--- kind: sector slug: data-centers title: Data Centers pillar: energy-production status: live as_of: 2026-08-26 updated_at: 2026-08-26 canonical_url: https://moreenergycleanplanet.com/sectors/data-centers/ json_url: https://moreenergycleanplanet.com/sectors/data-centers.json license: CC BY 4.0 --- # Data Centers > A leverage point, not a climate solution by default. Large, creditworthy loads can finance the power system we need—if they pay for new supply, the grid to deliver it, and the costs they create. **Pillar:** [Energy Production](https://moreenergycleanplanet.com/sectors/energy-production) · **Status:** live · **As of:** 2026-08-26 ## Key claims - Berkeley Lab estimates U.S. data centers used 192 TWh in 2024 and models a 649 TWh reference case for 2030, with a 521–843 TWh range. _(United States Data Center Energy Usage Report: 2025 Update)_ - ERCOT tracked 438 GW of large-load requests in June 2026, but its adjusted 2030 forecast assigns data centers 22.175 GW of coincident peak; a request pool is not expected load. _(PUCT approves ERCOT large-load interconnection review)_ - If load arrives before new supply, the near-term marginal response can be more gas generation and a slower coal decline rather than automatic clean-energy additions. _(Fossil generation could rise with faster-than-expected growth in data center power demand)_ - Creditworthy data-center customers can make new clean-firm projects bankable; Google's agreement with Kairos covers up to 500 MW of advanced nuclear through 2035. _(Google signs advanced nuclear agreement with Kairos Power)_ - The climate-positive bargain requires customers to pay full system costs and contract for additional, deliverable clean supply, auditable flexibility, and local safeguards. ## Analysis A leverage point, not a climate solution by default. Large, creditworthy loads can finance the power system we need—if they pay for new supply, the grid to deliver it, and the costs they create. ### What it is: The largest new electricity customer in a generation. Data centers turn electricity into computation. The largest campuses now arrive as gigawatt-scale industrial loads, with the same appetite for power as a steel mill or a small city and a much shorter clock. Berkeley Lab estimates that U.S. data centers used **192 terawatt-hours in 2024**, an average draw of 21.9 gigawatts. Its 2030 reference case reaches **649 terawatt-hours**, or 74.1 gigawatts around the clock. The modeled range is 521 to 843 terawatt-hours. That is a 3.4-fold increase in six years. Data centers account for one-third of Berkeley Lab's projected U.S. load growth through 2030. At the reference case, they would use 11.8 percent of U.S. electricity. This is the new fact every utility plan, generation project, transmission line, turbine order, and clean-power contract must absorb. The opportunity is real. So is the risk. A large, creditworthy buyer can make a new reactor, geothermal field, solar-and-storage portfolio, or transmission project financeable. The same buyer can also call forth a new fleet of gas plants. The load creates leverage. The power contract decides what that leverage does. ### National scale: The next grid is arriving inside this one. The 2024 and 2030 numbers below come from the same Berkeley Lab model. They are not a national meter total, and the 2030 column is not contracted load. The model begins with the computing equipment expected to ship, then estimates its electricity use, cooling load, facility type, and location. Cryptocurrency mining is excluded. The central case adds 457 terawatt-hours per year by 2030, equivalent to 52.2 gigawatts of continuous demand. For scale, the entire U.S. grid generated 4,308.6 terawatt-hours in 2024. The 2030 data-center reference case alone equals 15.1 percent of that 2024 output. That is a scale comparison, not a forecast share; Berkeley Lab's proper 2030 share is 11.8 percent. **Figure — U.S. data-center electricity use: estimated 2024 use and modeled 2030 range.** Interactive chart and accessible data table: [/embed/data-centers-load-scale](/embed/data-centers-load-scale). ### State buildout: A queue is a claim, not a power plant. Texas had 438 gigawatts of large-load requests in ERCOT's June 2026 review, 89 percent tied to data centers. ERCOT's finalized 2030 forecast, after adjusting for timing and probability, assigns data centers 22.2 gigawatts of coincident summer peak. Those two figures answer different questions. One describes everyone asking for a place in line. The other describes the load the system currently expects to serve. The same discipline matters elsewhere. Arizona reports more than 2 gigawatts operating and more than 10.6 gigawatts planned, while APS carries a separate 19-gigawatt uncommitted queue. Utah has 2.6 gigawatts of facility capacity under construction, but roughly 1.7 gigawatts is described as behind the meter or off-grid. Dominion measured 3.1 gigawatts of coincident data-center demand in 2023 and forecasts 7.7 gigawatts in 2032. Georgia Power's approved OpenAI contract is 3.2 gigawatts; its 73.1-gigawatt development pipeline is not a forecast. The categories must stay attached to the numbers. **Figure — Operating load, committed projects, utility forecasts, and request pipelines across five leading markets.