Reference brief — US (federal + states), Europe (EU framework + country markets), and Asia-Pacific (China, India, Singapore, Malaysia, Japan, South Korea and the wider region — Part G). Covers tax schemes, financing/grants/loans, ecological impact, electricity, water, energy mix and cost, the equipment supply chain, political positions with a fact-check layer, and grid/bills/water/land stress reports.
Used AI for compiling and double-checking all the references.
Prepared: January 2026 · Data vintage: most figures 2024–H1 2026 (sources listed at end)
See Part F (Addendum, 4 July 2026) for verified late-breaking updates that refine, but do not overturn, this report: Virginia’s finalized per-kWh data-center tax, Illinois’s and Ohio’s incentive pauses, and the CADA proposal’s legislative status.
0. How to read this / scope note
This is a working reference, not an exhaustive statute-by-statute compendium. A truly complete per-jurisdiction file (all 50 US states × every incentive clause, plus ~30 European countries × every dimension) runs to hundreds of pages and changes monthly; the sector is in the middle of a regulatory rewrite almost everywhere. What follows prioritizes:
- Depth on the jurisdictions that actually matter for siting decisions (Virginia, Texas, the Nordics, Ireland, France, Germany, UK, Netherlands).
- Breadth via summary tables for the long tail.
- The 2025–2026 inflection: the binding constraint has moved from tax to power and permitting, and the political mood has moved from courtship to scrutiny. Both shifts are reshaping the incentive landscape faster than the headline tax rules suggest.
Where a jurisdiction is thin below, it’s because public, current, reliable data is thin, flagged rather than invented.
1. Global framing (why the rules are moving)
- Capital scale: Data-center investment is projected toward ~$1 trillion by 2027 (PwC). In 2025, US tech firms committed well over $650bn in a single year; data centers accounted for ~78% of venture capital in the built environment.
- Demand shock: US utilities received interconnection requests for at least 700 GW in 2025 — more than the entire US consumed in 2023 (477 GW). European DC demand is forecast to grow from ~10 GW IT load (2024) to ~35 GW by 2030 (McKinsey).
- The bottleneck migrated. 2021–2024 the constraint was IT hardware (server lead times hit 52 weeks). From 2025 it is grid power and heavy electrical equipment (transformers, switchgear, substations) plus permitting. See §7.
- The politics flipped. Governments spent a decade competing to attract these facilities with tax breaks. In 2025–2026 many are adding conditions, caps, transparency mandates, or outright moratoria — driven by electricity-bill anxiety, water stress, land/housing competition, and grid stability.
- Sovereignty overlay (Europe). The EU hosts only ~5% of global AI compute vs ~75% for the US (Epoch AI). The response — AI Continent Action Plan, AI Factories, the Cloud & AI Development Act — is now explicitly industrial-policy, aiming to triple EU DC capacity in 5–7 years.
PART A — UNITED STATES
2. US federal layer
The US has no single federal data-center statute; the sector is shaped by federal tax and industrial tools layered under state incentives.
- 100% bonus depreciation (reinstated). The 2025 “One Big Beautiful Bill” reinstated full first-year expensing of qualified capital equipment/property placed in service before Jan 1, 2030, instead of multi-year MACRS depreciation. Large, immediate deduction of server/equipment cost — a major cash-flow incentive stacked on top of state breaks.
- Defense Production Act (grid equipment). In April 2026, the administration invoked DPA Section 303 to designate large-scale grid infrastructure as essential to national defense, authorizing emergency federal financing to expand domestic supply of transformers and related components (a direct response to the supply-chain crunch in §7).
- Federal land / energy pushes. Nuclear-adjacent siting is being actively encouraged (e.g. AWS ~$20bn Three Mile Island conversion; Google–Kairos 500 MW SMR targeted 2030).
3. US state incentives — how the machinery works
At least 41 states (Data Center Coalition) offer incentives; a CNBC analysis found 42 states provide full/partial sales-tax exemptions or have no sales tax at all, 37 with DC-specific legislation, and 16 that granted roughly $6bn in exemptions over five years. The common instruments:
- Sales & use tax exemption — the workhorse. Covers servers, networking, cooling/HVAC, power infrastructure, software, and in many states the electricity itself. This is the single biggest lever.
- Property / tangible-personal-property (TPP) abatement — administered locally, so it varies within states. Some states value DC equipment at salvage value (~5%) for ad-valorem purposes.
- Income / payroll tax credits — less common; tied to job and wage thresholds.
- Utility / electricity-tax relief — exempting power from sales or excise tax, a big deal given power is 30–50% of opex.
Qualification levers that differentiate states: minimum capex thresholds (from ~$1M in Iowa to $400M in Alabama), job-creation minimums (usually small — DCs employ few), wage floors (often county/state average or a multiple), rural vs urban tiers, and exemption duration (typically 10–50 years).
Structural caveat: DCs create large construction crews but only ~15–100 permanent jobs. That mismatch is the core of the current political backlash (§5).
4. US state-by-state (major markets detailed, then summary)
Tier-1 markets
Virginia — the global #1 market (“Data Center Alley,” Loudoun County). Sales & use tax exemption on qualifying equipment is the marquee incentive. Estimated foregone revenue: ~$730M+ (FY2024), rising to an estimated ~$1.6bn annually. Loudoun County alone expected ~$900M in DC property/real tax in FY2025 — nearly its whole operating budget. Reform wave (2025–26): ~100 bills nationally; Virginia bills include HB897 (tie the exemption to new energy standards), SB393 (a $3/sq-ft “land conservation tax” on facilities >25,000 sq ft), HB784 (annual transparency reporting). Gov. Spanberger prioritizing impact review. Residential power ~15.94¢/kWh.
