When you drive through Abilene, Texas on a clear afternoon, you may notice something that doesn’t quite fit the surroundings: a large construction project with buildings rising with the kind of urgency typically associated with wartime infrastructure, cranes moving against the vast sky. One of the biggest AI data center projects currently in progress worldwide is the Stargate facility, a partnership between OpenAI, Oracle, and SoftBank. It doesn’t appear to be what most people would consider valuable real estate. There are no shiny facades, no artwork in the lobby, and no corner offices overlooking the city. Rising out of West Texas as if the land itself had chosen to go digital, these massive, meticulously climate-controlled buildings were built to house servers and keep them cool.
The serious real estate money is increasingly going toward that image, which is practical, unglamorous, and almost brutalist in its disregard for aesthetics. By 2029, the global data center market is expected to grow from its estimated $342 billion in 2024 to $517 billion. 95% of the 92 major investors surveyed by CBRE in 2025 said they intended to increase their data center allocations, and 41% said they would invest at least $500 million in the industry that year.
| Topic | Data Centers as Premium Global Real Estate Asset Class |
|---|---|
| Global Market Value (2024) | Approximately $342 billion |
| Projected Market Value (2029) | ~$517 billion (CAGR of 8.6%) |
| Current Global Data Center Capacity | ~60 gigawatts of critical IT demand |
| Projected Capacity Needed by 2030 | 170–220 gigawatts (McKinsey estimate); unconstrained demand exceeds 250 GW |
| Goldman Sachs Power Forecast | Data center power demand up 50% by 2027; up 165% by 2030 |
| Hyperscale Facility Build Cost | ~$12 million per megawatt; modern facilities range 150–300 MW |
| AI-Centric Facility Cost (1 GW+) | Multi-billion-dollar investment required |
| Hyperscaler Capex (2024–2030) | ~$1.8 trillion needed (Boston Consulting Group estimate) |
| Global AI Data Center Spend (to 2029) | ~$3 trillion (investment bank estimates) |
| Investor Sentiment (CBRE Survey 2025) | 95% of major investors plan to increase data center allocations |
| Key US Locations | Northern Virginia, Silicon Valley, Dallas, Abilene (Texas — Stargate project) |
| Emerging Locations | Iowa, Wyoming, Wisconsin, Alabama — driven by power availability |
| Europe Investment (2024) | €8.4 billion — up 44% year-over-year |
| Key Players | OpenAI, Oracle, SoftBank (Stargate); CapitaLand, JLL, Gaw Capital, CBRE |
| Reference Website | CBRE Global Data Center Trends Report |
These figures depict a market that has quickly evolved from a specialized area of institutional real estate to something that fund managers and pension allocators now consider to be a core holding. The change is happening, and it’s happening more quickly than the majority of those in charge of it even predicted three years ago.
The CEO of CapitaLand Investment’s private funds and alternatives team, Kishore Moorjani, explained the change in terms that are worthwhile. At the Milken Institute Asia Summit in Singapore, he stated, “The world of real estate is changing from what I’d call the ‘visible’ to the ‘invisible,'”. He pointed out that the cloud is physically located in data centers, which is what is causing the value shift away from office towers and upscale hotels and toward structures that the majority of people will never see inside. It’s a subtle reinterpretation of what “location, location, location” really means when the tenants are AI workloads that run nonstop, day and night, rather than people commuting to desks.
Global data center power demand is expected to increase by 50% by 2027 and up to 165% by 2030, according to Goldman Sachs. The world will require between 170 and 220 gigawatts of critical IT capacity by the end of the decade, according to McKinsey’s proprietary demand modeling. If supply chains work together, this unrestricted demand could surpass 250 gigawatts. In practical terms, capacity must more than triple. There is currently no infrastructure in place to support that.
It must be constructed, financed, powered, and cooled—mostly from scratch—in a timeframe that puts pressure on every link in the supply chain. The most difficult limitation is power. If a data center can utilize the current power infrastructure, it can be finished in 18 to 30 months. The most valuable land for a data center isn’t always in a major city because new power generation projects frequently take longer than that, sometimes significantly longer. Wherever there is already electricity, that is.
If you have been following the real estate industry for a long time, you will find that this logic is changing the geography of real estate investment in ways that are truly bizarre. Although Silicon Valley, Dallas, and Northern Virginia continue to be major hubs, data center developers are increasingly banking land in states like Iowa, Wyoming, and Alabama.
These states are chosen not for their proximity to corporate headquarters or consumer populations, but rather for their availability of affordable power and the fiber networks required to transfer data. For both hyperscalers and real estate developers, “land banking”—purchasing desirable parcels five, seven, or ten years before a shovel goes into the ground—has become commonplace. These businesses are not engaging in heedless speculation. They are placing long-term wagers on locations where power can be produced or acquired within a reasonable amount of time, and they are patiently holding those positions.
Tension is beginning to appear on the financial side of all of this. The largest cloud providers construct and run hyperscale facilities, which are massive data centers that range in size from 150 to 300 megawatts. The cost of building a hyperscale facility is approximately $12 million per megawatt. Multi-billion-dollar investments are needed for each AI-centric facility that exceeds one gigawatt in capacity. According to Boston Consulting Group, in order to meet the demand for cloud computing and artificial intelligence, hyperscalers alone will need to deploy about $1.8 trillion between 2024 and 2030.
Conventional bank lending is not comfortably calibrated for capital requirements at that scale because it is based on risk frameworks intended for office parks and shopping centers. “Banks are certainly challenged around their data center exposures,” Moorjani said, “given the sheer volume and quantum of the build that’s going on.” Even though demand is far outpacing supply, the financing gap may become a real barrier to the market’s potential growth.
There is a sense that the real estate industry is undergoing an irreversible reorientation as institutional capital rotates toward this sector with the kind of conviction that was once reserved for office towers in the central business district. The same factors that drive demand for data centers—AI workloads, cloud computing, and the sheer amount of data generated daily by all connected devices—are structural rather than cyclical. Unlike traditional commercial property, they are not influenced by office occupancy rates, consumer sentiment, or interest rate cycles.
A tenant with a balance sheet that most landlords would find comforting typically signs a long-term lease with a major cloud provider or a government agency. When the traditional asset classes seem uncertain, institutional investors are searching for precisely that combination of qualities: steady income, rising demand, and protection from economic fluctuations. The story becomes complex when it comes to whether the build-out can keep up with market demands and whether the power grids underneath it can support the load. However, it appears that the direction of travel is fairly obvious.