** Interactive chart and accessible data table: [/embed/data-centers-state-status](/embed/data-centers-state-status). ### What is being built: The power race is choosing the grid we will inherit. There is no honest grand total for the projects below. A megawatt of solar nameplate is not a megawatt of firm generation. A battery's power rating is not its energy duration. A contract for existing nuclear output is not a new reactor. A campus rating is not the power plant beside it. The useful question is narrower: what did the customer help make possible, and what is its present status? Selected named projects. Capacities use each source's native basis and must not be added. Customer / project Power Status What it means [Google . Mesa, Arizona](https://blog.google/company-news/outreach-and-initiatives/sustainability/google-carbon-free-data-center-arizona-/) \>430 MW new-to-grid wind, solar, and storage Operating New variable clean supply. Not firm 24/7 power. [Google . Fervo / NV Energy](https://blog.google/feed/nevada-clean-energy/) 115 MW enhanced geothermal Approved / development Additional firm clean supply if built. [Google . Kairos Power](https://blog.google/company-news/outreach-and-initiatives/sustainability/google-kairos-power-nuclear-energy-agreement/) Up to 500 MW advanced nuclear Development through 2035 Potential new 24/7 clean supply. Not operating. [Microsoft . Crane Clean Energy Center](https://investors.constellationenergy.com/static-files/45978381-4afe-42b5-83ee-f27339cc2d8d) 835 MW nuclear restart Contracted; 2028 target Additive relative to the 2019 shutdown; NRC review remains. [Amazon . X-energy / Energy Northwest](https://www.aboutamazon.com/news/sustainability/amazon-nuclear-small-modular-reactor-net-carbon-zero) 320 MW first phase; expandable to 960 MW Development Potential new firm clean supply on a long clock. [AWS . Talen / Susquehanna](https://sustainability.aboutamazon.com/stories/how-amazon-is-powering-the-next-era-of-carbon-free-energy) Up to 1.9 GW access to existing nuclear Contracted; campus operating Low-carbon preservation or reallocation, not new nameplate supply. [Meta . Entergy Louisiana](https://www.entergy.com/news/entergy-louisiana-receives-lpsc-approval-for-major-infrastructure-investments-to-support-metas-data-center-and-improve-reliability) About 2,262 MW gas; up to 1,500 MW solar Approved / procurement New supply. The near-term firm component is gas-led. [Microsoft . Pecos, Texas](https://blogs.microsoft.com/blog/2026/06/22/powering-the-next-wave-of-ai-expanding-capacity-with-our-new-datacenter-in-pecos/) About 2 GW campus; generation undisclosed Announced / development Initial behind-the-meter gas; later grid interconnection. [Oracle . Vantage Frontier, Texas](https://vantage-dc.com/news/vantage-data-centers-unveils-plans-for-frontier-a-25b-mega-campus-in-texas-to-meet-unprecedented-ai-demand/) 1.4 GW critical IT campus Under construction / development New onsite natural-gas microgrid. Campus load is not generation capacity. [xAI . Southaven, Mississippi](https://www.mlgw.com/images/content/files/pdf/new/xAI%202025%20Update.pdf) 1.2 GW permanent gas generation Under construction New firm gas, separate from Memphis grid service. ### The Fairchild Pattern: The Fairchild Pattern, now measured in gigawatts. Fairchild Semiconductor's early chips were too expensive for a mass market. NASA and the Department of Defense bought them anyway, because performance mattered more than price. Those first customers financed the factories, the yield improvements, and the accumulated know-how that made the next chip cheaper. Premium demand built a commodity industry. The hyperscalers have begun to play the same role in energy. Google is backing up to 500 megawatts from Kairos Power and 115 megawatts of enhanced geothermal through Fervo and NV Energy. Amazon is backing X-energy's first 320-megawatt Energy Northwest project and a broader development program. Microsoft contracted for the 835-megawatt restart of Crane. Meta's nuclear agreements reach across existing plants, uprates, and new reactors. Deep-pocketed customers are converting speculative engineering into projects that can raise capital and place equipment orders. That is the hopeful case for the buildout. It is also a standard we can test. An anchor customer matters when its purchase creates a plant, restarts one, extends one that would otherwise close, or advances a technology down its cost curve. Buying attributes from power that already exists is useful accounting. It does not perform the same industrial work. ### Thesis role: A leverage point, not a climate solution by default. When new supply lags, the grid reaches first for the plants it already has. EIA's 2026 high-load sensitivity found that faster data-center growth could add 123 terawatt-hours of U.S. gas generation between 2025 and 2027, compared with 29 terawatt-hours in its baseline, while slowing the decline in coal generation. In ERCOT, the modeled 2027 wholesale price rose 79 percent. That is the default path when the load arrives before the power. The better path uses the customer's balance sheet to build ahead: new clean generation, firm capacity, storage, transmission, and demand flexibility in one package. The customer pays the full cost. The power is deliverable to the load. Clean-energy claims match time and place, not merely annual totals. Flexible work moves when the grid is tight and proves that movement at the meter. Local water, air, noise, land, and community impacts remain real costs, not footnotes. Data centers will not