firms, but disruption and the disappearance of industry leaders have been a persistent theme. Not only were individual companies uprooted by the computer revolution, but successive waves of computing technologies kept changing the center of gravity. The mainframe yielded to the minicomputer, the minicomputer to the personal computer. And today the personal computer arguably faces disruption from smartphones and tablets. 

In an industry characterized by rapid technological progress, it can be hard to know whom to back and risky to cling to incumbents. 

If the optimists are right, we are stepping into a period of rapid innovation akin to earlier upheavals. There are already early frontrunners: OpenAI and Anthropic have reached lofty valuations, and AI-linked equities have helped drive the S&P 500 to more than double in recent years. 

The market seems to be signaling something, but it remains unclear who will ultimately prevail. Will today’s leaders flame out like the Detroit Automobile Company or the seven dwarfs of early computing? Or, perhaps more optimistically, will they lead for a time before fading like DEC did?

The technology itself faces its own disruption questions. Will AI prove immensely valuable, or, like mainframes, will it be superseded by smaller, cheaper iterations? Perhaps these models could even be offshored in the long run.

Beyond the usual difficulty of predicting winners in eras of change, a particular AI challenge is distinguishing empowered outcomes from endangered ones in advance. 

At the job level, for instance, researchers struggle to determine which roles will be replaced by AI and which will merely be transformed by it. As economist Daniel Rock, co-author of one of the most-cited AI-exposure studies, notes, “Mapping capabilities to tasks that may change is easier than forecasting the precise changes themselves.”

This uncertainty extends to firms and industries too. Consider software-as-a-service (SaaS). Companies in this space, such as Salesforce and Workday, have faced a selloff dubbed the SaaSpocalypse due to fears that the ease of “vibecoding” software could reduce demand for experts. Salesforce CEO Marc Benioff counters that the opportunity is greater than ever and that AI will not replace their software but rather enhance it. Whether history vindicates him or not, some observers believe software companies remain under pressure.

Industries, firms, and occupations deeply affected by technology appear to be contending with a blend of uncertainty and opportunity. If we are searching for the safer winners, where should we look?

To identify winners we can feel confident about, we must outline plausible ways in which AI will reshape the macroeconomic landscape in a manner that holds up across various eventualities. We need shifts that endure across multiple possible futures. 

What can we say with some degree of certainty? Put aside the scenario where AI fails to deliver. If it carries economic significance, I think we can be fairly sure that average incomes will rise faster than in the past. Here are several reasons why. 

First, even if AI triggers substantial job displacement—though I doubt it will do so to the extent feared—this alone won’t be enough to stall mean income growth. 

That’s because even with widespread disruption affecting many workers, gains will accrue to those at the top, lifting mean income overall even if the bottom half benefits less. 

I would not say the same about median income growth, which reflects the experience of the typical worker. If substantial job losses occur, median wages could lag productivity gains as advantage concentrates among high earners. Yet mean income could still rise even if benefits are skewed toward the upper tier and capital owners. 

Let me be explicit: that would not necessarily be good news. It would be preferable if AI reduced inequality and elevated median wages as well. 

Second, looking to history shows that mean income in the United States has historically climbed robustly even during stretches when median wage growth stagnates. This is often treated as a problem, and rightfully so. Yet in the U.S. economy, GDP growth has consistently translated into higher mean incomes. 

As far as we can tell, AI is likely to push mean income higher.

If you’ve followed the logic so far, we can be reasonably confident about two things in an AI-haunted economy: 

So what will prosper under these twin forces? My answer is land. We can’t conjure more land with AI, and we can’t substitute land with AI. Yet demand for land rises as mean income climbs. 

This aligns with a thoughtful recent essay by Alex Imas, who argued that to determine what will be in demand in an economy rich in AI, we must identify what will be scarce: 

Imas’s piece focused on workers, but the same framework can be applied to other factors of production. Undoubtedly, land remains a scarce resource. In fact, he notes in a footnote that land itself may absorb a sizable share of income in the future.

Common sense suggests that rising mean incomes lift land values. This sits comfortably with the scarcity framework outlined by Imas. If you want more evidence, here are a few data points. 

At the very top of the income distribution, demand for land appears to rise quite noticeably. A telling illustration is the late billionaire Ted Turner, who reportedly could traverse his vast holdings without leaving his property—though the tale isn’t exact, he did accumulate about 2 million acres by the time of his death in May. 

Even that record doesn’t top the registries. Stan Kroenke owns 2.7 million acres—more than the entire state of Delaware. Forbes puts Kroenke’s fortune at about $24.3 billion. 

