Utilities Waited. SoftBank, Google, and the Army Didn't.

SoftBank's 10-gigawatt Ohio campus, Google's $5 billion Anthropic financing in Texas, and the Army's 2.5-gigawatt Fort Bliss complex all landed in the same news cycle, confirming that AI infrastructure has entered a new phase where developers build their own power systems rather than wait for utilities. The common thread: generation, transmission, and compute are being planned as a single integrated platform, with projects measured in gigawatts, not megawatts.

Texas sits at the center of gravity. Google's Anthropic deal puts 7.7 GW of behind-the-meter gas generation near Enterprise and Energy Transfer pipelines, bypassing ERCOT entirely. The Army tapped Carlyle to build self-contained power and closed-loop water systems at Fort Bliss. Meanwhile, Central Texas's corridor from Temple to San Antonio now carries 7,823 MW of planned capacity against just 1,154 MW operating, with 96% of construction pre-leased.

ERCOT is processing over 230 GW of new load requests, with data centers driving 70% of volume. System peak stress has shifted from 4-5 p.m. to 8-9 p.m. as solar fades, making firm gas capacity indispensable. Batteries averaging 1.5-hour duration can't bridge winter double-peaks. PJM's capacity auction cleared gas at roughly 800% price increases over two cycles, hitting the $329.17/MW-day cap. SPP saw thermal interconnection requests jump tenfold in a single year.

Anthropic's Texas buildout exemplifies the new model: behind-the-meter gas turbines near major pipelines, no grid dependency. SoftBank's Ohio project bundles 9.2 GW of gas generation with $4.2 billion in transmission and new 765-kV lines. The Army's Fort Bliss complex designs self-contained generation into the project from day one. Developers still waiting on utility service timelines face a widening competitive gap. Risk: permitting bottlenecks for gas infrastructure and pipeline capacity could stall projects that assume rapid power deployment. MISO flagged "fewer synchronous resources and more inverters" as an active stability concern in its February 2026 report.

Water has become a first-order site selection variable. The Army's Fort Bliss project commits to closed-loop water systems. Edged Data Centers pitches a waterless design in Fort Worth. Western states face 7 billion gallons of annual data center consumption by 2035. WUE is emerging as the standard metric, but it masks real problems: it doesn't capture water quality, discharge chemistry, or competition with agricultural users.

California's regulatory framework could set a national disclosure standard. Over 230 environmental groups have called for a data center moratorium, and $98 billion in projects were blocked or delayed in Q2 2025 alone. In Vineland, New Jersey, DataOne's 20-million-gallon annual estimate for a Microsoft-linked facility faces community skepticism despite closed-loop claims. Risk: developers treating water as an afterthought face moratoriums and rate battles. Those integrating reclaimed water, onsite treatment, and watershed investment into project design gain permitting advantage.

Fort Worth's City Council unanimously tabled Edged Data Centers' $1 billion tax abatement, demanding noise ordinances, water limits, quarterly compliance reports, and ERCOT approval before groundbreaking. Council member Martinez called for restrictions on crypto and data mining facilities. The 2871 Community Coalition is pushing for data center-specific zoning. This is local governance catching up to industrial reality.

At the federal level, AI-backed super PACs have poured at least $2.8 million into seven Texas congressional races. American Mission, funded by OpenAI co-founder Greg Brockman and Palantir co-founder Joe Lonsdale, spent $372,000 backing Chris Gober in the 10th District. Opposing groups funded by Anthropic's $20 million to Public First Action want state-level regulatory authority and whistleblower protections. The outcome shapes whether Texas datacenter permitting gets lighter or tighter. Risk: developers relying on tax incentives without clear operating criteria face extended approval timelines. Fort Worth's May 12 vote and the 2027 Texas legislative session will set precedents.

Global geothermal investment is projected to hit $9 billion by 2030. Fervo Energy's Cape Station, a 500 MW EGS project in Utah, comes online in June as the first large-scale commercial enhanced geothermal system in the U.S. Advanced drilling and closed-loop technology are expanding viable geography far beyond traditional volcanic zones.

Texas is conspicuously absent. Despite abundant subsurface heat, strong power markets, and ERCOT's need for continuous baseload, no Texas geothermal activity appears in current reporting. Risk: capital flows to proven markets first. If Cape Station validates EGS economics at scale this summer, Texas developers without geothermal pilot projects will have missed the positioning window.