solve the climate crisis by existing. They can finance some of the machinery that will. The load is coming. The fuel is a choice. ### Rules that make the bargain work: Make growth pay for growth. Utilities and regulators are writing the bargain now. Large-load tariffs can require site control, deposits, minimum bills, long contracts, collateral, direct payment for grid upgrades, and exit fees. Those terms keep a speculative request out of the forecast and keep a departing customer from leaving other ratepayers with the bill. A serious bring-your-own-power offer must also survive five tests: the generation is additional; it can physically reach the load; clean supply is matched when and where the load runs; backup and standby service are fully priced; and any promised flexibility is contracted, metered, and auditable. Expedite projects that meet the tests. Price or reject the ones that do not. The prize is larger than one industry. A stronger grid and a new generation fleet can serve factories, homes, electric vehicles, and carbon removal long after today's chips are obsolete. That is how an AI buildout becomes public infrastructure instead of a private claim on a scarce system. ## Data | indicator | value | goal (year) | as of | source | |---|---|---|---|---| | U.S. data-center electricity use, historical modeled estimate | 192 TWh/yr ⚠ stale | — | 2024-12-31 | Energy and AI | | U.S. data-center electricity use, 2030 reference case | 649 TWh/yr ⚠ stale | 649 TWh/yr (2030) | 2024-12-01 | 2024 United States Data Center Energy Usage Report | | U.S. data-center electricity use, 2030 low case | 521 TWh/yr | 521 TWh/yr (2030) | 2026-06-01 | United States Data Center Energy Usage Report: 2025 Update | | U.S. data-center electricity use, 2030 high case | 843 TWh/yr | 843 TWh/yr (2030) | 2026-06-01 | United States Data Center Energy Usage Report: 2025 Update | | U.S. net electricity generation, present-grid scale reference | 4,308.608 TWh/yr ⚠ stale | — | 2024-12-31 | United States Electricity Profile 2024 | | ERCOT observed energized large-load projects | 5.778 GW | — | 2026-03-26 | Large Load Update to Texas House State Affairs | | ERCOT large loads approved to energize | 3.241 GW | — | 2026-03-26 | Large Load Update to Texas House State Affairs | | ERCOT data-center contribution to adjusted 2030 coincident peak forecast | 22.175 GW ⚠ stale | 22.175 GW (2030) | 2025-04-01 | 2025 Long-Term Load Forecast | | ERCOT tracked large-load requests | 438 GW | — | 2026-06-18 | PUCT approves ERCOT large-load interconnection review | | Arizona operating data-center facility capacity | 2 GW | — | 2026-04-20 | ACC Large Load / Data Center Workshop Highlights | | TEP contracted data-center load | 0.286 GW | — | 2026-04-20 | ACC Large Load / Data Center Workshop Highlights | | APS total-system peak forecast | 12 GW | 12 GW (2035) | 2026-06-30 | Data Centers: Protecting Customers While Planning for Big Energy Needs | | APS uncommitted extra-large-load requests | 19 GW | — | 2026-04-20 | Data Centers: Protecting Customers While Planning for Big Energy Needs | | Utah operational data-center facility and IT capacity | 0.92 GW | — | 2026-08-01 | Data Centers in Utah | | Utah data-center facility capacity under construction | 2.602 GW | — | 2026-08-01 | Data Centers in Utah | | Rocky Mountain Power large-load requests evaluated in six months | 3 GW | — | 2026-08-01 | Data Centers in Utah | | Dominion Virginia metered coincident data-center demand | 3.127 GW ⚠ stale | — | 2023-12-31 | 2024 Integrated Resource Plan | | Dominion Virginia data-center load under electric service agreements | 8.012 GW ⚠ stale | — | 2024-12-31 | 2024 Integrated Resource Plan | | Dominion Virginia data-center coincident demand forecast | 7.686 GW ⚠ stale | 7.686 GW (2032) | 2024-06-30 | 2024 Integrated Resource Plan | | Georgia Power contracted OpenAI load | 3.2 GW | — | 2026-08-26 | OpenAI contract approved as part of Georgia Power resource portfolio | | Georgia Power total-system winter peak forecast | 24.5 GW | 24.5 GW (2030) | 2026-06-30 | OpenAI contract approved as part of Georgia Power resource portfolio | | Georgia Power large-load economic-development pipeline | 73.1 GW | — | 2026-08-26 | OpenAI contract approved as part of Georgia Power resource portfolio | ## Companies - **[Amazon Web Services](https://moreenergycleanplanet.com/ecosystem#company/amazon-web-services)** (commercial) — Largest US cloud + AI-infrastructure operator (subsidiary of Amazon NASDAQ:AMZN). 2026 capex >$125B (Morningstar 2025). Talen / Vistra nuclear PPAs anchor 24/7 carbon-free-energy posture; multi-GW data-center campuses across PJM + ERCOT + SPP. - **[Apple](https://moreenergycleanplanet.com/ecosystem#company/apple)** (commercial) — Consumer-device and cloud operator with a material U.S. data-center footprint and long-running renewable-electricity procurement, but a smaller causal role in the current AI power buildout than the leading cloud and model companies. - **[Bloom Energy](https://moreenergycleanplanet.com/ecosystem#company/bloom-energy)** (public) — US solid-oxide fuel-cell OEM (NYSE: BE). BTM on-site-power deployments at AWS / Equinix / Oracle / Walmart at hyperscaler + colo scale. Hydrogen-ready stack; multi-MW deployments. - **[Constellation Energy](https://moreenergycleanplanet.com/ecosystem#company/constellation-energy)** (commercial) _(primary: coal)_ — Largest US nuclear-generation operator (NYSE: CEG). 