Why smaller data centers fit rural Opportunity Zones

In this guide, a small data center draws roughly 1 to 50 megawatts (MW) from the grid. There is no legal definition, so the range follows where regulators and utilities draw their lines. On June 18, 2026, FERC ordered the six regional grid operators to justify or reform their rules for large loads, and proposed defining a large load as a site with peak demand of 50 MW or more connecting to the transmission grid above 69 kV.^1,2 Utilities set their own thresholds: AEP Ohio's data center tariff applies above 25 MW, and Georgia's large-load rule applies to new customers above 100 MW.^8,9

At the low end are edge sites. NREL describes edge data centers as facilities under 20 MW, often placed on distribution feeders close to users.⁴ Above them sit regional colocation buildings, enterprise facilities and AI inference halls, often built in phases from prefabricated power and cooling modules. Containerized units, which arrive as complete enclosures, are the smallest building block.

Size changes the budget, not the tax rules. Cushman & Wakefield's 2026 cost guide puts all-in greenfield costs at about $8.9 million to $23.3 million per MW, excluding chips and GPUs, with power infrastructure the largest single component.¹¹ At those rates a 10 MW facility is roughly a $90 million to $230 million project before servers. The Opportunity Zone tests that apply are the same as for any data center; see data centers in Opportunity Zones.

Because they can often avoid the transmission process that governs large campuses. A load of a few megawatts can sometimes be served from a feeder or substation with spare capacity, without new high-voltage lines. A campus of several hundred megawatts usually needs new transmission and substations. Dominion Energy said in 2024 that connecting data centers above 100 MW could take one to three years longer than before, for total waits of up to seven years in Northern Virginia.³

Smaller is not automatically faster. NREL's November 2025 study warns that edge data centers under 20 MW are often sited on distribution feeders that are already constrained, and that several small sites on one feeder can add up to a large load.⁴ The advantage exists only where the utility confirms headroom, so the first question for any small site is a written capacity answer from the utility.

Four ways a small load can get power sooner

• Existing substation headroom. Rural substations built for a closed mill, plant or irrigation load may have spare transformer capacity. Electric cooperatives own about 42% of U.S. distribution lines and their lines cover about three-quarters of the country's land area, so in many rural zones the co-op is the first call.¹²
• Flexible load. Duke University's Nicholas Institute estimated that the largest U.S. balancing authorities could absorb about 76 GW of new load if it accepted curtailment 0.25% of the time, and about 126 GW at 1%.¹³ A small site with batteries, or with work that can shift in time, is easier to run flexibly.
• Curtailed or stranded generation. Where wind or solar output is curtailed for lack of transmission, a nearby load can use power that would otherwise be wasted. EIA reports that California's grid operator curtailed about 3.4 million MWh of wind and solar in 2024, up 29% from 2023.¹⁴
• Co-location behind the meter. Placing load at a generator can avoid some grid upgrades, but the rules are still forming: in December 2025 FERC directed PJM to create clearer rules and service options for co-located load.¹⁵ New on-site generation that needs its own grid connection faces generator queues where the median wait exceeded four years for projects built in 2018 to 2024.¹⁶

How these options compare is covered in power interconnection for Opportunity Zone data centers.

Often, yes, if power is committed before the fund invests. A Qualified Opportunity Zone Business (QOZB), the operating company a Qualified Opportunity Fund (QOF) invests in, can treat cash as a qualifying asset for up to 31 months if it has a written plan designating the cash for zone property, a written schedule consistent with the ordinary start-up of a business, and spends the cash substantially as planned. Overlapping or sequential plans can extend the period to 62 months from the first cash infusion.^5,17 A purchased existing building must also be substantially improved within a 30-month period.¹⁸

A multi-phase campus often cannot spend that fast, because power for later phases may be years away. A small facility built from factory-made modules has fewer moving parts: one utility feed, one building or pad, and equipment assembled off site while site work proceeds. Modular methods are spreading to large builds too: NTT has put the traditional design and construction period for large data centers at three to four years and has targeted 2028 to use modular construction to halve it.¹⁹

Relief for delay is narrow. The working capital period is tolled for delay caused by government action on a completed application, and a utility's study backlog is not clearly covered.⁵ Notice 2026-55 asks whether a QOZB should be allowed to modify an existing plan; that guidance is still pending.²⁰ Details are in the working capital safe harbor and data center power timelines.

Two things: a larger basis step-up for investors in a rural fund, and a lower improvement bar for existing buildings. A rural area is any area other than a city or town with more than 50,000 people and any urbanized area contiguous and adjacent to it.²¹ A Qualified Rural Opportunity Fund (QROF) is a QOF that holds at least 90% of its assets in Opportunity Zone property in zones made up entirely of rural area. For investments made after December 31, 2026 and held five years, its investors get a 30% basis step-up on the deferred gain instead of 10%.^21,6

The amount taxed at year five is based on the lesser of the deferred gain and the investment's fair market value, so a fund that has fallen in value produces a smaller inclusion.²¹ The step-up math is explained in the basis step-up: 10% and 30%; confirm how it applies to your own gain with your tax adviser.

