AI compute regions, GPU supply, and grid stress

PJM

Largest US cloud/data-center concentration; strong hyperscale and specialty GPU provider presence, but premium enterprise capacity can be tight.

PJM's 2026 forecast process explicitly evaluates large-load adjustments and distinguishes firm from non-firm projects.

Data-center-driven load growth is a core regional issue; near-term requests are being vetted more strictly.

Use PJM and Interconnection.fyi generation queues as supply-pipeline context; queue depth does not equal deliverable capacity.

Loudoun shifted data centers from by-right treatment toward conditional/SPEX review for many sites.

Medium: water and generator permitting scrutiny are rising, though power/interconnection is the more visible bottleneck.

Dense transmission and substation footprint, but public HIFLD layers do not reveal spare capacity at nodes.

ERCOT

Strong mix of cloud regions, colocation, and emerging on-site generation campuses; spot/marketplace prices can be attractive but operational quality varies.

ERCOT forecasts and public updates show major load-growth pressure from data centers and other large loads.

ERCOT reported roughly 410 GW of large loads seeking interconnection, about 87% data centers, as of March 26, 2026.

ERCOT reported more than 453 GW of active generation interconnection requests in February 2026, led by storage and solar.

State-level large-load interconnection process is changing; public concern focuses on reliability, cost allocation, and water.

Medium to high in water-stressed subregions; site-level cooling design matters.

Large state grid with visible congestion and price nodes; public data helps market analysis but not exact spare-capacity claims.

PJM

Strong cloud-region relevance through AWS us-east-2 and expanding central-Ohio data-center demand.

PJM's load-forecast process includes large-load adjustment scrutiny; Ohio utilities are part of the debate.

AEP Ohio and manufacturing stakeholder disputes show material uncertainty in data-center load assumptions.

Use PJM generation queue and Interconnection.fyi as supply-pipeline context; queue capacity can be speculative.

Debate focuses on tariffs, cost allocation, and whether forecasts overstate committed load.

Medium: less severe than desert markets, but site-level cooling and community impacts still matter.

Central Ohio has data-center concentration, but public transmission layers cannot confirm energization timing.

Western Interconnection

Emerging hyperscale and western compute market; public GPU marketplace depth is weaker than established cloud regions.

Utah policy research reports substantial existing, under-construction, and planned data-center capacity.

Proposed multi-GW campuses make Utah a high-attention market even when specific cloud GPU offers are not yet visible.

On-site gas and utility-specific interconnection plans matter more than generic western queue totals.

Backlash centers on scale, water, emissions, and Great Salt Lake/desert impacts.

High: arid-region water stress and cooling choices are core buyer and community risks.

Transmission access and gas pipeline proximity are relevant, but proposed off-grid designs complicate public scoring.

Arizona utilities / Western Interconnection

Important western cloud and colocation market with growing AI demand, but fewer transparent GPU-market signals than PJM or ERCOT.

Arizona policy and media reports identify data centers and other large loads as a reliability and electricity-cost challenge.

Large-load pressure is visible but less standardized publicly than ERCOT's queue updates.

Use Western interconnection, utility planning, and EIA signals for context; public project-level supply visibility is limited.

Public debate centers on heat, water, and energy burden rather than a single state-wide moratorium.

High: desert water constraints and cooling demand are central buyer-risk factors.

Transmission proximity can be mapped, but desert heat, cooling load, and water constraints are central site variables.

Southeast utilities

Growing southeast compute market with strong fiber and enterprise demand; public GPU price transparency is weaker than marketplace-heavy regions.

Data-center ordinances and local planning activity show a fast-moving development market.

Large-load interest is visible through ordinances, utility planning, and project announcements, but not one unified public queue.

Interconnection.fyi and regional utility planning are the best V1 supply-pipeline proxies.

Local governments are actively drafting data-center ordinance language and review standards.

Medium: regional water concerns are site-specific and should be verified by county/utility.

Transmission access should be mapped as context only; deliverability remains utility-specific.

PJM / MISO

Major Midwest connectivity and enterprise market; cloud GPU buyer value depends on provider-specific region availability.

State policy debate shows electricity-cost and grid-impact concerns around continued data-center incentives.

Large-load risk is visible through policy and market reports, but public project-level queues are fragmented.

Use PJM, MISO, and Interconnection.fyi queue context rather than a single market-only score.

Illinois officials have moved to slow or review tax-break driven data-center development amid power and water concerns.

Medium: generally lower scarcity risk than Southwest markets, but community and cooling concerns remain.

Chicago-area grid and fiber advantages are material, but deliverability must be verified by utility.

Pacific Northwest utilities / BPA context

Strong cloud-region relevance through AWS us-west-2 and Pacific Northwest data-center clusters.

Pacific Northwest reports and advisory processes identify data centers as a major electricity-demand driver.

Large-load pressure is visible in Oregon and Washington policy work, but public queues are utility-specific.

Use EIA, regional planning, and Interconnection.fyi as broad context; hydro and transmission constraints require local review.

Washington and Oregon policymakers are debating cost allocation, water, and grid expansion impacts.

Medium: lower scarcity than desert regions, but water quality/use and river-system impacts are active issues.

BPA/hydro context is a major regional differentiator, but public line maps do not reveal deliverable capacity.