21 reactors / ~33 GW. Three Mile Island Unit 1 restart agreed with Microsoft (Sep 2024, 20-year PPA, targeted 2028 ops). Coal divestiture complete pre-spinoff. Primary sector: nuclear-fission; cross-tagged here for managed-decline-of-coal context. - **[CoreWeave](https://moreenergycleanplanet.com/ecosystem#company/coreweave)** (scaling) — GPU-cloud operator leasing and developing high-density AI capacity across U.S. markets; a material demand counterparty whose public power disclosures remain less complete than the largest hyperscalers. - **[Crusoe Energy](https://moreenergycleanplanet.com/ecosystem#company/crusoe-energy)** (scaling) — AI-data-center developer + BTM-generation operator. Originally stranded-gas-to-Bitcoin Digital Flare Mitigation; now AI-infrastructure-as-a-service with Oracle anchor + Stargate JV. Multiple GW-scale projects in pipeline. Status: scaling-to-hyperscaler. - **[CyrusOne](https://moreenergycleanplanet.com/ecosystem#company/cyrusone)** (commercial) — Data-center developer and operator serving hyperscale and enterprise customers; landlord capacity is tracked separately from tenant demand to avoid double counting. - **[Digital Realty Trust](https://moreenergycleanplanet.com/ecosystem#company/digital-realty)** (commercial) — US data-center REIT (NYSE: DLR). ~300 facilities globally; hyperscaler build-to-suit + colocation. 100% renewable-electricity for US colo footprint; long-term green-bond financing. - **[Entergy Louisiana](https://moreenergycleanplanet.com/ecosystem#company/entergy-louisiana)** (commercial) _(primary: the-grid)_ — Regulated utility building the initial generation and grid package for Meta Hyperion, including about 2,262 MW of gas generation and procurement for up to 1,500 MW of solar. - **[Equinix](https://moreenergycleanplanet.com/ecosystem#company/equinix)** (commercial) — Largest US carrier-neutral colocation operator. NASDAQ: EQIX. ~260 data centers in 70 markets globally. 100% renewable-electricity commitment by 2030; PPAs across US + Europe; Bloom Energy fuel-cell deployments at multiple sites. - **[Fervo Energy](https://moreenergycleanplanet.com/ecosystem#company/fervo-energy)** (public) _(primary: geothermal)_ — Enhanced geothermal developer using horizontal drilling and AI-enabled subsurface analytics to deliver 24/7 carbon-free baseload power. - **[Google (Alphabet)](https://moreenergycleanplanet.com/ecosystem#company/google)** (commercial) — Google Cloud + Search + Workspace + AI hyperscaler (subsidiary of Alphabet NASDAQ:GOOGL). 2026 capex ~$75B (Morningstar 2025). 24/7 carbon-free-energy commitment (originator of the framework); Fervo geothermal PPA; multi-GW data-center campuses across US. - **[Kairos Power](https://moreenergycleanplanet.com/ecosystem#company/kairos-power)** (growth) _(primary: nuclear-fission)_ — Developer of fluoride salt-cooled high-temperature pebble-bed nuclear reactors using TRISO fuel, with the Hermes demonstration reactor under construction in Oak Ridge and a 500 MW master agreement with Google. - **[Mainspring Energy](https://moreenergycleanplanet.com/ecosystem#company/mainspring-energy)** (commercial) — US BTM on-site-power OEM; proprietary linear generator runs on natural gas, hydrogen, biogas, or ammonia. Deployments at hyperscaler + colo sites (AWS, Walmart, Lineage); positioned as fast-deploy bridge generation pending grid energization. - **[Meta Platforms](https://moreenergycleanplanet.com/ecosystem#company/meta)** (commercial) — Hyperscaler + AI-infrastructure operator (Facebook, Instagram, WhatsApp, AI). NASDAQ: META. 2026 capex ~$60-65B (Morningstar 2025). 100% renewable-energy match commitment achieved 2020; net-zero value-chain by 2030; multi-GW data-center campuses across SE US. - **[Microsoft](https://moreenergycleanplanet.com/ecosystem#company/microsoft)** (commercial) — Hyperscaler + AI-infrastructure operator. NASDAQ: MSFT. 2026 capex ~$90B+ (Morningstar 2025). Three Mile Island Unit 1 restart PPA with Constellation (20-year, Sep 2024); Helion fusion PPA (2028 target); 100% renewable-energy target by 2025; net-zero by 2030. - **[Oklo](https://moreenergycleanplanet.com/ecosystem#company/oklo)** (public) _(primary: nuclear-fission)_ — Public advanced fission developer building the Aurora Powerhouse, a 15-75 MW liquid-metal-cooled fast reactor with a ~14 GW non-binding pipeline targeting AI data centers, with Atomic Alchemy subsidiary targeting first criticality by July 4 2026. - **[OpenAI / Stargate](https://moreenergycleanplanet.com/ecosystem#company/openai-stargate)** (scaling) — AI developer and Stargate demand counterparty behind multiple gigawatt-scale U.S. campuses, including an approved 3.2 GW Georgia Power service contract. - **[Oracle Cloud Infrastructure](https://moreenergycleanplanet.com/ecosystem#company/oracle-cloud)** (commercial) — Cloud + AI-infrastructure operator (subsidiary of Oracle NYSE:ORCL). 