Think about the last time you walked through a half-empty office building in a big U.S. city. Vacancy rates hovered high after remote work took hold, and landlords wondered if demand would ever return. Now picture that same space filling up fast, not with traditional tenants, but with firms racing to power the next wave of artificial intelligence. The expansion of AI data centers is quietly turning the U.S. real estate leasing market upside down, creating pockets of intense activity amid broader softness.
AI models need massive computing power, which means hyperscalers like those from big tech pour billions into data centers. These facilities guzzle electricity and space, pushing operators to lease aggressively in markets with reliable power grids. North America saw net absorption double to 1,668.5 megawatts over the past year, with vacancy rates dropping to a record low of 1.6% by mid-2025. Landlords in hubs like Northern Virginia, Atlanta, and Phoenix report operators snapping up capacity years ahead, often at premium rates due to supply constraints. Power availability has become the key factor, shifting focus from low-latency spots to anywhere scalable infrastructure exists.
This rush benefits real estate investment trusts focused on data centers, as long-term leases provide steady income. Operators secure sites for 15 to 19 years, locking in revenue while betting on AI’s growth. Emerging areas like Des Moines and Richmond draw interest for their untapped power and fiber lines, diversifying beyond traditional markets. Rental rates climb in tight spots, though construction costs and grid limits slow new builds.
Beyond data centers, AI firms lease office space for engineers and support teams, sparking recovery in tech-heavy cities. San Francisco, with 35% office vacancy, sees AI companies take 4.3 million square feet since 2022, including big deals like OpenAI’s 486,000-square-foot Mission Bay lease. Landlords note optimism, as startups grab flexible class B space and established players chase class A buildings. This trend spreads to New York, with 237 AI firms, and spots like Seattle and Boston, where professional services expand AI offerings.
Law firms and financial outfits pile in too, hunting Bay Area space to build their own AI tools. Vacancy eases as leases shorten for growth flexibility, yet tenants prioritize landlords open to future expansions. Office owners outside core tech zones feel less direct lift, but the overall vibe hints at broader stabilization.
Oracle Corporation (NYSE: ORCL) offers a clear case of this shift. The company revealed lease obligations for data centers and cloud capacity jumping 148% to $248 billion as of late November 2025, spanning 15 to 19 years. This covers $10 billion in immediate cloud deals, fueled by clients like OpenAI, which tapped Oracle for projects including a Texas site funded by partners. Oracle competes with Amazon, Microsoft, and Google by leasing from specialists like Crusoe, delaying its own upfront costs until facilities deliver.
Such moves highlight risks and rewards. Hyperscalers overbuild to avoid falling behind, but grid strains and financing tests resilience. Real estate players adapt by targeting power-rich zones and flexible terms.
Landlords who solve power puzzles stand to gain most, as AI keeps demanding more room to run. Data center deals reshape portfolios, while office revivals add upside in key cities. Operators and REITs navigate tight supply, but the leasing pipeline looks full for years. Tech’s infrastructure bet redefines U.S. property value in tangible ways.Â
