What is a hyperscale data center?

Data centers have existed in Florida for decades. Small office-park facilities storing business files, hosting websites, or supporting regional companies. Most people never noticed them. "Hyperscale" is a relatively new category, tied to the explosion of cloud computing and, more recently, artificial intelligence. The scale is genuinely different — and that difference explains why a proposed data center today is a materially bigger thing than one proposed ten years ago.

The size difference, plainly

A traditional corporate data center might occupy 5,000 to 50,000 square feet — roughly the footprint of a medium office building — and consume 1 to 10 megawatts of electricity. That is enough power for 1,000 to 8,000 typical homes, concentrated in one facility.

A hyperscale data center campus typically starts at 500,000 square feet and runs to several million square feet. Power consumption starts at 50 megawatts and routinely reaches 500 to 1,200 megawatts for the largest AI-focused facilities. That is enough electricity to power 400,000 to 1,000,000 homes. A single campus. In one place.

The industry is not just adding more traditional data centers. It is building a fundamentally different category of facility, at a pace and size that most local zoning frameworks were not designed to handle.

What drives the size

• Cloud computing consolidation. Amazon Web Services, Microsoft Azure, Google Cloud, Meta, and Oracle together operate most hyperscale facilities. Running a small fleet of enormous facilities is more efficient than running a large fleet of small ones.
• Artificial intelligence training and inference. Training modern AI models requires thousands of high-end graphics processors (GPUs) running continuously for weeks or months. Each GPU consumes hundreds of watts and generates significant heat. A single AI training cluster can consume 50-100 megawatts by itself. Recent AI-specific facilities are being built at 500-1,200 megawatts and beyond.

What makes hyperscale different from a regular data center

Water use

Traditional data centers often use air cooling. Hyperscale facilities, because of their heat density, frequently rely on evaporative cooling, which can consume 1-5 million gallons of water per day. Single large facilities can use water equivalent to 10,000 to 30,000 households.

Continuous, always-on power draw

A traditional data center's power use fluctuates with demand. A hyperscale facility running AI training workloads draws close to peak power 24 hours a day, every day, for months at a time. This "baseload" demand stresses the grid in ways that intermittent industrial demand does not.

Infrastructure footprint

To serve 500 megawatts of demand, the local electric utility may need to build new substations, transmission lines, or even new generation capacity. The facility itself may be one or two buildings, but the infrastructure chain behind it often reshapes the regional grid.

The "hyperscale" threshold

No universally standardized definition, but industry and regulators generally use these thresholds:

• Florida's SB 484 (2026): defines "large-scale data center" as facilities with an anticipated monthly peak load of 50 megawatts or more.
• Florida's HB 1007 (original version): used "25 megawatts of energy resources" as the hyperscale threshold.
• Industry usage: often uses "hyperscale" for facilities operated by major cloud providers (Amazon, Google, Microsoft, Meta, Oracle, Apple) at campus scale, regardless of specific megawatt thresholds.

Why Florida is a major growth market

• Fiber connectivity: extensive fiber backbone along I-4 and the east coast
• Power availability: major generation capacity from FPL, Duke Energy Florida, and Tampa Electric
• Rural land availability: large contiguous parcels at accessible prices
• Tax incentive programs: sales tax exemptions for qualifying equipment
• Climate stability: relatively low hurricane-related grid failure rates in interior counties

Our Risk Calculator rates each of Florida's 67 counties on these structural factors.

The terms you'll hear

• Campus: cluster of multiple data center buildings on a single parcel, sharing infrastructure
• Gigawatt (GW): 1,000 megawatts. Several currently-proposed AI facilities will eventually reach this scale
• Interconnection: physical and contractual connection between the data center and the electric grid
• Large-load customer: the term SB 484 uses for facilities drawing 50+ MW
• Tariff: the utility's rate structure for a specific class of customer. Under SB 484, large-load customers must be on specific tariffs that allocate infrastructure costs to them, not to residential ratepayers
• Comprehensive plan amendment: first major county-level approval most hyperscale facilities require. Critical homeowner input opportunity
• Consumptive use permit: water permit from one of Florida's five regional Water Management Districts

This guide is educational and not legal advice. Before taking action that may affect your property or your legal rights, consult a Florida-licensed attorney.