Oracle’s stock recently surged 36 percent in a single day after the announcement of a new deal with OpenAI—a spike that generated an extra $244 billion in market cap for the company. The move fueled increased speculation about a potential AI bubble brewing. Such high-flying stock prices recall the dot-com bubble, when the NASDAQ stock index lost more than 70 percent of its value, dropping from a high of 5,048 in March 2000 to a low of 1,139 in October 2002.
Urban Land: Do you see an AI bubble? How is commercial real estate likely to be affected if there is an AI bubble that bursts?
Michael Acton, head of research and strategy for North America, AEW Capital Management
Recent increases in the market capitalization of AI and AI-related companies does not necessarily indicate an “AI bubble,” but it does give us pause. Warren Buffet’s preferred measure of equity market risk—the ratio of market capitalization to GDP, is at an all-time high—much higher than the early 2000s tech crash.
Rapid propagation of technologies broadly referred to as AI is occurring at a demographic watershed, with population and labor force growth slowing across most advanced economies. Economic growth will increasingly depend on productivity, largely tied to labor substitution and enhancement. If this assumption ultimately proves incorrect (i.e., the AI bubble bursts), the impact on commercial property could be significant.
The most direct impact would be demand for data centers, one of the more active areas of property investment in recent years. There could, however, be offsetting positive demand from office tenants needing to increase hiring as hoped for productivity growth fizzles.
Beyond property specifics, a meaningful reversal of the AI tech stock boom could slash household net worth if trillions of dollars in equity market gains became losses; possibly disrupting household and business consumption and investment for years to come. For property investors, this could trigger a negative denominator effect with falling equity values forcing reductions in real estate holdings.
Ryan Severino, CFA, managing director, chief economist, and head of research at BGO
It is difficult to spot a bubble in advance. Typically, we only know we have been in one after the fact. But sometimes bubbles are important signals of long-term structural economic shifts, even if they cause short-term pain. Such was the case with the dot-com bubble of the late 1990s. And such could be the case again.
It is difficult to know if there is an AI bubble forming in public markets. But irrespective, AI will almost certainly change the economy in profound ways, similar to how the Internet changed the economy in profound ways. AI, including generative AI, will only continue to improve over time, like many technologies we use today—smartphones, the Internet, and personal computers. As that occurs, commercial real estate stands to benefit, most obviously data centers, but other property types, as well.
It is hard to prepare for a bubble when it is impossible to determine if we are even in one. But for commercial real estate, it is far better to prepare for a structurally different future than fixate on the potential for a bubble. Those who banked on the long-term prospects of the Internet benefited greatly, including those within commercial real estate. Those who fled permanently missed out on some of the best returns in history. Long-term AI-related prospects look transformative, and investors who focus on that should, ultimately, fare well once again.
Christopher Thornberg, Ph.D., founding partner, Beacon Economics
It is obvious how IT has dramatically altered and improved the lives of all Americans. Yet we nevertheless have to beware the tendency the industry has to oversell specific ideas, creating economic booms and busts along the way—something that the hyper-cyclical real estate industry always needs to be wary of.
The equity market frenzy around AI seems yet another example of this phenomenon, with rapidly rising asset prices and a sharp uptick in investment in both information technology equipment as well as in physical data centers. If the bubble pops, investments in both areas will slow sharply, and this can have an impact on overall growth—not unlike the mild tech recession at the start of the century when the promises of the “new economy” turned out to be in large part hyperbole.
But this specific tech boom is occurring within a broader context that makes the potential collapse of an AI bubble that much more worrisome. Even as AI investments have taken off, the Federal government has been in the midst of a spending and borrowing binge of $2 trillion per year. This surge in borrowing has had less of an impact on interest rates, in part because of a huge surge in portfolio investment into the U.S. economy by foreign investors who are chasing the outsized promised returns of AI.
When the bubble collapses, the U.S. will be forced to self-fund the Federal deficit, leading to a sharp rise in interest rates. This will, in turn, cause the carry cost of Federal debt to rise, making the deficit even worse. In other words, the AI bubble collapse could be the trigger to a much broader public debt crisis, and this is something every commercial real estate investor should rightly fear.