That’s the most extreme example, of course. But even the 100th largest landholder in the United States—the Irwin family—controls roughly 170,000 acres, larger than the city of Chicago. 

For the very wealthy, land accumulation can accumulate rapidly. How many AI-rich individuals will there be? Likely more than a few. Yet the link between income growth and land demand extends beyond a handful of millionaires. 

The chart below, drawn from a study by Schuyler Louie, John Mondragon, and Johannes Wieland, compares mean per-capita income growth with metro-area house-price growth from 2000 to 2020. I don’t endorse every result in their sometimes controversial paper, but the takeaway—that income growth helps push house prices upward—seems fairly clear. If house prices rise, land prices are very likely to rise as well. 

Microeconometric estimates of land demand are relatively scarce, but they line up with this broader trend. Gyourko and Voith found land prices rise about 1.5 percent for every 1 percent rise in income, which would render land a “luxury good.” That study, however, focused on a single Pennsylvania county. A broader analysis by Glaeser, Kahn, and Rappaport indicates land prices likely increase by roughly 0.25–0.5 percent for each 1 percent rise in income, suggesting land behaves more like a “normal good,” where higher incomes spur greater demand. 

Thus, in the AI era, land demand isn’t limited to a wealthy minority acquiring more property. Wider income growth—even if concentrated among the upper half—will fuel land demand. 

So when mean income climbs, I’m confident land demand will rise as well. AI is unlikely to disrupt this trend. 

If the narrative thus far reads like a dystopian forecast of AI-driven inequality, remember I’m not predicting benefits will accrue only to a few; I’m simply saying land demand remains robust even in that scenario. 

Still, the outline I’ve sketched hints at two policy moves that could greatly improve the outcome. Two steps are all that’s needed. 

Guarantee broad wage growth with a wage subsidy.

We can be reasonably sure that mean income will rise, but some workers may miss out on the upside. To promote broader prosperity, we can convert substantial mean income growth into meaningful median wage growth through a wage subsidy. (For a deeper look, see my longer piece on how such a subsidy would function.)

Ensure housing demand translates into homes.

Once we anchor strong wage growth, there should be demand for land as well as for housing in general. The next policy push is to pursue zoning reform at federal, state, and local levels. If we can carry forward the YIMBY momentum that has been building in politics in recent years, housing demand will express itself as more housing rather than simply higher prices. 

If these two policy challenges are tackled—or if AI merely delivers sturdy median wage growth—there remains one sector likely to thrive in an AI-enabled economy: construction. 

Construction stands out as one of the most productivity-resistant industries. Studies indicate productivity in construction has trended downward over time. That is usually bad news, but here it implies AI is unlikely to dramatically shrink the number of workers needed per home. After all, if broad computerization has failed to lift productivity in this field, smarter systems are unlikely to magically succeed. Even with any productivity gains, the sector will still demand substantial human labor for the foreseeable future. 

The future of building isn’t limited to housing. In this discussion I’ve focused on how mean income growth will reliably feed demand for land and housing. But let’s zoom out further. If AI simply fuels overall GDP growth, it opens up opportunities to construct all kinds of real-world projects. Through vigorous private demand, redistribution, or direct government support, there is no shortage of things we can build if there is money to fund them. 

Build more houses, more bridges, more roads, more public spaces, and more entertainment venues. 

Long before his death, Walt Disney envisioned Epcot as a futuristic city. We could indeed bring that city to life. We could realize California Forever. We could conceive of a dozen new cities and a hundred new theme parks. 

AI cannot replace dining out, but rising incomes fuel demand for meals away from home. So we can open additional restaurants. 

We could even plant forests if we’re prepared to fund it. 

The drive to build things in the physical world has three useful traits in an AI-powered future. Demand rises with income, it requires a lot of workers, and it offers a wide spectrum of possibilities. Land may emerge as the biggest winner from AI, but anyone involved in producing or consuming tangible goods can benefit as well. 

Video games can be protected by copyrights, which grants them protection for roughly 70 to 120 years. The first copyrights on video games appeared in the late 1970s and became widely asserted and litigated in the 1980s. That means it will be at least 2050 before the earliest video game source code enters the public domain. Those IP protections pose a serious hurdle for historians and archivists attempting to preserve classic games and provide access to them. 

Hardware patents, by contrast, expire after 20 years. This introduces ongoing competition and innovation in hardware. For example, Palmer Luckey’s latest venture ModRetro released a striking Nintendo 64–compatible console, the M64, with numerous features beyond the original. 

We have unleashed the market for classic video game hardware, but not for software. The outcomes are visible in real time. 

There is broad support for free trade nowadays.