Texas — #2 market, ERCOT grid. Sales-tax exemption on equipment and electricity for facilities ≥100,000 sq ft, ≥$200M investment over 5 years, ≥20 jobs at above-average wages → 10–15 year abatement. Estimated >$1bn in subsidies in 2025; cheap commercial power (~8.35¢/kWh; deregulated 100+ retail providers). Texas DC water use projected at 49bn gallons (2025) → up to 399bn (2030) (HARC/Univ. Houston).
Other significant incentive states
| State | Headline mechanism | Threshold / notable terms |
|---|---|---|
| Georgia | Sales-tax exemption on equipment | $100M–$250M by county population; ~$296M value by 2025 |
| Arizona | Computer Data Center Program (TPT + Use tax exemption) | $25M/$50M min by location; 10 or 20-yr certification |
| Ohio | Sales-tax abatement; no TPP tax | ≥$100M invest + $1.5M annual payroll |
| Illinois | Sales-tax exemptions (equipment + energy) | ≥$250M capex; ~$370M value; itemizes recipients |
| Iowa | 100% sales/use abatement incl. cooling & electricity | as low as $1M; larger tiers >$200M; no equipment property tax |
| Nevada | Up to 75% abatement (personal property/sales/use) | investment + job criteria; itemizes recipients |
| Washington | Sales/use exemption (servers, power infra) | HB1846 (2022) expanded to rural + urban; annual certificate cap |
| West Virginia (2025) | Sales/use exemption + salvage-value (~5%) property valuation | new 2025 law; consolidated guidance still pending |
| Kentucky | Sales/use exemption | HB775 (2025) expanded eligibility; thresholds by county pop. |
| Louisiana | Sales/use rebate on equipment + construction | Act 730 |
| Kansas | 20-yr sales/use exemption | SB98 |
| Alabama | Up to 30-yr abatement | ≥$400M + 20 jobs @ $40k avg comp |
| Minnesota | 20-yr exemption via refund model | ≥$30M + 25,000 sq ft |
| Mississippi | Sales/use exemption + income & franchise tax exemption | ≥$50M + 50 jobs @ 150% state avg wage |
| North Carolina | Sales-tax exemption incl. electricity | $150M (poorer counties) / $225M (others) |
| Indiana | Personal-property tax exemption | ≥$10M |
| Wyoming | Sales-tax exemption on equipment | ≥$5M; +power/cooling break at ≥$50M |
| Oregon | No sales tax + enterprise-zone property exemptions | e.g. Hillsboro zones |
| Nebraska | Nebraska Advantage Act (abatement + credits) | comprehensive structure |
| Tennessee | Sales-tax breaks (equipment + power) + jobs credit | ≥$250M |
| South Carolina | Sales-tax exemption (equipment + electricity) | ≥$50M + 25 jobs |
No dedicated incentive / structural note: Delaware (no sales or property tax anyway), Alaska (no statewide sales tax), plus several states with no DC-specific legislation. New York exempts equipment for internet DCs. Eight sales-tax states give DCs no exemption.
5. US backlash & reform (the live risk to incentives)
- The jobs-per-dollar math is politically exposed: one Microsoft DC in Illinois reportedly received $38M+ in exemptions for ~20 permanent jobs.
- Virginia’s own JLARC classed the exemption a “moderate economic benefit” that “does not pay for itself” — better than most other incentives but still net-negative on pure revenue.
- Expect more: transparency mandates, energy-standard conditionality, land/impact taxes, and specialized large-load rate classes (“beneficiary-pays” tariffs) so households don’t subsidize DC grid upgrades.
6. US electricity & water
Electricity (2026): National residential avg ~18.83¢/kWh (April 2026); commercial ~13.51¢. Cheapest commercial: North Dakota 7.05¢, Oklahoma 7.77¢, Texas 8.35¢. Most expensive: Hawaii 46.62¢, California 35.25¢. Utilities requested >$29bn in rate increases in H1 2025 (double H1 2024); prices could rise up to 40% by 2030. Nuance / contested: Several 2025–26 analyses (IER, Columbia CGEP, LBNL) find no clean correlation between DC concentration and higher retail rates so far — top-10 DC states averaged ~14.46¢ vs ~14.39¢ elsewhere. But near-term PJM capacity price spikes show DCs can drive localized increases.
Water: US DCs consumed ~17.4bn gallons directly (2023); projected 38–73bn by 2028; hyperscale can hit ~5M gal/day. Indirect footprint (power generation) is far larger (~211bn gallons in 2023). Google: ~7.8bn gallons withdrawn (2024), ~78% consumed. Local flashpoints: rural Georgia (Meta), Arizona (ag competition), South Carolina. MSCI: ~1 in 4 of 14,000 DC assets may face increased water scarcity by 2050.
PART B — EUROPE
7-EU. The EU framework (applies across member states)
Europe has no single binding efficiency standard at EU level, but a layered framework member states implement (or exceed):
- Energy Efficiency Directive (EU) 2023/1791 — Article 12. DCs with IT power ≥500 kW must monitor and report KPIs: total energy, PUE, temperature set-points, waste-heat use, water usage, renewable share → into a centralized EU database. Crucially, it sets no binding minimum efficiency targets — those are left to member states, which is why national regimes diverge sharply.
- Energy Tax Directive. Sets minimum electricity tax and allows reduced rates for business/energy-intensive use — the legal basis for both the Nordic reduced rates and their recent removal.
