Regional guides
The same machine, six different design bases. The matrix is the durable shape of each market — every volatile figure behind it is date-stamped on the region pages, pulled live from the numbers register.
| Region | Speed to power | Power cost | Gating & risk | Water |
|---|---|---|---|---|
| Northern Virginia & PJM | Slowest of the majors — PJM queues run in years | Moderate, rising with transmission buildout | County land-use politics; FERC co-location rules in flux | Moderate stress |
| Texas & ERCOT | Fastest interconnect of the US majors | Low average — with scarcity-price spikes you own | Emergency curtailment regimes (SB6-class kill switch) | West Texas: scarce — plan dry or hybrid cooling |
| Ireland & Dublin | Reopened on paper (Dec 2025) — Greater Dublin still fully constrained | High — SEM prices ran ~a third above the EU benchmark in 2025 | Matched dispatchable generation + ≥80% additional renewables to connect | Sector use is small, but Dublin supply is near capacity |
| The Nordics | Queues now real — maturity-gated allocation (SE), inquiry glut (FI) | Europe's cheapest power in the far north (~€9–17/MWh, 2025) | Waste-heat mandates in statute; DC tax breaks abolished | Low stress; free cooling most of the year |
| The Middle East | Fast — sovereign-backed power delivery | Low energy; imported-supply-chain premium | Export-control geometry on the silicon itself | Scarce — dry/hybrid cooling is the design basis |
| Asia-Pacific | Bimodal — allocation-gated hubs vs fast scale markets | High in the hubs; moderate in scale markets | Capacity-gated allocations (Singapore); grid-frequency splits (Japan) | Varies; tropical humidity in the scale belt |
Northern Virginia & PJM
The densest data-center market on earth sits inside PJM, where interconnection queues, transmission buildout, and county-level land-use politics now set the pace — not capital. Data-center alley's grid position makes speed-to-power the defining regional constraint.
Verdict: Deepest ecosystem, longest queue — buy speed with an existing position, or don't play.
Texas & ERCOT
ERCOT's energy-only market, fast interconnection paths, and co-location with West Texas wind, solar, and gas make it the speed play — priced in exposure to scarcity pricing, curtailment economics, and grid-islanding risk that the design basis must own.
Verdict: Fastest to energize — priced in scarcity, curtailment, and islanding risk you must design for.
Ireland & Dublin
Dublin was the canonical moratorium market; since December 2025 it is the canonical conditional market. CRU's connection policy reopened applications — priced in matched dispatchable generation and an 80%-additional-renewables obligation — while EirGrid still classifies Greater Dublin itself as fully constrained. It remains the leading indicator for how European grid operators gate data-center load.
Verdict: Bring your own generation and renewables — Dublin itself is still a 2030s conversation.
The Nordics
Cheap northern hydro and cool climates made the Nordics the efficiency play — 2025 day-ahead averages ran ~€9–17/MWh in the far-northern zones. But the tax honeymoon is over (Sweden 2023, Finland July 2026), waste-heat mandates are hardening into statute, and the grid queues are now real: 7 GW of data-center requests pending at Svenska kraftnät and >100 GW of inquiries at Fingrid shape what actually gets built.
Verdict: Cheap hydro and free cooling, priced in mandates, ended tax breaks, and distance — built for training, not latency.
The Middle East
Sovereign-backed gigawatt campuses in the UAE and Saudi Arabia trade the West's interconnection-queue problem for a different design basis: hot-arid climate hardening, water-scarce cooling, imported supply chains, and export-control geometry on the silicon itself.
Verdict: Power is the easy part — silicon allocation and climate hardening set the real schedule.
Asia-Pacific
APAC splits into land-and-power-scarce hubs (Singapore's capacity-gated allocations, Japan's grid-frequency split) and scale markets (India, Malaysia, Indonesia) where subsea-cable geography and 50 Hz grids drive different electrical and network design bases than the US default.
Verdict: Two different games: win allocations in the hubs, win execution in the scale markets.