Data centers in NWGA: Floyd County at a crossroads

Floyd County is weighing whether to attract data centers that could expand the local tax base and help lower property taxes, but officials stress there will be no tax incentives. The article highlights major policy and infrastructure issues around zoning, water use, electricity demand, and legal limits on banning a lawful industry, with no specific deal or company announced. Broader data center sentiment remains mixed nationally, with public support weakening in some Virginia markets.

Analysis

The investable signal is not “data centers are good or bad,” but that local permitting is becoming the binding constraint on AI capex. As more counties copy a playbook of zoning, setbacks, water caps and moratoriums, the industry’s growth rate will increasingly depend on execution friction rather than demand, which favors incumbent operators with existing power and land positions over new entrants trying to greenfield projects. That shifts value upstream toward utilities, switchgear, transformers, cable, and gas-turbine vendors while pressuring developers whose returns depend on fast entitlement and cheap public infrastructure.

The second-order risk is that political resistance arrives after capital is already committed, creating stranded assets, longer payback periods, and more costly financing. If localities insist on public-record water deals, no tax incentives, and bankruptcy protections, the economics move from “land grab” to “option value,” where only the largest balance sheets can absorb permitting delays and policy surprises. That environment is bearish for speculative data-center REITs and private developers with aggressive pipeline narratives, but constructive for regulated utilities with rate-base growth and for equipment suppliers selling into an elongated build cycle.

The contrarian point is that the market may be overestimating how quickly a local backlash can actually stop the buildout. AI inference and training demand will not vanish on sentiment, and counties that refuse projects may simply export them to more accommodative jurisdictions, preserving national demand while redistributing local winners and losers. In that scenario, the near-term headline risk is high, but the medium-term capex pie stays large; the main trade is dispersion, not an outright short on the theme.

Key Decisions for Investors

• Long REGI/utility-adjacent infrastructure via XLU vs short data-center levered REIT exposure for a 3-6 month dispersion trade; thesis is that permitting friction lengthens utility capex visibility while squeezing speculative land-banking economics.
• Buy LECO/ETN/PH-style electrical infrastructure names on pullbacks for 6-12 months; best risk/reward is that every incremental data-center project requires more grid hardware even if site approvals slow.
• Short over-owned data-center REITs with aggressive growth expectations versus long utility names in a pair trade for 1-2 quarters; stop if state-level incentives or federal fast-tracking reduce entitlement friction.
• Add optionality on gas-turbine and backup-power suppliers (e.g., GNRC or lower-beta industrial power names) into any AI pullback; if local power constraints tighten, on-site generation demand can accelerate within 6-18 months.
• Avoid or fade private-market stories relying on rapid greenfield buildouts until zoning/water agreements are signed; the key risk is a 12-24 month delay that compresses IRRs more than revenue assumptions imply.