- AI Continent Action Plan (April 2025). InvestAI aims to mobilise ~€200bn (incl. €20bn for AI “gigafactories”); target to triple EU DC capacity in 5–7 years; ≥19 AI Factories.
- Cloud and AI Development Act (CADA). Commission proposal adopted June 2026; accelerates DC deployment, energy-efficient capacity, and public-sector cloud. Central to closing the ~5%-vs-75% compute gap with the US.
- Climate Neutral Data Centre Pact (industry self-regulation). Targets incl. a water-use cap of 400 ml/kWh by 2040 and PUE/renewable commitments.
- No EU-wide direct environmental tax on DCs exists yet. Member states diverge: Finland/Sweden removing preferential treatment; others still offering generous investment incentives.
8-EU. European electricity prices (Eurostat, H2 2025, non-household*)
Non-household = medium industrial band, 500–2,000 MWh/yr, incl. non-recoverable taxes. Large DCs negotiate below these via PPAs/wholesale, but the ranking is directionally indicative.
| Country | Non-household price (€/kWh) | Note |
|---|---|---|
| Finland | 0.0748 | lowest in EU; cold climate + nuclear/wind |
| Sweden | 0.0970 | 2nd lowest; hydro + wind |
| EU average | 0.1837 | — |
| Germany | 0.2264 | 3rd highest; Energiewende surcharges + grid fees |
| Cyprus | 0.2429 | island grid |
| Ireland | 0.2552 | highest in EU; grid-constrained |
Context: EU industrial electricity remains 2–4× higher than US (European Commission), a persistent competitiveness drag. Cheapest-power = Nordics; most expensive = Ireland/Germany/Belgium/Denmark.
9-EU. Country profiles
Ireland — from magnet to gatekeeper
- Market: ~97% of DCs in the Dublin region; DCs consumed ~22% of national electricity (2024), up from ~5% in 2015; EirGrid forecasts ~31% by 2034.
- Policy history: de-facto moratorium on new Dublin grid connections from 2021.
- Reset (Dec 2025): CRU Decision CRU/2025236 ended the blanket moratorium for applications after 12 Dec 2025. New terms for connections >1 MVA: must provide own dispatchable generation/storage matching import capacity (MIC), source 80% of annual demand from renewables (6-year glide path), and feed power back to the grid.
- Read: clarity restored, but capex-per-facility rises sharply.
Netherlands — hyperscale largely closed
- Policy: since 1 Jan 2024, no new hyperscale DCs (>10 ha AND ≥70 MW) except two designated zones — Eemshaven (Groningen) and Agriport (Noord-Holland).
- Reality: ~7 hyperscales still under construction on pre-existing permits; Microsoft advanced an Amsterdam project by splitting it into 3 separate permits. Amsterdam grid congestion now competes with housing (30,000 homes) and schools.
- Local bans spreading: Utrecht (2021), plus Leiden, Breda, Westland.
Germany — the strictest efficiency regime (EnEfG)
The Energy Efficiency Act (EnEfG, in force Jan 2024) implements the EED with binding DC rules: Renewable electricity ≥50% from 2024, 100% from 2027; PUE ≤1.2 for new DCs (commissioned ≥1 Jul 2026) within 2 years, existing ≤1.5 by Jul 2027, ≤1.3 by Jul 2030; waste-heat reuse (ERF) ≥10% (2026), 15% (2027), 20% (2028); mandatory energy management + annual reporting. 3rd-highest EU industrial power (€0.2264). Industry (GDA) warns it risks being a “data-center-prevention law.”
France — the low-carbon nuclear play
- Energy edge: 95.2% low-carbon electricity (2025, RTE); EDF ~€70/MWh nuclear price ceiling from 2026.
- Tax / incentives: TICFE reduced for DCs from €22.5 to €12/MWh; ~50% energy-tax reduction for efficiency-compliant DCs; 30% R&D tax credit (CIR); fast-track permits under France 2030.
- Market: Paris overtook Amsterdam in 2024 as Europe’s 3rd-largest market; IT load ~2,070 MW (2025) → ~4,550 MW by 2030.
- Investment (Feb 2025 AI Action Summit): ~€109–110bn pledged — UAE/MGX, Brookfield €20bn, Data4 €20bn, Fluidstack €10bn, Microsoft €4bn.
United Kingdom — AI Growth Zones
- Market: 520+ DCs, largest in Western Europe.
- Framework: AI Opportunities Action Plan → “Delivering AI Growth Zones” (Nov 2025). DCs designated Critical National Infrastructure; Infrastructure Planning Bill gives NSIP status. Claims it can cut time-to-power by up to 5 years, unlocking ~£100bn.
- Anchors: Microsoft $30bn (2025–28), Google £5bn, Equinix £3.9bn.
The Nordics — the sustainable-growth model
Cheap renewable power, free cooling (PUE <1.10), fast permitting, grid access up to 24 months faster than FLAP-D. Sweden: SEK 756M power-cost compensation (2024); DC electricity tax now ~4.0 c/kWh (ultra-low rate phasing out); Brookfield $10bn/750 MW. Norway: ~98% hydro; withdrew reduced electricity tax (2023) but offers real-estate relief; OpenAI Stargate Norway. Finland: ultra-low electricity tax €0.0005/kWh removed from 2026 (~40× rise), rattling ~€10bn of planned investment; offset by corporate tax cut to 18%. Iceland: 100% renewable.
