Published September 12, 2026 — Washington, D.C. AI data center electricity demand is driving unprecedented new US power plant construction. AI data centers consumed approximately 4.4% of US electricity in 2024; projected 6.7-12% by 2028 per IEA, Lawrence Berkeley National Lab, and Goldman Sachs. 92% of new US electricity generation in 2025 came from solar + battery storage.
Data last verified September 12, 2026 from IEA Electricity 2026 report, Lawrence Berkeley National Lab 2024-2026 data center energy reports, Goldman Sachs research, EIA capacity additions data, and FERC interconnection queue reports.
Quick Answer
AI data centers are driving new US power plant construction at unprecedented scale. AI consumed 4.4% of US electricity in 2024; projected 6.7-12% by 2028 (IEA, LBNL, Goldman Sachs). 92% of 2025 new US generation was solar + battery (EIA). Top markets: Northern Virginia (2.6 GW operational), Texas, Phoenix, Columbus, Atlanta. Power constraints emerging. 100 MW data center electricity: $87M/year; 1 GW: $876M/year. Mitigations: nuclear restarts (Three Mile Island for Microsoft 2028), SMRs, new transmission ($30B pipeline), AI efficiency (IEA; Lawrence Berkeley National Lab; Goldman Sachs, 2026).
US data center electricity demand growth
| Year | Data center electricity (TWh) | Share of US electricity | Annual growth |
|---|---|---|---|
| 2018 | ~200 TWh | ~4.0% | Baseline |
| 2020 | ~210 TWh | ~4.1% | +2-3% |
| 2022 | ~230 TWh | ~4.5% | +5-7% |
| 2024 | ~176 TWh (AI only) | ~4.4% | +12-15% |
| 2026 (projected) | ~280 TWh (AI + traditional) | ~6.7% | +15-20% |
| 2028 (projected) | ~400-500 TWh | ~9-12% | +20-25% |
| 2030 (projected) | ~600-800 TWh | ~14-18% | +20%+ |
Source: Lawrence Berkeley National Lab (2024-2026); IEA Electricity 2026; Goldman Sachs data center research.
How the US is meeting AI demand
Solar + battery storage (92% of 2025 new generation)
The US is building solar and battery at unprecedented scale:
- Solar additions: 32 GW in 2024, 50+ GW projected 2025, 60+ GW projected 2026 (EIA).
- Battery storage additions: 10 GW in 2024, 18+ GW projected 2025, 25+ GW projected 2026.
- Solar + battery costs: $0.04-0.06/kWh levelized (utility-scale); falling due to learning curve and Chinese supply.
- Data center solar PPAs: 20-year fixed-price contracts at $0.04-0.06/kWh - very favorable for data center economics.
- Behind-the-meter solar + battery: data centers building dedicated solar + battery at the campus.
92% of new US electricity generation in 2025 was solar + battery (EIA). This is driven by: (1) cost competitiveness. (2) fast construction (12-18 months for solar, 6-12 months for battery). (3) corporate sustainability goals. (4) data center PPA demand (EIA, 2025-2026).
Natural gas (firm backup)
Natural gas provides firm backup and baseload where needed:
- New gas plants: 5-10 GW approved annually for 2025-2028 (EIA).
- Behind-the-meter gas: data centers building dedicated gas turbines or reciprocating engines.
- Gas + carbon capture: pilot projects for low-carbon firm power.
- Costs: $0.05-0.08/kWh levelized; lower than coal; competitive with nuclear.
Nuclear restarts and new builds
Nuclear is seeing renewed interest for AI data centers:
- Three Mile Island restart: Constellation Energy restarting Unit 1 for Microsoft (20-year PPA, 2028).
- Talen Energy - Amazon: Susquehanna nuclear plant powering AWS data center campus.
- Small Modular Reactors (SMRs): NuScale, TerraPower, X-energy, Holtec - first commercial deployments 2028-2030.
- Costs: $0.06-0.10/kWh levelized; long-term price stability attractive to data centers.
- Time to deploy: restarts 4-6 years; new SMRs 6-10 years; new large reactors 10+ years.
Grid expansion
Grid investment is accelerating to meet demand:
- FERC Order 2023 (2023): reformed interconnection queue process; faster approvals.
- FERC Order 2024 (2024): transmission planning and cost allocation reforms.
- Major transmission projects: $30B+ in pipeline across major US grid operators (PJM, MISO, ERCOT, CAISO).
- New HVDC lines: several 500+ kV HVDC projects in development.
Top US data center markets
| Market | Operational MW (2026) | Pipeline MW | Primary power source | Constraints |
|---|---|---|---|---|
| Northern Virginia | 2,600+ MW | 5,000+ MW | PJM grid (mixed) | Severe capacity shortfalls; transmission |
| Texas (Dallas, Austin) | 1,500+ MW | 4,000+ MW | ERCOT (gas, wind, solar) | Summer/winter reliability; interconnection wait |
| Phoenix, Arizona | 1,200+ MW | 3,500+ MW | SRP + APS (mixed) | Water stress; capacity constraints |
| Columbus, Ohio | 1,000+ MW | 2,500+ MW | AEP grid (mixed) | Transmission; coal plant retirements |
| Atlanta, Georgia | 800+ MW | 2,000+ MW | Georgia Power (nuclear, gas, solar) | Capacity tight; nuclear expansions |
| Dallas-Fort Worth | 700+ MW | 1,800+ MW | ERCOT | Reliability; cost allocation |
| Quincy, Washington | 600+ MW | 1,200+ MW | Hydro + nuclear | Limited; environmentally constrained |
| Hillsboro, Oregon | 500+ MW | 1,000+ MW | BPA (hydro, wind) | Power shortages; transmission |
Source: JLL; CBRE; Cushman & Wakefield data center reports (2025-2026).
