Nvidia’s recently announced deal to invest $100 billion in ChatGPT parent OpenAI is the latest in a seemingly never-ending run of ambitious artificial intelligence projects, a euphoric climate that may mask underlying risks in the development of high-demand data storage real estate.
The latest deal, in which Nvidia and OpenAI are expected to build new data centers using Nvidia’s chips, is expected to use 10 gigawatts of power, enough to power about 7.5 million homes simultaneously.
Projects such as the Nvidia-OpenAI venture are so ambitious they dwarf data storage facilities built just a few years ago, a sign of the growing presence of AI — and the continued need for cloud data storage.
Ever-larger investments by the likes of Google and Meta demonstrate a bullish outlook for the sector because AI promises a transformative future with unprecedented leaps in automation, personalization and productivity. But some analysts are sounding the alarm about a complex environment that could conspire to cause some data storage projects to go awry.
Here are some factors that could upend a project:
Power
Rising demand for energy is expected to drive the largest expansion of power capacity in history over the next five years, fueled partly by the data storage boom, according to a Boston Consulting Group report on data centers.
Demand for data centers is expected to account for up to 60% of total power load growth between now and 2030, outpacing the expansion in sectors such as transportation and electrification, the report said.
The need for massive amounts of new power is a “critical bottleneck” that creates a mismatch between the typical two- to three-year development cycle for a greenfield data center and the typical eight-year period to make power-system upgrades to a site, according to BCG.
Tech giants are lining up deals directly with power providers as they launch huge data storage developments, such as a Meta deal with Entergy Louisiana to power a 4 million-square-foot data center in northeast Louisiana.
Many new projects will be powered by nuclear energy.
That includes an agreement Microsoft struck with energy provider Constellation to restart a Pennsylvania reactor that has sat idle since 2019 to provide 835 megawatts of new carbon-free power for Microsoft’s data centers. That reactor is at Three Mile Island, next to the reactor that sustained a partial meltdown in 1979.
Planning time
Power backlogs and other shortages, such as chips and various technology components, are among factors that have pushed overall North American data storage vacancy to an unprecedented low of 2.3%, despite inventory reaching a record 15.5 gigawatts, according a recent JLL report.
Low vacancy and high demand could lead to as much as $1 trillion in new data centers built on the continent by 2030, the report said.
Today’s market conditions require tenants to act 18 to 24 months, or even further, in advance to secure storage capacity, according to JLL. In the past, that time frame was six to 12 months, giving tenants a better chance to gauge market conditions before committing to space.
Long lead times on these developments are especially challenging because of rapid changes happening within AI, including the emergence of Chinese upstart DeepSeek early this year, which signaled some AI projects could be carried out far more inexpensively than initially thought and left some industry professionals questioning how the cost of AI projects might change in the years to come.
“The AI and high-performance computing market is such in its infancy right now, we can’t even project beyond two or three years out,” Michael Rareshide, a partner at Site Selection Group who specializes in data center site selection, told CoStar News in January.
Potential bubble
AI projects continue to proliferate amid an unclear picture of the long-term economy. That could mean an investment bubble that could burst, analysts say.
Goldman Sachs research strategist Ryan Hammond recently wrote of an “inevitable slowdown” in capital expenditures by major tech players such as Amazon, Microsoft and Oracle.
In a research note, Hammond wrote that an “extreme scenario in which hyperscalers cut capex [capital expenditure] back to 2020 levels” could lead to a loss of 15% to 20% in the S&P 500, Fortune reported earlier this month.
Some projects with massive price tags include Stargate, a $500 billion investment in data centers and other infrastructure by OpenAI, Oracle and SoftBank in Abilene, Texas. Developers of that project recently landed a $7.1 billion loan for the second phase of construction.
Sticker shock
For companies developing or leasing data centers to dive into AI initiatives, costs are eye-opening. Even longtime tenants with expiring leases are experiencing sticker shock, with market rents up more than 50% over the past five years, according to JLL.
Despite high costs to enter the AI space, there is no guarantee it will net tangible benefits — at least not right away.
A recent study by the Massachusetts Institute of Technology found that despite a $30 billion to $40 billion investment into generative AI to create new content, relatively few organizations achieved any measurable results.
The MIT study reviewed more than 300 publicly disclosed AI initiatives, finding that 95% of organizations got zero return on investment.
“Despite high-profile investment and widespread pilot activity, only a small fraction of organizations have moved beyond experimentation to achieve meaningful business transformation,” the report found.
Community opposition
As data centers have grown to massive proportions, new projects have faced opposition from neighbors as previously seen with properties such as distribution centers and factories.
Data Center Watch, a project led by AI research firm 10a Labs, estimated that $64 billion in data center projects were blocked or delayed in the United States over a recent two-year period.
Several challenged projects cited in the report are in Northern Virginia, the nation’s largest data center market. There are 42 activist groups campaigning to stop, slow or further regulate the development of data centers, according to Data Center Watch.
In all, the report says, there are at least 142 activist groups across 24 states that are organizing to block the construction and expansion of data centers.
Reasons cited for opposing projects include water consumption, higher utility bills, noise, decreased property values and loss of open space.