Spain — permitting speed as the edge
“Positive silence” permitting: a 2022 Royal Decree simplified EIAs, reportedly halving timelines (the model the EU CADA is studying). Below-EU-avg power; strong solar. AI Factory: Barcelona (MareNostrum/BSC).
Others (brief)
Italy: Milan growth; AI Factory hosted. Poland/Czechia/Romania: emerging Tier-2; among the highest household prices in PPS terms. Overall EU trend: >half of forecast AI capacity growth expected in the Nordics and Tier-2/secondary markets as FLAP-D hits grid/land limits.
PART C — CROSS-CUTTING DIMENSIONS
10. Energy mix & cost — comparative snapshot
| Jurisdiction | Grid character | Indicative large-user power cost | Policy signal |
|---|---|---|---|
| US – Texas (ERCOT) | Gas + wind + solar, deregulated | ~8¢/kWh commercial | cheap, light-touch, water-stressed |
| US – Virginia (PJM) | Gas + nuclear + growing solar | ~10–12¢/kWh; capacity prices spiking | incentives under review |
| US – North Dakota | Wind + lignite + hydro | lowest US (~7¢ commercial) | cheap, small market |
| France | ~95% low-carbon (nuclear-led) | EDF ~€70/MWh nuclear ceiling | stable, decarbonized, subsidized |
| Nordics (SE/NO/FI/IS) | Hydro / wind / geothermal, 90–100% renewable | €0.03–0.10/kWh; PPAs common | cheapest + greenest; some tax breaks phasing out |
| Germany | Renewables + gas, Energiewende | ~€0.226/kWh (high) | strictest efficiency mandates |
| Ireland | Gas-heavy, wind growing | ~€0.255/kWh (highest EU) | conditional connection + 80% renewable rule |
| UK | Gas + wind + nuclear | high, but AI Growth Zones cut DC power cost | planning + grid acceleration |
| Spain | Solar-heavy, cheap gas | below EU avg | fast permitting |
Takeaway for siting: the winning combinations are (a) cheap + green + fast grid (Nordics), (b) stable low-carbon at scale (France), or (c) cheap + fast + light-touch but water/grid-exposed (Texas). The losing combination is expensive + grid-constrained + politically hostile (Ireland/Amsterdam/Frankfurt for new hyperscale).
11. Ecological impact — the three pressure points
- Electricity / carbon. DCs ~1.5–2% of global electricity; US DCs estimated 4–7% of national electricity (2025). Carbon depends entirely on grid mix. AI could add 24–44 Mt CO₂/yr by 2030.
- Water. Evaporative cooling consumes water. Mitigations: direct-to-chip and immersion liquid cooling, closed-loop, reclaimed/non-potable sourcing. EU Pact targets 400 ml/kWh by 2040.
- Land + heat + community. Housing/school competition (Amsterdam), farmland (Arizona/Georgia). Waste-heat reuse into district heating turns a liability into value — mandated in Germany.
12. Supply chain — the real 2025–2026 constraint
- Transformers: demand +119% (2019–25); HV lead times 2–4 years (up to 5). China ~60% of global production, order books full through 2027.
- Switchgear: lead times >60 weeks; HV substations 3–5 years.
- Grid connection: new HV transmission can take >a decade. Nearly half of US DCs planned for 2026 may be delayed/cancelled on equipment shortages.
- Response — “BYOP” (bring your own power): off-grid “energy islands” and on-site generation (gas, fuel cells, planned SMRs); plus US DPA Section 303 (April 2026).
13. Financing & public money — quick map
- US: tax expenditure (state exemptions) + federal bonus depreciation; DPA financing; little direct grant money.
- UK: National Wealth Fund, British Business Bank £25.6bn, UK Export Finance.
- EU: InvestAI (~€200bn target), €20bn gigafactory facility, EuroHPC €10bn.
- France: France 2030, BPI France, CIR, green loans, EDF price ceiling.
- Nordics: mostly market-led via PPAs + targeted tax relief.
14. What to watch (12–24 months)
- US: Virginia reform bills; spread of large-load “beneficiary-pays” tariffs; whether bonus depreciation survives to 2030.
- EU: final shape of CADA; gigafactory site selection; “special compute zones.”
- Everywhere: transformer/switchgear lead times; rise of on-site/nuclear (SMR) power; liquid cooling as water-policy compliance tool.
Sources (selected)
US tax/incentives: NAIOP; Tax Foundation; NCSL; CNBC (Jun 2025); MultiState (Feb 2026). US power: EIA; Columbia CGEP; IER; EESI. US water: Lincoln Institute; LBNL; EESI. EU framework: European Commission; Eurostat (H2 2025). Country sources for Ireland, Netherlands, Germany, France, UK, Nordics, Spain as cited inline. Supply chain: Wood Mackenzie; Transformer Magazine; pv-magazine. Market: JLL EMEA 2025; McKinsey; PwC; Epoch AI.
Figures are drawn from the above as of mid-2026; regulatory terms are actively evolving — verify against primary sources before relying on any specific threshold for a live decision.
PART D — THE POLITICS: WHO’S FOR, WHO’S AGAINST, AND WHETHER THE ARGUMENTS HOLD UP
D.0 How to read this part
Three ground rules: (1) this is not an exhaustive roster — it maps the named, sourced actors driving policy plus party-level patterns; (2) “for/against” is a spectrum: accelerate → accept with conditions → pause/moratorium → ban, and most of the fight is in the middle; (3) “factual or not” is applied only to empirical claims (§D.4), not to value judgments. A structural note: the backlash is bipartisan and cross-ideological — it unites figures who agree on little else (Sanders and DeSantis), while support also crosses party lines.