Cost of powering AI data centers
| Data center size | Capacity (MW) | Annual electricity cost @ $0.10/kWh | Annual electricity cost @ $0.06/kWh (PPA) |
|---|---|---|---|
| Small edge facility | 10 MW | $8.76M | $5.26M |
| Mid-sized enterprise | 100 MW | $87.6M | $52.6M |
| Large hyperscaler campus | 1 GW (1,000 MW) | $876M | $526M |
| Massive AI training campus | 5 GW (5,000 MW) | $4.38B | $2.63B |
Source: EIA; utility data; data center industry analysis (2026).
For AI training campuses at 5 GW, electricity is one of the largest operating costs (along with chips, networking, cooling, real estate). At $0.08/kWh blended, a 5 GW campus at 90% capacity factor spends $3.15B/year on electricity - larger than most Fortune 500 companies' annual operating budgets (Goldman Sachs, 2026).
Environmental impact
Greenhouse gas emissions
US data center GHG emissions:
- 2024: ~105 million metric tons CO2e (~2.0% of US emissions).
- 2030 (projected): 200+ million metric tons CO2e (~3.5% of US emissions).
- Driver: electricity consumption growth and grid carbon intensity.
- Mitigation: renewable PPAs, behind-the-meter renewable, nuclear, efficiency improvements.
Water consumption
Data center cooling water use:
- Water usage effectiveness (WUE): 0.5-2.0 L/kWh (1.0 L/kWh = average; 0.5 L/kWh = efficient).
- 100 MW data center water use: 50-200 million gallons/day (depends on cooling tech, location).
- Water-stressed regions: Phoenix, Las Vegas, Los Angeles, Texas - data center water use is increasingly controversial.
- Mitigations: closed-loop cooling, air cooling (where climate allows), recycled water, waterless cooling (immersion).
Land use
Data center land requirements:
- 1 GW data center campus: 500-1,000+ acres (including data center buildings, substation, parking, buffer).
- 5 GW data center campus: 2,500-5,000+ acres.
- Including dedicated power plant: 10,000+ acres (solar farms, SMR sites).
AI efficiency improvements
AI efficiency is improving rapidly, partially offsetting demand growth:
| Efficiency metric | 2022 | 2026 | 2030 (projected) |
|---|---|---|---|
| Training compute efficiency (FLOPs/W) | Baseline | 4-5x better | 10-20x better |
| Inference efficiency (FLOPs/W) | Baseline | 3-4x better | 8-15x better |
| Chip (NVIDIA Blackwell vs Hopper) | Hopper | Blackwell (2-3x perf/W) | Next-gen (further 2-3x) |
| Cooling PUE | 1.4-1.6 | 1.1-1.2 (liquid cooling) | 1.05-1.10 (immersion) |
Source: NVIDIA; data center industry analysis; IEA (2022-2026).
Despite efficiency gains, total AI electricity demand continues to grow because total AI compute deployed is growing faster than efficiency improvements. AI usage, model size, and deployment scale are all growing exponentially (Lawrence Berkeley National Lab; IEA, 2026).
Grid constraints and solutions
Major US grid constraints for data center demand:
- Interconnection queue: average 4-5 year wait for grid connection in major data center markets (FERC).
- Transmission constraints: 7-10 year timeline for new transmission permits + build.
- Capacity shortfalls: Northern Virginia, Phoenix, Atlanta face near-term shortfalls.
- Reliability concerns: ERCOT (Texas) summer/winter reliability with growing data center demand.
Mitigations and solutions:
- Behind-the-meter generation: data centers building dedicated gas, solar, battery, nuclear.
- Demand response: data centers reducing load during peak periods (paid by utilities).
- AI efficiency: continue improving chips, cooling, algorithms.
- Regulatory reform: FERC Orders 2023, 2024 streamlining interconnection and transmission planning.
- New transmission: $30B+ in pipeline; multi-state HVDC projects.
- Site selection: data centers moving to less-constrained markets (rural Midwest, smaller metros).
FAQ
Why are data centers concentrated in Virginia, Texas, and Phoenix?
The concentration reflects: (1) Power availability - regions with surplus generation or new generation projects. (2) Tax incentives - Virginia, Texas, and several other states offer data center sales tax exemptions, property tax abatements, and other incentives. (3) Fiber infrastructure - long-haul fiber routes. (4) Land availability - large parcels (500+ acres) for hyperscaler campuses. (5) Workforce and water - though increasingly in tension. (6) Climate - cooler climates (Virginia, Pacific Northwest) reduce cooling costs. (7) Proximity to customers - serving East Coast, Texas, and West Coast demand. (JLL; CBRE; Cushman & Wakefield, 2026).
Will AI electricity demand cause blackouts?
Unlikely. The US grid has not had a major AI-related blackout in 2024-2026 despite demand growth. Blackout risk is concentrated in: (1) Winter Storm Uri-type events (Texas 2021) - ERCOT faces summer/winter reliability concerns but has added ~20 GW of solar and battery since 2021. (2) PJM capacity shortfalls - PJM added capacity auctions with high clearing prices in 2025-2026 to incentivize new generation. The risk is more about: higher electricity prices in capacity-constrained markets, slower data center growth in those markets, and behind-the-meter generation requirements (FERC; NERC; ERCOT; PJM, 2026).
Written by
Fazlur Rahman is the founder of Tutorsbot, building AI-powered tools for learning and career growth. He writes about applying AI in real products and the practi… Read more
Fazlur Rahman is the founder of Tutorsbot, building AI-powered tools for learning and career growth. He writes about applying AI in real products and the practical side of building an ed-tech startup.