2026 capex ramping (~$25B+ AI infrastructure spend); anchor tenant in Crusoe AI-data-center pipeline (Stargate JV with OpenAI + SoftBank). Aggressive nuclear + behind-the-meter strategy. - **[QTS Data Centers / Blackstone](https://moreenergycleanplanet.com/ecosystem#company/qts-data-centers)** (commercial) — Large U.S. data-center developer and operator owned by Blackstone, serving hyperscale and enterprise customers across major power markets. - **[Sage Geosystems](https://moreenergycleanplanet.com/ecosystem#company/sage-geosystems)** (foak) _(primary: geothermal)_ — US enhanced + geopressured-geothermal developer. Geopressured Geothermal System (GGS) couples energy generation with subsurface storage. San Miguel TX pilot operational. Meta data-center PPA signed Aug 2024. - **[Schneider Electric (Data Centers)](https://moreenergycleanplanet.com/ecosystem#company/schneider-electric-data-centers)** (commercial) — Global power-management + cooling OEM (Euronext: SU). Schneider Electric's data-center BU is the largest single market; APC UPS line + EcoStruxure microgrid software + Galaxy power-distribution. - **[STACK Infrastructure](https://moreenergycleanplanet.com/ecosystem#company/stack-infrastructure)** (commercial) — Hyperscale data-center developer and operator with large U.S. campus portfolios; landlord capacity may house demand disclosed by a tenant and is not additive to it. - **[Submer](https://moreenergycleanplanet.com/ecosystem#company/submer)** (commercial) — Spanish immersion-cooling OEM for high-density GPU + AI racks. SmartPod + SmartPod GO products at hyperscaler + colo deployments. Backed by Mubadala, Cisco Ventures, M&G. - **[Talen Energy](https://moreenergycleanplanet.com/ecosystem#company/talen-energy)** (commercial) _(primary: coal)_ — US IPP. Divested entire coal generation portfolio 2023 (Brunner Island, Montour, etc.); now owns Susquehanna nuclear (2.5 GW boiling-water reactor). $650M PPA to AWS for Cumulus data-center campus. Emerged from Chapter 11 May 2023 with clean balance sheet. The cleanest single-name coal-to-nuclear-IPP pivot on record. - **[TerraPower](https://moreenergycleanplanet.com/ecosystem#company/terrapower)** (foak) _(primary: nuclear-fission)_ — US advanced-reactor developer (Bill Gates + Khosla Ventures backed). Natrium sodium-cooled fast reactor with integrated molten-salt storage; 345 MWe + 500 MWe peaking. Kemmerer WY first-of-a-kind under construction since 2024 (non-nuclear systems first; targeting 2030 ops). DOE ARDP $2B cost-share. - **[Vantage Data Centers](https://moreenergycleanplanet.com/ecosystem#company/vantage-data-centers)** (commercial) — Hyperscale campus developer and operator; its 1.4 GW Frontier campus in Texas pairs Oracle demand with a large onsite natural-gas microgrid. - **[Vertiv](https://moreenergycleanplanet.com/ecosystem#company/vertiv)** (commercial) — Data-center power + cooling OEM (NYSE: VRT). Cold-plate + immersion liquid-cooling for GPU racks (the H100/B200 generation); modular UPS + power-distribution; the BoS exemplar for high-density AI infrastructure. - **[Vistra Energy](https://moreenergycleanplanet.com/ecosystem#company/vistra-energy)** (commercial) _(primary: coal)_ — US IPP (NYSE: VST). Owns Comanche Peak nuclear (2.4 GW). Coal retirements: Vermilion + Joppa + Edwards (IL) retired; published 2030 coal exit. AWS PPA at Comanche Peak (2024). Energy Harbor acquisition (Mar 2024) added Davis-Besse + Perry + Beaver Valley nuclear (Sustainable Energy segment). - **[VoltaGrid](https://moreenergycleanplanet.com/ecosystem#company/voltagrid)** (scaling) _(primary: natural-gas)_ — Onsite natural-gas generation and microgrid provider serving fast-to-power industrial and AI data-center projects. - **[X-energy](https://moreenergycleanplanet.com/ecosystem#company/x-energy)** (public) _(primary: nuclear-fission)_ — Developer of high-temperature gas-cooled small modular reactors using TRISO ceramic fuel pebbles, with first commercial deployment at Dow's Seadrift Texas petrochemical site under DOE's Advanced Reactor Demonstration Program. - **[xAI / SpaceXAI](https://moreenergycleanplanet.com/ecosystem#company/xai-spacexai)** (scaling) — AI infrastructure operator using grid service and dedicated onsite generation; its Southaven project is a clear speed-to-power case built around permanent natural-gas turbines. ## FAQ ### How much electricity do U.S. data centers use? Berkeley Lab estimates **192 TWh in 2024**, equal to 21.9 GW of average demand. Its 2030 reference case is **649 TWh**, or 74.1 GW average and 11.8% of projected U.S. electricity. The modeled 2030 range is **521–843 TWh**. These are modeled estimates, not a national meter total or contracted load. _(United States Data Center Energy Usage Report: 2025 Update)_ ### Does Texas really have 438 GW of data centers coming? No. ERCOT's **438 GW** figure was a June 2026 pool of large-load requests, 89% associated with data centers. Its finalized, probability-adjusted 2030 forecast assigns data centers **22.175 GW** of coincident summer peak. Requests include projects that move, duplicate sites, miss dates, or withdraw. _(PUCT approves ERCOT large-load interconnection review)_ ### Will data centers run on clean energy? Some will help build new clean supply; others are already pulling new gas plants and onsite turbines into the market. The project, fuel, status, grid relation, and additivity matter more than an annual renewable-energy claim. Demand creates leverage. The power contract decides what it builds. _(Fossil generation could rise with faster-than-expected growth in data center power demand)_ ### How can the data-center buildout help with climate? Large, creditworthy customers can serve as anchor buyers for first-of-a-kind and grid-scale projects. Google / Kairos, Amazon / X-energy, Microsoft / Crane, and Google / Fervo show the mechanism. The climate value appears when the contract brings forward additional, deliverable clean power rather than merely reallocating existing attributes. _(Google signs advanced nuclear agreement with Kairos Power)_ ### Who should pay for the new power and grid infrastructure? The customer creating the load should pay the full cost of generation, transmission, interconnection, standby service, and any stranded assets. Minimum bills, collateral, direct upgrade charges, staged commitments, and exit fees keep speculative projects out of utility forecasts and protect other ratepayers. ## Charts (embeddable) - **The next grid is arriving inside this one.** — https://moreenergycleanplanet.com/embed/data-centers-load-scale (https://moreenergycleanplanet.com/embed/data-centers-load-scale.svg · https://moreenergycleanplanet.com/embed/data-centers-load-scale.png) - **A request is not a forecast.** — https://moreenergycleanplanet.com/embed/data-centers-state-status (https://moreenergycleanplanet.com/embed/data-centers-state-status.svg · https://moreenergycleanplanet.com/embed/data-centers-state-status.png) ## Sources - **iea-energy-ai-2025** — [Energy and AI](https://www.iea.org/reports/energy-and-ai) _(primary, report)_ - **lbnl-dc-energy-2024** — [2024 United States Data Center Energy Usage Report](https://eta-publications.lbl.gov/sites/default/files/2024-12/lbnl-2024-united-states-data-center-energy-usage-report_1.pdf) _(primary, report)_ - **lbnl-dc-energy-2025-update** — [United States Data Center Energy Usage Report: 2025 Update](https://escholarship.org/content/qt33m6w3x0/qt33m6w3x0.pdf) _(primary)_ - **eia-us-electricity-profile-2024** — [United States Electricity Profile 2024](https://www.eia.gov/electricity/state/unitedstates/) _(primary)_ - **ercot-large-load-apr-2026** — [Large Load Update to Texas House State Affairs](https://www.ercot.com/files/docs/2026/04/09/ERCOTLargeLoadUpdate-April9HouseStateAffairsHearing.pdf) _(primary)_ - **ercot-long-term-load-forecast-2025** — [2025 Long-Term Load Forecast](https://www.ercot.com/files/docs/2025/04/08/ERCOT-2025-Long-Term-Load-Forecast-Report.pdf) _(primary)_ - **ercot-batch-zero-jun-2026** — [PUCT approves ERCOT large-load interconnection review](https://www.ercot.com/news/release/06182026-puct-approves-ercots) _(primary)_ - **azcc-data-center-workshop-2026** — [ACC Large Load / Data Center Workshop Highlights](https://azcc.gov/kevin-thompson/news/2026/04/20/acc-large-load-data-center-workshop-highlights) _(primary)_ - **aps-data-centers-2026** — [Data Centers: Protecting Customers While Planning for Big Energy Needs](https://www.aps.com/en/Utility/Company-wide-Events-and-Messages/Data_Centers_How_Were_Protecting_Customers_While_Planning-for_Big_Energy_Needs) _(primary)_ - **utah-data-centers-aug-2026** — [Data Centers in Utah](https://le.utah.gov/interim/2026/pdf/PB_DataCenters.pdf) _(primary)_ - **dominion-virginia-irp-2024** — [2024 Integrated Resource Plan](https://www.dominionenergy.com/-/media/content/about/our-company/irp/pdfs/2024-irp-w_o-appendices.pdf) _(primary)_ - **georgia-power-openai-2026** — [OpenAI contract approved as part of Georgia Power resource portfolio](https://www.georgiapower.com/news-hub/press-releases/contract-openai-approved-part-portfolio-delivering-950-million-annual-savings.html) _(primary)_ - **eia-data-center-high-load-2026** — [Fossil generation could rise with faster-than-expected growth in data center power demand](https://www.eia.gov/todayinenergy/detail.php?id=67344) _(primary)_ - **google-kairos-2024** — [Google signs advanced nuclear agreement with Kairos Power](https://blog.google/company-news/outreach-and-initiatives/sustainability/google-kairos-power-nuclear-energy-agreement/) _(self-reported)_ ## Related sectors [Bioenergy](https://moreenergycleanplanet.com/sectors/bioenergy.md) · [Coal](https://moreenergycleanplanet.com/sectors/coal.md) · [Fusion](https://moreenergycleanplanet.com/sectors/fusion.md) · [Geothermal](https://moreenergycleanplanet.com/sectors/geothermal.md) · [Hydro](https://moreenergycleanplanet.com/sectors/hydro.md) · [Natural Gas](https://moreenergycleanplanet.com/sectors/natural-gas.md) · [Nuclear Fission](https://moreenergycleanplanet.com/sectors/nuclear-fission.md) · [Oil](https://moreenergycleanplanet.com/sectors/oil.md) · [Solar](https://moreenergycleanplanet.com/sectors/solar.md) · [Energy Storage](https://moreenergycleanplanet.com/sectors/storage.md) · [The Grid](https://moreenergycleanplanet.com/sectors/the-grid.md) · [Wind](https://moreenergycleanplanet.com/sectors/wind.md) ## Machine access - Structured data: https://moreenergycleanplanet.com/sectors/data-centers.json - Live data via MCP: https://mcp.moreenergycleanplanet.com/mcp — tools: get_sector, get_cost_curve, list_indicators, get_sources - License: CC BY 4.0 ```json { "schema_version": "1.2", "kind": "sector", "slug": "data-centers", "title": "Data Centers", "pillar": { "slug": "energy-production", "title": "Energy Production" }, "status": "live", "theme_color": "#FFB627", "summary": "A leverage point, not a climate solution by default. Large, creditworthy loads can finance the power system we need—if they pay for new supply, the grid to deliver it, and the costs they create.", "analysis": "## What it is: The largest new electricity customer in a generation.\n\nData centers turn electricity into computation. The largest campuses now arrive as gigawatt-scale industrial loads, with the same appetite for power as a steel mill or a small city and a much shorter clock. Berkeley Lab estimates that U.S. data centers used **192 terawatt-hours in 2024**, an average draw of 21.9 gigawatts. Its 2030 reference case reaches **649 terawatt-hours**, or 74.1 gigawatts around the clock. The modeled range is 521 to 843 terawatt-hours.\n\nThat is a 3.4-fold increase in six years. Data centers account for one-third of Berkeley Lab's projected U.S. load growth through 2030. At the reference case, they would use 11.8 percent of U.S. electricity. This is the new fact every utility plan, generation project, transmission line, turbine order, and clean-power contract must absorb.\n\nThe opportunity is real. So is the risk. A large, creditworthy buyer can make a new reactor, geothermal field, solar-and-storage portfolio, or transmission project financeable. The same buyer can also call forth a new fleet of gas plants. The load creates leverage. The power contract decides what that leverage does.\n\n## National scale: The next grid is arriving inside this one.\n\nThe 2024 and 2030 numbers below come from the same Berkeley Lab model. They are not a national meter total, and the 2030 column is not contracted load. The model begins with the computing equipment expected to ship, then estimates its electricity use, cooling load, facility type, and location. Cryptocurrency mining is excluded.\n\nThe central case adds 457 terawatt-hours per year by 2030, equivalent to 52.2 gigawatts of continuous demand. For scale, the entire U.S. grid generated 4,308.6 terawatt-hours in 2024. The 2030 data-center reference case alone equals 15.1 percent of that 2024 output. That is a scale comparison, not a forecast share; Berkeley Lab's proper 2030 share is 11.8 percent.\n\n**Figure — U.S. data-center electricity use: estimated 2024 use and modeled 2030 range.** Interactive chart and accessible data table: [/embed/data-centers-load-scale](/embed/data-centers-load-scale).\n\n## State buildout: A queue is a claim, not a power plant.\n\nTexas had 438 gigawatts of large-load requests in ERCOT's June 2026 review, 89 percent tied to data centers. ERCOT's finalized 2030 forecast, after adjusting for timing and probability, assigns data centers 22.2 gigawatts of coincident summer peak. Those two figures answer different questions. One describes everyone asking for a place in line. The other describes the load the system currently expects to serve.\n\nThe same discipline matters elsewhere. Arizona reports more than 2 gigawatts operating and more than 10.6 gigawatts planned, while APS carries a separate 19-gigawatt uncommitted queue. Utah has 2.6 gigawatts of facility capacity under construction, but roughly 1.7 gigawatts is described as behind the meter or off-grid. Dominion measured 3.1 gigawatts of coincident data-center demand in 2023 and forecasts 7.7 gigawatts in 2032. Georgia Power's approved OpenAI contract is 3.2 gigawatts; its 73.1-gigawatt development pipeline is not a forecast. The categories must stay attached to the numbers.\n\n**Figure — Operating load, committed projects, utility forecasts, and request pipelines across five leading markets.** Interactive chart and accessible data table: [/embed/data-centers-state-status](/embed/data-centers-state-status).\n\n## What is being built: The power race is choosing the grid we will inherit.\n\nThere is no honest grand total for the projects below. A megawatt of solar nameplate is not a megawatt of firm generation. A battery's power rating is not its energy duration. A contract for existing nuclear output is not a new reactor. A campus rating is not the power plant beside it. The useful question is narrower: what did the customer help make possible, and what is its present status?\n\nSelected named projects. Capacities use each source's native basis and must not be added.\n\nCustomer / project\n\nPower\n\nStatus\n\nWhat it means\n\n[Google . Mesa, Arizona](https://blog.google/company-news/outreach-and-initiatives/sustainability/google-carbon-free-data-center-arizona-/)\n\n\\>430 MW new-to-grid wind, solar, and storage\n\nOperating\n\nNew variable clean supply. Not firm 24/7 power.\n\n[Google . Fervo / NV Energy](https://blog.google/feed/nevada-clean-energy/)\n\n115 MW enhanced geothermal\n\nApproved / development\n\nAdditional firm clean supply if built.\n\n[Google . Kairos Power](https://blog.google/company-news/outreach-and-initiatives/sustainability/google-kairos-power-nuclear-energy-agreement/)\n\nUp to 500 MW advanced nuclear\n\nDevelopment through 2035\n\nPotential new 24/7 clean supply. Not operating.\n\n[Microsoft . Crane Clean Energy Center](https://investors.constellationenergy.com/static-files/45978381-4afe-42b5-83ee-f27339cc2d8d)\n\n835 MW nuclear restart\n\nContracted; 2028 target\n\nAdditive relative to the 2019 shutdown; NRC review remains.\n\n[Amazon . X-energy / Energy Northwest](https://www.aboutamazon.com/news/sustainability/amazon-nuclear-small-modular-reactor-net-carbon-zero)\n\n320 MW first phase; expandable to 960 MW\n\nDevelopment\n\nPotential new firm clean supply on a long clock.\n\n[AWS . Talen / Susquehanna](https://sustainability.aboutamazon.com/stories/how-amazon-is-powering-the-next-era-of-carbon-free-energy)\n\nUp to 1.9 GW access to existing nuclear\n\nContracted; campus operating\n\nLow-carbon preservation or reallocation, not new nameplate supply.\n\n[Meta . Entergy Louisiana](https://www.entergy.com/news/entergy-louisiana-receives-lpsc-approval-for-major-infrastructure-investments-to-support-metas-data-center-and-improve-reliability)\n\nAbout 2,262 MW gas; up to 1,500 MW solar\n\nApproved / procurement\n\nNew supply. The near-term firm component is gas-led.