D.1 United States — federal level
Skeptics / restrict: Sen. Bernie Sanders (I-VT) called (Dec 2025) for a national moratorium and co-introduced the AI Data Center Moratorium Act of 2026 with Rep. Ocasio-Cortez (D-NY). Sens. Warren, Van Hollen, Blumenthal opened an investigation into DC energy use and bills. ~200 advocacy groups sought a federal moratorium. Skeptics on the right: Gov. Ron DeSantis (R-FL) proposed a local right to block DCs, arguing the grid can’t support the buildout. Accelerate / pro: the Trump administration (EO to curb state AI regulation; DPA to expand transformer supply); the Data Center Coalition and building-trades unions (IBEW). Town/county moratoria have passed since Aug 2025 in MO, IN, GA, NC and elsewhere.
D.2 United States — Virginia (the defining battleground)
Virginia matters because it’s the world’s largest market and because the 2026 fight was Democrat-vs-Democrat. Defend incentives: Gov. Spanberger, Speaker Don Scott, Approps. Chair Luke Torian (jobs ~31,500 created/maintained; honoring contracts). Curb/condition: Sen. Louise Lucas (curb the ~$1.6–1.9bn/yr exemption), Sen. Danica Roem (cancel outright), Del. Rip Sullivan & Sen. Creigh Deeds (condition on efficiency), Del. Josh Thomas (impact-study, signed into law). Republican angle: former Gov. Youngkin argued for local control and vetoed study bills in 2025. Non-party actors: Dominion Energy, Clean Virginia, the VEA, IBEW Local 26. Outcome: a compromise budget created Virginia’s first statewide tax on DC energy consumption plus a study commission — a “warning shot,” not a repeal.
D.3 Europe — party and government positions
Europe’s pattern: less “moratorium vs build,” more conditions, efficiency mandates, siting control — with Green parties the most consistent drivers of restrictions. Ireland: the Green Party presided over the 2021 moratorium era and lost 11 of 12 seats in Nov 2024 (cost-of-living backlash). Netherlands: Parliament passed a motion to halt large-scale construction; ChristenUnie, PvdD, GroenLinks-PvdA, SP critical; D66 defended DCs. Germany: the Greens drove the strict EnEfG; the GDA (industry) warns of a “prevention law.” France: Macron/Renaissance strongly pro; Les Écologistes and LFI the main critics. EU level: Greens/EFA push efficiency; the EPP/Renew/S&D bloc backs CADA. An Oct 2025 Savanta poll: 72% want DCs built only if paired with new renewables, 85% want mandatory environmental disclosure.
D.4 Are the arguments factual? (empirical claims only)
| Claim | Verdict | What the evidence shows |
|---|---|---|
| “Data centers raise electricity bills” | Contested nationally, real locally | No clean national correlation yet (IER, CGEP, LBNL); but PJM capacity prices spiked with ~$23bn attributable to DCs, and Virginia/PJM face real near-term pressure. |
| “They drain community water” | True, but local | Real in specific basins (rural GA, AZ); US direct ~17.4bn gal (2023) → 38–73bn by 2028; not a uniform national crisis. |
| “Very few permanent jobs” | True | ~15–100 permanent staff each; big employment is temporary construction. |
| “Tax breaks don’t pay for themselves” | Supported, with nuance | Virginia JLARC: “moderate economic benefit” that “does not pay for itself,” though better than most incentives. |
| “Not enough grid capacity” | Largely true near-term | 700+ GW of 2025 requests vs 477 GW total 2023 use; 2–5 yr equipment lead times. |
| “Slow down / pay fair share / oligarchs benefit” | Value judgment | Priority/values arguments, not empirical claims. |
| Pro: “unmitigated economic engine / no rate impact” | Overstated | Investment is real, but “unmitigated” ignores documented local water/grid/rate pressures. |
D.5 Honest synthesis
The empirical core is genuinely mixed and localized: national price effects aren’t proven, but specific grids (Virginia/PJM) and basins face real pressure. The most defensible position most systems are converging on is neither ban nor blank check but conditionality. The electorally risky lesson (Irish Greens 2024; Virginia’s intra-Democratic fight) is that cost-of-living now dominates.
PART E — STRESS REPORTS: IS THE GRID / BILLS / WATER / LAND ACTUALLY BREAKING?
Short answer: Yes for most mature markets — but the pattern is localized and grid-specific, not uniform. Markets under real strain are the mature hubs (Virginia, Dublin, Amsterdam, Frankfurt) and water-stressed newcomers (Aragón). A macro signal: EMEA capacity growth slumped ~11% in 2025, and ~40% of US projects are behind schedule. Rating key: High = documented crisis/pauses · Moderate = real pressure · Emerging = early/localized · Low = largely not a constraint.