Timothy H. Savage, Ph.D., CRE, professor of practice at the NYU Schack Institute of Real Estate and inaugural faculty director of the CREFC Center for Real Estate Finance at New York University
It’s always challenging to predict an asset bubble until it actually appears because you don’t see the bubble growing. Once it reaches its peak, it either pops or it begins to lose energy and collapses in on itself relatively slowly.
At the university level, what I see with my students is that the genie is out of the bottle. Students are using various large language models on a regular basis, and there’s no going back. Although AI has limitations, such as gathering incorrect information or reproducing human bias, it can be a very useful tool. It’s the growing use of AI that takes us to the question of the AI bubble and potential impact on commercial real estate.
The manifestation of AI in commercial real estate is in data centers. Data centers are tremendously large, and they require huge amounts of power to drive the chips that run the large language models (LLMs). Where the potential for a bubble arises is in these very large data centers because the LLMs rely on a class of algorithms that, under current technology, we can’t parallelize well.
An analogy I like to use is the string of holiday lights. Traditionally, holiday lights were strung in a serial capacity. Energy moved along the wire, lit one bulb, and then another and another on down the line. If one bulb failed, portions of the entire strand would fail. As holiday lights evolved, we began to string them not in a serial fashion, but in a parallel fashion so that if one bulb went out, it didn’t affect anything else.
If we can get to that stage with these LLMs, then the footprint of the current data centers is far too large, and the data centers themselves may lose value. That might be where the true bubble lies, not in the AI itself, but in the underlying technology that drives AI.
Chad Littell, national director, capital markets analytics, CoStar Group
In thinking about the burgeoning AI sector and its potential impact on commercial real estate, it may be instructive to consider what happened during the collapse of the dot-com boom, a similar technological leap. To be sure, the office market today is not as frothy as it was in 2000. In the second quarter of 2000, vacancy rates in Boston and San Francisco were 2.9 percent and 1.3 percent. But by 2003, those figures reached 12.2 percent and 15.5 percent, with asking rents declining 31 percent in Boston and 56 percent in San Francisco. A full recovery took 15 years.
In contrast, vacancies in these markets today have been rising since before the pandemic and now stand at 14.5 percent and 23 percent as of the second quarter of 2025. Over that period, asking rents fell 2 percent in Boston and 32 percent in San Francisco. The market is already working through a multi-year correction, with less risk of a sharp collapse from being overextended.
The data center market presents a fundamentally different situation than it did in 2000. Between 1990 and 2000, fewer than 40 data centers over 200,000 square feet were built nationally, with only eight exceeding 500,000 square feet. Much of the data storage was co-located within traditional office sites, and the third-party data hosting market was still in its infancy.
Since 2020, 282 data centers have been delivered or are under construction, with 79 exceeding 500,000 square feet and some reaching as much as four million square feet. The largest exposures are in markets such as Washington, DC, Atlanta, and Richmond in the east; Omaha, Des Moines, and Chicago in the Midwest; and Phoenix, Dallas, and Albuquerque in the Southwest. The scale and distribution of these developments suggest that if an AI bubble were to burst, the impact could be spread across a broad set of geographies, rather than concentrated in a single segment or market.
John Chang, chief intelligence and analytics officer, Marcus & Millichap
Although companies operating in the Artificial Intelligence space have achieved lofty valuations, the risk of a bursting AI bubble severely impacting commercial real estate remains limited. In contrast to the dot-com era, AI is creating value and influencing business practices more quickly. The technology is being integrated into business applications across a wide array of mature industries, so AI may be less vulnerable to a bubble burst.
However, if the AI sector is in a bubble and it were to rupture, the impact on commercial real estate would likely be less traumatic than the bursting of the dot-com bubble. Today, an estimated 50 percent to 67 percent of technology employees work remotely, which contrasts sharply with the single-digit percentage in 2000. That lessens the risk a breached bubble would pose to the already battered office sector.
In addition, AI workers are also less geographically concentrated than their dot-com counterparts. In the early 2000s, the Bay Area was the epicenter of the dot-com workforce. Although AI companies still have a heavy presence in the Bay Area, the sector’s workforce is more dispersed. As a result, the impact of an AI bubble burst on commercial real estate would be diluted, spreading across many markets with a less substantial direct impact on any one of them.
So, while the popping of an AI bubble would certainly have significant financial market and economic implications, potentially even spurring a recession, its direct impact on the commercial real estate market would likely be limited.
Recent Snapshots:
