Data centers are moving inland, away from some traditional locations

U.S. data center construction is shifting inland, with Texas leading the pipeline and Midwest states such as Wisconsin, Indiana, Michigan and Missouri attracting major projects from Amazon, Google, Meta, Microsoft, OpenAI and CoreWeave. Sightline Climate says 16GW are slated to open in the U.S. this year, but only 5GW are under construction and 30%–50% of projects may be delayed; for 2027, 25GW are announced but only 6GW are under construction. Growth is being constrained by power availability, component shortages and rising local resistance, including a Seminole nation ban on data centers on tribal lands.

Analysis

The key market implication is not “more data centers,” but a re-pricing of where the bottleneck rent accrues. As projects migrate inland, the scarce assets become grid interconnects, transmission, gas peaker support, and electrical gear rather than coastal land or metro proximity; that favors utilities, power-equipment vendors, and local industrial landlords in the interior more than the hyperscalers themselves. The biggest second-order effect is that every delay elongates the cash conversion cycle for AI capacity, which should compress near-term utilization assumptions for the builders while extending the earnings duration for upstream suppliers with already-booked backlogs.

The more important near-term risk is that the delay curve is becoming self-reinforcing: shortages in transformers, breakers, batteries, and switchgear don’t just slow one site, they cascade across portfolios and push procurement into a longer lead-time regime. That creates a gap between headline announced capacity and actually billable compute, which is a subtle negative for the infrastructure-heavy AI names if investors have been capitalizing future revenue too aggressively. Over the next 3-9 months, the market is likely to reward the picks-and-shovels layer with pricing power, while punishing operators exposed to construction slippage and carrying costs.

There is also a regulatory trap in the “move inland” narrative: lower land cost does not equal lower friction. Interior states may offer better power economics, but permitting, community pushback, and water/electricity constraints can still cap realized deployment, which means the current geography shift may be more gradual than consensus expects. The contrarian read is that this is constructive for select power and grid names, but not yet a clean bullish signal for the hyperscalers until there is evidence that delayed megawatts are actually converting into energized capacity.

Key Decisions for Investors

• Long ETN/EMR/KR? No direct ticker set provided; preferred expression is long power-equipment suppliers via EMR on a 6-12 month horizon, betting on backlog re-pricing and pricing power from transformer/switchgear scarcity; stop if order growth rolls over for two consecutive quarters.
• Long DUK/EXC-style regulated utilities with Texas/Midwest load exposure on a 12-month view; these names can monetize load growth if they can win interconnect rights, but trim if regulators cap data-center tariffs.
• Short a basket of the most capex-exposed hyperscalers on the first rally after capacity announcements, especially AMZN/MSFT/GOOGL, using 3-6 month put spreads; thesis is that revenue recognition lags and delivery delays force multiple compression before the market sees the cash flow.