E.1 Stress matrix (documented reports)
| Jurisdiction | Grid | Bills | Water | Land / permitting |
|---|---|---|---|---|
| US – Virginia | High — Dominion filed ~70 GW new load | High — rates +28% 2020–25; PJM ~$23bn | Moderate | High — new statewide DC energy tax; siting bills |
| US – Texas | Moderate — ERCOT absorbing load; BYOP rising | Emerging | High — 49→399bn gal; drought-exposed | Low–Moderate |
| US – Georgia/Ohio | Moderate — new tariffs shift cost to DCs | Moderate | Moderate — rural GA depletion | Moderate — local moratoria |
| Ireland | High — DCs ~22%; 2021 moratorium; ~€1bn emergency gas | High — highest EU price; poorest paid ~€209 extra | Emerging | High — 97% in Dublin; conditional rules |
| Netherlands | High — Amsterdam congestion; 30k homes/50 schools at risk | Moderate | Emerging | High — national hyperscale ban |
| Germany | High — grid “at its limits”; Frankfurt constrained | Moderate — 3rd-highest industrial price | Emerging | Moderate — strict EnEfG |
| UK | High — grid queue = “single biggest blocker” | Moderate | Emerging | Moderate→improving — CNI + NSIP |
| France | Low — nuclear surplus; land available | Low — EDF ~€70/MWh; prices fell ~12.5% | Emerging | Low–Moderate — fast permits |
| Spain (Aragón) | Moderate–High — 2,790 MW demand vs 402 MW plant; Apr 2025 blackout | Emerging | High — DC water +76–124% regional | Moderate — fast permits, rising opposition |
| Nordics | Low — spare capacity, 400 kV corridors | Low — cheapest EU power (Finland tax rise aside) | Low — cold, hydro | Low — fast permitting |
E.2 Evidence notes
Pan-Europe: IEA (Nov 2025): EU grid-connection waits 2–10 years (FLAP-D 7–10); congestion cost €4.3bn in 2024 (ACER). Dublin and Amsterdam paused projects; capital fleeing to secondary markets. US: 700+ GW of requests vs 477 GW total 2023 use; ~40% of projects behind schedule; Virginia the sharpest case. Ireland: the textbook case (22% of power, 2021 moratorium, €1bn emergency gas, poorest households ~€209 extra 2021–23). Netherlands: ~30,000 homes and 50 schools reported at risk over grid. Germany: the boom is “pushing the grid to its limits.” UK: grid connection is the “single biggest blocker.” France: the positive outlier (95%+ low-carbon, prices fell ~12.5% in 2025). Spain: grid capacity drew investment but water is the acute risk (Aragón +76–124%). Nordics: the counter-example on all four axes.
E.3 The signal that ties it together
The constraints are repricing and relocating the industry in real time: buildout slowing (EMEA −11%; ~40% US behind schedule); capital relocating to spare-capacity secondary markets; “bring your own power” going mainstream (reintroducing the emissions problem clean grids avoid); policy converging on “make DCs pay and prove efficiency.” Net read: the binding 2026 question is no longer “what’s the tax incentive?” but “can I get power and permits here in <5 years, and does this basin have water?”
PART F — ADDENDUM: VERIFIED UPDATES (as of 4 July 2026)
These items emerged around or just after the main report’s cutoff; each was checked against primary sources and reinforces the central thesis rather than contradicting it.
F.1 Virginia — the per-kWh data-center tax is now law (VERIFIED)
- Data Center Electricity Consumption Tax: $0.011/kWh on all electricity consumed at each Virginia DC, monthly. The first statewide per-kWh tax on DC consumption in the US.
- Effective 1 July 2026; sunsets 1 July 2028. First collection September 2026. Applies to utility-supplied and self-generated (behind-the-meter) power.
- Annual cap ~$600M; estimated ~$1.2bn over the biennium. The sales-and-use tax exemption was preserved — Virginia added a cost rather than removing a benefit. A 500 MW facility owes ≈$48M/year.
- New water rule: DEQ to define “Cooling Water Scarcity Areas” by 1 July 2027; in the Eastern Virginia Groundwater Management Area, new DCs filing air permits after 1 Jan 2027 must use air cooling, closed-loop, or 100% recycled/stormwater cooling.
F.2 Illinois & Ohio — incentive pauses (VERIFIED)
- Illinois: Gov. Pritzker (D) directed DCEO (5 June 2026) to pause processing all new Data Center Investment Program agreements from 1 July 2026; existing agreements honored. The pause was a direct consequence of the legislature failing to pass the POWER Act (SB4016/HB5513) by the 31 May 2026 deadline. Backlash fueled by hard grid-cost numbers: data-center demand raised costs by ~$13 billion over two PJM capacity auctions; the Union of Concerned Scientists projected a further ~$37 billion in Illinois over 25 years; ComEd attributed ~$12/month of added residential bills to rising supply costs.
- Ohio: Gov. DeWine (R) the prior week halted new DC sales-tax-exemption requests pending a study — a Republican parallel underscoring the bipartisan pattern.
F.3 EU — CADA is a proposal, not enacted law (CLARIFICATION)
The Cloud and AI Development Act was adopted by the Commission as a proposal on 3 June 2026 and entered the ordinary legislative procedure; it is not yet final law. CADA is explicitly a technological-sovereignty instrument — a uniform EU legal framework that reduces dependencies on non-EU cloud/AI providers and safeguards public-order/security interests, not merely to add capacity.
F.4 Smaller confirmations
- OBBBA: the bill restoring 100% bonus depreciation was signed 4 July 2025.
- Finland: the ultra-low Category II electricity-tax rate was removed effective 1 July 2026 (~40× rise).
- State reform wave: >25 states advancing DC-related measures (grid cost, tax, siting, transparency).
- Supply chain: no relief — HV transformer/substation lead times still 2–5 years; DPA Section 303 confirmed.
F.5 Items treated as reported, not confirmed
The exact Finland €47M revenue figure; a Henrico County VA schools anecdote; the precise CADA “COM(2026) 502” document number; specific Iowa cap terms and a “~38 states” count. Directionally consistent but check primary sources.