The 50% rule favors reusing old buildings

In a zone made up entirely of rural area, an existing building is substantially improved if additions to its basis within any 30-month period exceed 50% of its adjusted basis at the start of that period, instead of 100%. The rule has applied since July 4, 2025, including to wholly rural zones on the current map, and land is excluded from the test.⁷ It applies whatever kind of fund owns the building.

Illustrative example: a QOZB buys a closed textile mill for $3 million, allocating $500,000 to land and $2.5 million to the building. In a wholly rural zone it must add more than $1.25 million to the building's basis within 30 months; in other zones, more than $2.5 million. Only additions to the building's basis count, such as structural, roof, envelope and building electrical work. Much of the power and cooling equipment in a data hall is separately depreciable and is tested on its own as original-use property, so the cost split between building and equipment decides whether the bar is met. The 50% rule matters most where the building is large and the first phase is small, which is the typical phased small data center. See substantial improvement and converting industrial buildings into data centers.

They let a fund put capital to work in smaller pieces and spread risk across sites. Investors generally have 180 days from a gain to invest it in a QOF.²¹ A fund that holds several small projects can direct each closing to a project that is ready to spend, instead of holding cash while one large campus waits for power.

• Each QOZB is tested on its own. At least 70% of the tangible property it owns or leases must be Opportunity Zone business property, and at least 50% of its gross income must come from the active conduct of business in a zone.¹⁸
• The fund is tested as a whole. At least 90% of its assets must be qualifying Opportunity Zone property, averaged over two semiannual testing dates.²¹
• A QROF must keep at least 90% of its assets in wholly rural zones, so one urban site can cost every investor the 30% step-up.²¹
• Holding each site in its own QOZB keeps a weak site from pulling another site below the 70% test.
• Operated, multi-tenant facilities fit the active trade or business requirement more naturally than a single triple-net lease, which the regulations say is not enough on its own.¹⁸

The trade-offs are in single-asset versus multi-asset Opportunity Zone funds and what a Qualified Rural Opportunity Fund is.

They avoid the most common objections, but approval is never automatic. Gallup found in March 2026 that 71% of Americans opposed building AI data centers in their local area, compared with 53% opposing a local nuclear plant; half of opponents cited excessive resource use, including 18% each who named water and energy.²² Data Center Watch, a tracker run by the AI research firm 10a Labs, counted at least 75 projects worth about $130 billion blocked or delayed in the first quarter of 2026, and local opposition groups rising from 396 at the end of 2025 to 833 by the end of March.²³

A 10 MW facility draws a small fraction of a gigawatt campus's power and needs far less land than a campus seeking 200 acres or more.¹⁰ It can be cooled with air or closed-loop systems that use little water; Microsoft says all its new data center designs since August 2024 use closed-loop, zero-evaporation cooling.²⁴

Reuse can still draw opposition. In Jay, Maine, a plan to turn a former paper mill site into a data center collapsed after local resistance.²⁵ Moratoriums can also be broad: Pima County, Arizona adopted a 120-day moratorium on new data center projects in September 2026.²⁶ Early, specific disclosure of power, water, noise and tax effects is the practical answer. The community side is in the economic impact of data centers in Opportunity Zones.

Yes, driven mainly by AI inference, enterprise and regional colocation needs. JLL's 2026 Global Data Center Outlook says AI was about a quarter of data center workloads in 2025, that inference could overtake training as the dominant AI requirement in 2027, and that AI could be half of all workloads by 2030.²⁷ NREL's November 2025 study expects inference to make up 90% of AI workloads by 2030, served in part by many low-latency edge data centers under 20 MW.⁴

Building is moving to rural places. Pew Research Center found in April 2026 that 67% of planned U.S. data centers are in rural areas, compared with 13% of those operating today.²⁸ Pew counts facilities, not megawatts, so the figure shows where projects are headed rather than how large they are. Latency-sensitive inference favors rural sites with good fiber routes to nearby metros; see edge data centers in Opportunity Zones.

Cost per megawatt, connectivity, staffing, tenants, the multi-site tests and the map. None is a reason to avoid small sites, but each needs a plan before fund equity is called.

The map is the largest open question. As of October 27, 2026, Treasury had not published a national list of zones taking effect January 1, 2027. Treasury certified Idaho's full slate of 25 tracts in early October, and states using the single 30-day nomination extension had until about October 28.^31,32 Governors are not required to nominate rural tracts, and under Notice 2026-40, property acquired after 2026 in a current zone that is not redesignated generally does not qualify.³³ Track status in the new Opportunity Zone maps for 2027.

Policy risk is the other open item. Bills introduced in September 2026 would exclude data centers from Opportunity Zone benefits; none had been enacted as of this review.³⁴ Investors should model current law and read how a sponsor discloses the proposals. For fiber planning, see fiber for rural Opportunity Zone data centers.

OZ Data Centers sponsors funds of this kind. The rural rules in full are in rural data centers and Opportunity Zones and rural Opportunity Zones, explained.