The underlying logic of commercial real estate investment in the United States is being reshaped by the artificial intelligence (AI) boom. While this trend brings enormous potential returns for investors, it also exposes their portfolios to unprecedented levels of risk in a single industry.
According to a report by The Wall Street Journal on the 23rd, citing data from the U.S. Census Bureau, expenditures on data center construction are expected to surpass those for office buildings as early as next year. Driven by demand for AI, the return on investment for data centers has outperformed the market, with data from the National Association of Real Estate Investment Trusts indicating that the return on such assets reached 11.2% last year, outperforming all other real estate sectors except for mobile homes.
This construction frenzy is reflected in astonishing capital investments. Real estate services firm JLL predicts that the scale of new data centers in North America could reach $1 trillion between 2025 and 2030. Due to the urgent need to expand computing power, tech giants are changing their strategies from building their own facilities to leasing, with Oracle alone currently having future leasing commitments totaling $248 billion.
However, this shift has raised market concerns about the &#34;AI bubble&#34; and its impact on physical assets. Unlike the relatively stable performance of commercial real estate during the bursting of the internet bubble in 2000, the current real estate industry is more tightly coupled with the tech sector than ever before. As investors bet that AI technology will generate trillions of dollars in new revenue, real estate funds with increased risk exposure may face severe challenges if demand retracts or construction deliveries fall short of expectations.
<h2>Trillion-Dollar Infrastructure Boom and the Decline of Office Buildings</h2>
The surge in data center construction expenditures comes at a time when U.S. cities are struggling to absorb decades of excess office space. Currently, new office building projects are nearly at a standstill, with vacancy rates hovering at historical highs. In contrast, the thirst for computing power makes the expansion trend of data centers difficult to reverse in the short term.
To accelerate the construction of AI networks, major cloud computing companies like Meta, Amazon, and Oracle are increasingly leasing data center space from owners rather than solely relying on building their own. According to McKinsey, 40% of the capacity of U.S. hyperscale data centers this year is obtained through leasing, up from 35% in 2023.
This tilt towards data centers is not just a change in the flow of funds but a fundamental shift in the nature of commercial real estate investment. Traditionally, office buildings, apartments, and shopping centers were viewed as diversified and stable assets capable of hedging against fluctuations in the tech sector. During the tech stock sell-off from 2000 to 2002, despite the Nasdaq index plummeting nearly 80%, the value of commercial real estate remained relatively stable or only slightly declined. Today, that decoupling relationship no longer exists.
<h2>Capital Influx and Big Tech Bets</h2>
Major asset management firms' infrastructure and real estate divisions, such as Blackstone and Brookfield, have significantly increased their exposure to data center assets in recent years. According to data from the National Association of Real Estate Investment Trusts, publicly traded real estate companies are increasing their investments in data centers by 15% in 2024 while reducing investments in traditional sectors like office buildings and apartments. Private capital is also following suit, with a survey by CBRE of 92 major investors (including private equity and pension funds) showing that 95% of respondents plan to increase their investments in data centers
In this round of expansion, Oracle has been particularly aggressive. According to recent public documents, the company has $248 billion in future lease commitments off its balance sheet. This massive capital expenditure plan also caused market turbulence earlier this month, leading to a significant drop in Oracle's stock price after announcing an increase in AI infrastructure spending.
Green Street analyst David Guarino pointed out that despite the enormous amount, the ultra-large enterprises have balance sheets sufficient to support these transactions. To hedge risks, owners and lenders have also designed special structures for the transactions. For example, Meta's upcoming Hyperion data center in Louisiana is being developed by a partnership in which Blue Owl Capital holds a majority stake. Its 20-year lease is divided into five four-year terms, and if Meta wishes to terminate the contract after the first lease term, it must repay the outstanding debt at that time and pay additional cash to Blue Owl.
<h2>New Risks from Single Tenants and Stringent Terms</h2>
The long-standing appeal of commercial real estate lies in its &#34;diversification&#34; attribute, as office buildings are typically leased to different companies across multiple industries. However, as the primary growth driver for data centers comes from artificial intelligence, this asset class is becoming highly dependent on a single type of tenant, whose business models have not yet fully achieved profitability.
In addition to market risks, the risk of technical defaults due to operational disruptions cannot be overlooked. Sharon Haran, Chief Business Officer of Parametrix Insurance, pointed out that many lease agreements contain clauses that protect tech tenants. For example, if the monthly power outage exceeds a certain limit, owners may face penalties equivalent to several months' rent; long-term power, cooling, or connectivity issues may even grant tenants the right to cancel the entire lease. These technical indicators could lead to significant financial losses and are also the reason many conservative investors have yet to enter this industry.
Despite strong demand, the supply side is facing multiple challenges. The data center industry is struggling with a sluggish labor market, which could lead to a shortage of construction crews. At the same time, the supply chain for building materials is threatened by tariffs, and power acquisition remains a huge question mark in many regions.
These operational obstacles could cause developers to be hindered during construction, triggering delay clauses that lead to lease termination. Wes Cummins, CEO of data center operator Applied Digital, warned that many projects are expected to encounter issues during the actual delivery phase. He noted:
<blockquote>
&#34;If you deliver too late, it usually triggers cancellation clauses.&#34;
</blockquote>
This means that in this trillion-dollar infrastructure gamble, execution will be the key factor determining whether investors &#34;make a fortune&#34; or &#34;catch flying knives.&#34;

AMZN

ORCL

The AI boom is reshaping the U.S. commercial real estate market, with data center construction spending expected to surpass office buildings next year, attracting over a trillion dollars in funding from Blackstone, Brookfield, and tech giants. With reliance on single tenants, stringent operating terms, and challenges in power supply, investors' risk exposure has reached historic peaks. If AI demand cools or delivery is delayed, this gamble will face severe tests

- The surge in U.S. commercial real estate investment is being reshaped by the AI boom, exposing portfolios to unprecedented single-industry risks.  
- Data center construction spending is projected to surpass office space investment next year, driven by AI demand, with returns leading the market at 11.2%.  
- This shift sees tech giants moving from building data centers to leasing, heightening concerns about an "AI bubble" and the intertwined risks between real estate and tech sectors.

Wallstreetcn

Amazon

Oracle

Make big money or catch flying knives? U.S. AI infrastructure spending is about to surpass office buildings for the first time, with Oracle alone signing a massive leasing deal worth $248 billion