F.6 Net effect
Nothing overturns the main analysis. The Virginia tax is the concrete flagship of the “beneficiary-pays” turn; Illinois/Ohio are concrete incentive-pause instances; CADA’s status is a precision fix. The central 2026 thesis stands and is strengthened: power, permitting, and equipment — not tax rates — are the binding constraints.
PART G — ASIA-PACIFIC (including China & India)
Asia-Pacific is now the world’s fastest-growing data-center region — APAC DC electricity is projected to climb from ~320 TWh (2024) to ~780 TWh by 2030 (+165%), with only ~32% met by renewables. The defining feature: the binding constraint is power availability and permitting, and the political model is mostly state-directed/technocratic rather than the electoral backlash seen in the US/Europe.
G.0 Regional framework & patterns
- Incentives are dominant and state-led. Where the US/EU are adding conditions, most of Asia is still courting — tax holidays, electricity-duty waivers, land subsidies. Competition is often sub-national.
- The “capacity-allocation” model. Singapore, and increasingly Malaysia and China, ration growth: capacity is awarded to projects meeting efficiency/sustainability criteria.
- Constraint-driven geographic spillover. Singapore’s moratorium pushed capacity to Johor and Batam; China’s east-coast limits push workloads west; grid limits push Korea out of Seoul, Japan beyond Tokyo.
- Climate/water exposure is acute. Tropical heat (SEA), monsoon/drought (India), arid west/north (China), seismic/flood risk (Japan, Taiwan, HK).
G.1 China — the state-directed model (~25% of global DC consumption)
“East Data West Computing” (东数西算, EDWC). Launched 2021–22, builds 8 western computing hubs + 10 eastern data clusters, migrating non-realtime workloads west (Inner Mongolia, Gansu, Ningxia, Guizhou). PUE ≤1.25 by 2025; new hub DCs ≥80% renewable by 2030 (vs ~70% of DC power historically from coal). Green Electricity Certificates the only official proof since March 2025. Nov 2025: up to 50% electricity subsidy for DCs using domestic semiconductors. DC electricity ~130–200 TWh (2022–25) → 400–600 TWh by 2030. Reality check: EDWC underperformed — many western DCs sit idle; Beijing imposed ≥60% utilization rules and barred new large DCs in cities below 50% used. Fallback remains the coal-dominated grid.
G.2 India — the state-competition model (fastest-growing major market)
National: the Draft National Data Centre Policy 2025 proposes a 20-year tax holiday, infrastructure status, 100% electricity-duty exemption, single-window clearance. 100% FDI under the automatic route. State layer (the real driver): Maharashtra (up to 60% electricity-duty exemption for 15 yrs; AWS $8.2bn, Microsoft $3bn), Tamil Nadu (dedicated feeders, uninterrupted water), Karnataka (7% capital subsidy), Uttar Pradesh (25–50% land subsidies), Rajasthan (2025) (ZLD, wastewater recycling), Andhra Pradesh (seawater cooling). Power = ~65% of opex; open-access PPAs ₹4–7 vs ₹7–10 utility. Gaps (CEEW): water-governance deficit, no PUE/WUE standards.
G.3 Singapore — the capacity-allocation pioneer
- 2019 moratorium (DCs ~7% of national electricity); lifted in phases from July 2022 via DC-CFA.
- Green Data Centre Roadmap (2024): ≥300 MW contingent on green energy; PUE <1.3. DC-CFA2 (Dec 2025): ≥200 MW, ≥50% green power, PUE 1.25.
- Jurong Island 700 MW low-carbon park (Oct 2025). Energy imports up to 6 GW low-carbon by 2035. Construction cost US$14.53/watt (2nd-highest globally).
G.4 Malaysia / Johor — the constraint spillover, now selective
- Johor went from ~10 MW (2021) to ~1.3 GW (2024); SEA’s fastest-growing hub. DayOne $3.5bn; Amazon $6bn by 2037; heavy Chinese capital.
- Informal moratorium since ~mid-2024 (PM Anwar): a filter, not a withdrawal. Infrastructure-upgrade costs borne by developers. New dedicated DC water tariff; potable-water cooling ban forces reclamation.
G.5 Japan & South Korea — grid-constrained mature markets
Japan: IT capacity ~2.3 GW (2025) → ~3.7 GW (2030); DC boom to drive ~60% of power-demand growth; restarted 14 nuclear reactors. Grid connection 5–10 years is the core bottleneck; plus seismic risk and Tokyo’s top construction cost. South Korea: ~75% of capacity concentrated in Greater Seoul (66% self-sufficient); large DCs must pass a Power Grid Impact Assessment; permitting 2–3 years; $1.2 trillion national semiconductor/AI push.
G.6 The rest of the region (brief)
- Taiwan: TSMC anchor; Google 40 MW microgrid; seismic exposure.
- Indonesia: Microsoft $1.7bn; government mandates hyperscalers contribute to grid; high water stress.
- Thailand: Amazon $5bn, Google $1bn; DC power demand grew 400% (2020–24) vs ~8% generation; severe water stress.
- Vietnam: cheap land + streamlined approvals, but weekly power cuts in summer.
