This is how difficult it has gotten for tech giants needing more data centers: Meta, not willing to wait years for conventional data centers to be completed, is placing AI servers inside massive tent-like structures that can be erected in a matter of months.
Meta has built a series of what it calls rapid deployment structures at its New Albany, Ohio campus. According to data center tracking firm Cleanview Energy, Meta has built or is constructing six large structures at the Ohio site. Five of the facilities measure approximately 125,000 square feet each.
According to Cleanview, local permit records indicate construction began earlier this year, while satellite imagery suggests the facilities were completed far more quickly than conventional data center buildings.
The speed difference is of course enormous. Traditional structures at the same campus reportedly required between two and three years to complete. By contrast, the new tent-like facilities moved from permitting to deployment within months.
Meta CEO Mark Zuckerberg previously disclosed plans to use weather-resistant structures to accelerate deployment of data center capacity. Similar facilities are now reportedly being developed at a Meta site in Tennessee.
The tent-based facilities reveal the pace of competition among tech companies: data center construction is now arguably as important as advances in AI models.
A Growing Urgency
While the tent-like structures themselves have attracted attention, the larger story is the growing urgency surrounding AI infrastructure. Building AI systems requires vast quantities of graphics processors, networking equipment and electrical power, and demand for those resources has increased so quickly that construction timelines are a competitive disadvantage.
The challenge extends beyond buildings. Meta is also investing in dedicated power generation to support its AI expansion. The Ohio facility is supported by a nearby 200-megawatt power installation using modular gas turbines. The company has also pursued off-grid energy arrangements that allow facilities to operate independently of local utility infrastructure. Such approaches reduce dependence on grid interconnections, which can take years to secure in some regions.
The use of on-site generation is becoming common across the AI sector as operators seek faster deployment schedules. Obtaining electrical capacity has become a real roadblock facing large AI projects.
Cleanview estimates that approximately 2 gigawatts of behind-the-meter data center capacity is currently available in the US, with another gigawatt expected to come online this year. But the pace is quickening: if current projects remain on schedule, total capacity could reach roughly 13 gigawatts by the end of 2027.
The 13-gigawatt figure shows the gargantuan scale of investment now flowing into AI infrastructure. That amount of generating capacity is comparable to the output of multiple large nuclear power facilities and exceeds the electricity consumption of some major cities.
Meta has already indicated that it intends to spend as much as $145 billion on data centers and related capital projects. That level of investment is creating a new phase of AI development in which traditional approaches are being replaced by faster, more flexible alternatives designed to keep pace with escalating demand.