G.7 Energy mix & cost — Asia snapshot
| Market | Grid character | Indicative large-user power cost | Policy signal |
|---|---|---|---|
| China (west) | Wind/solar/hydro-rich | ~US$0.026/kWh | EDWC steers builds here; ≥80% renewable by 2030 |
| China (east) | Coal-heavy, congested | ~US$0.06/kWh | utilization caps |
| India | Coal-dominant + growing renewables | PPA ₹4–7; utility ₹7–10 | state incentives; draft tax holiday |
| Singapore | Gas ~95%, import-dependent | high | rationed capacity; ≥50% green |
| Malaysia/Johor | Gas + coal + hydro | cheaper than SG | selective approvals; developer-pays grid |
| Japan | Gas/coal + restarting nuclear | high | nuclear restarts; 5–10 yr grid queues |
| South Korea | Nuclear + gas + coal | moderate–high | grid-impact assessment |
| Thailand/Indonesia/Vietnam | Coal + gas + hydro | cheap but supply-constrained | tax holidays; grid/water risk |
G.8 Stress reports — Asia
| Market | Grid | Bills / cost | Water | Land / permitting |
|---|---|---|---|---|
| China | Moderate–High — congestion; outages; western idle capacity | Moderate | High — arid west/north | High (top-down) — utilization caps |
| India | Moderate–High — dedicated feeders needed | Emerging | High — Mumbai/Bengaluru/Chennai stressed | Moderate — fragmented policy |
| Singapore | High — Oracle abandoned 150 MW after power denial | High — 2nd-priciest build cost | Moderate–High — WUE limits | High — capacity rationed since 2019 |
| Malaysia/Johor | Moderate — ~40% utilization; 18-mo approvals | Moderate | High — potable-water ban | High — ~30% of applications rejected (2024) |
| Japan | High — 5–10 yr grid connection; seismic | Moderate | Emerging | Moderate — sites moving to regions |
| South Korea | High — Seoul 66% self-sufficient; grid assessment | Moderate | Emerging | Moderate–High — de-clustering push |
| Thailand/Indonesia/Vietnam | High — TH demand +400% vs +8% gen; VN summer cuts | Emerging | High — 11 of 28 DCs face medium-high→extreme stress by 2030 | Emerging–Moderate |
Regional signal: power availability is the #1 barrier to APAC project completion. Hard stops include Oracle abandoning 150 MW in Singapore and Microsoft building its own power plants in Indonesia — the Asian “bring your own power” move.
G.9 Politics & positions — how Asia differs
Asia’s DC politics is less electoral, more state-directed/technocratic — no analogue to the Sanders/DeSantis push. China: no opposition politics; the contest is intra-state (central government reining in local overbuilding). Singapore: top-down by design. Malaysia: PM Anwar and the Deputy Energy Minister the visible actors; NGO Sahabat Alam Malaysia the clearest civil-society critic; Johor’s state government gatekeeps. India: inter-state competition, not organized opposition; think tanks (CEEW) the main counterweight. Japan/Korea: the “opposition” is effectively the interconnection queue.
G.10 Supply chain (Asia-centric)
Asia both drives and relieves the crunch: China controls ~60% of global transformer production; Asian majors (Hyundai Electric, Toshiba, Doosan, Delta) are scaling T&D. TSMC (Taiwan) and Samsung/SK Hynix (Korea) anchor GPU/HBM. Microgrids, SMR offtakes, and on-site gas are the region’s workarounds for slow grids.
G.11 Verified refinements & 2026 updates
India — Budget 2026-27 tax holiday now enacted. In the Union Budget 2026-27 (1 Feb 2026), a tax holiday until 31 March 2047 for foreign companies serving global (non-Indian) customers from Indian DCs (Indian customers routed through a reseller entity), plus a 15% safe-harbour designed to simplify transfer-pricing compliance for related-party transactions. Investments: Google ~$15bn (AdaniConneX, Visakhapatnam); Microsoft ~$17.5bn (largest-ever Asia commitment); Microsoft + Amazon >$50bn combined (Dec 2025). Build cost ~$5M/MW (30–50% below London/Singapore); IT load ~1.35 GW → ~5–8 GW by 2030; India generates ~20% of the world’s data but hosts only ~5% of DC capacity. Nuclear Energy Mission: 100 GW by 2047. FDI caveat: Press Note 3 (2020) requires prior approval for land-border countries. Maharashtra duty figure flagged as unsettled (national draft 100% vs Maharashtra’s verified industrial-tariff ~40%-cheaper classification).
Reported, not confirmed: Japan ~$65bn by 2030; Korea 15–25% AI-DC tax credits; Singapore WUE ≤2.0; China GB 40879-2021 / RMB 1M reward / late-2025 FDI easing. Directionally consistent but check primary sources.
BIBLIOGRAPHY & NOTE ON METHOD
The full consolidated source list — government/regulatory filings, statistical agencies (EIA, Eurostat, IEA), law-firm regulatory analyses, industry/market research, journalism, NGO/advocacy research, and academic work — is maintained in the working reference file. Key primary/official anchors: EIA Electric Power Monthly; Eurostat electricity price statistics (H2 2025); European Commission (AI Continent, CADA proposal 3 June 2026, AI Factories); Ireland CRU Decision CRU/2025236; UK DSIT “Delivering AI Growth Zones”; Virginia HB 30 Item 3-5.24#1c; Illinois Governor’s Office / DCEO; IEA (Nov 2025 grid-constraint commentary and 2026 Energy & AI update); India Union Budget 2026-27 coverage (CNBC, Forbes India, Mondaq); China EDWC analyses (OIES, Carbon Brief, ASPI); Singapore IMDA Green DC Roadmap.
Where sources conflict (notably on whether data centers raise consumer electricity prices), the divergence is presented rather than resolved, and empirical claims are separated from value judgments (Part D). Advocacy and industry sources are used for documented positions/quotes, not as neutral arbiters. All URLs accessed mid-2026; live policy items (US state bills, CADA, Nordic tax changes, India/China rules) evolve — verify against primary sources before relying on any specific figure for a decision or citation.