\n\n[Microsoft . Pecos, Texas](https://blogs.microsoft.com/blog/2026/06/22/powering-the-next-wave-of-ai-expanding-capacity-with-our-new-datacenter-in-pecos/)\n\nAbout 2 GW campus; generation undisclosed\n\nAnnounced / development\n\nInitial behind-the-meter gas; later grid interconnection.\n\n[Oracle . Vantage Frontier, Texas](https://vantage-dc.com/news/vantage-data-centers-unveils-plans-for-frontier-a-25b-mega-campus-in-texas-to-meet-unprecedented-ai-demand/)\n\n1.4 GW critical IT campus\n\nUnder construction / development\n\nNew onsite natural-gas microgrid. Campus load is not generation capacity.\n\n[xAI . Southaven, Mississippi](https://www.mlgw.com/images/content/files/pdf/new/xAI%202025%20Update.pdf)\n\n1.2 GW permanent gas generation\n\nUnder construction\n\nNew firm gas, separate from Memphis grid service.\n\n## The Fairchild Pattern: The Fairchild Pattern, now measured in gigawatts.\n\nFairchild Semiconductor's early chips were too expensive for a mass market. NASA and the Department of Defense bought them anyway, because performance mattered more than price. Those first customers financed the factories, the yield improvements, and the accumulated know-how that made the next chip cheaper. Premium demand built a commodity industry.\n\nThe hyperscalers have begun to play the same role in energy. Google is backing up to 500 megawatts from Kairos Power and 115 megawatts of enhanced geothermal through Fervo and NV Energy. Amazon is backing X-energy's first 320-megawatt Energy Northwest project and a broader development program. Microsoft contracted for the 835-megawatt restart of Crane. Meta's nuclear agreements reach across existing plants, uprates, and new reactors. Deep-pocketed customers are converting speculative engineering into projects that can raise capital and place equipment orders.\n\nThat is the hopeful case for the buildout. It is also a standard we can test. An anchor customer matters when its purchase creates a plant, restarts one, extends one that would otherwise close, or advances a technology down its cost curve. Buying attributes from power that already exists is useful accounting. It does not perform the same industrial work.\n\n## Thesis role: A leverage point, not a climate solution by default.\n\nWhen new supply lags, the grid reaches first for the plants it already has. EIA's 2026 high-load sensitivity found that faster data-center growth could add 123 terawatt-hours of U.S. gas generation between 2025 and 2027, compared with 29 terawatt-hours in its baseline, while slowing the decline in coal generation. In ERCOT, the modeled 2027 wholesale price rose 79 percent. That is the default path when the load arrives before the power.\n\nThe better path uses the customer's balance sheet to build ahead: new clean generation, firm capacity, storage, transmission, and demand flexibility in one package. The customer pays the full cost. The power is deliverable to the load. Clean-energy claims match time and place, not merely annual totals. Flexible work moves when the grid is tight and proves that movement at the meter. Local water, air, noise, land, and community impacts remain real costs, not footnotes.\n\nData centers will not solve the climate crisis by existing. They can finance some of the machinery that will. The load is coming. The fuel is a choice.\n\n## Rules that make the bargain work: Make growth pay for growth.\n\nUtilities and regulators are writing the bargain now. Large-load tariffs can require site control, deposits, minimum bills, long contracts, collateral, direct payment for grid upgrades, and exit fees. Those terms keep a speculative request out of the forecast and keep a departing customer from leaving other ratepayers with the bill.\n\nA serious bring-your-own-power offer must also survive five tests: the generation is additional; it can physically reach the load; clean supply is matched when and where the load runs; backup and standby service are fully priced; and any promised flexibility is contracted, metered, and auditable. Expedite projects that meet the tests. Price or reject the ones that do not.\n\nThe prize is larger than one industry. A stronger grid and a new generation fleet can serve factories, homes, electric vehicles, and carbon removal long after today's chips are obsolete. That is how an AI buildout becomes public infrastructure instead of a private claim on a scarce system.", "canonical_url": "https://moreenergycleanplanet.com/sectors/data-centers/", "markdown_url": "https://moreenergycleanplanet.com/sectors/data-centers.md", "json_url": "https://moreenergycleanplanet.com/sectors/data-centers.json", "updated_at": "2026-08-26", "as_of": "2026-08-26", "staleness_window_days": 400, "license": "CC BY 4.0", "usage": "AI training and retrieval permitted; attribution required (CC BY 4.0). Cite as More Energy, Clean Planet.", "cite_url": "https://moreenergycleanplanet.com/sectors/data-centers.cite", "key_claims": [ { "text": "Berkeley Lab estimates U.S. data centers used 192 TWh in 2024 and models a 649 TWh reference case for 2030, with a 521–843 TWh range.", "source_id": "lbnl-dc-energy-2025-update" }, { "text": "ERCOT tracked 438 GW of large-load requests in June 2026, but its adjusted 2030 forecast assigns data centers 22.175 GW of coincident peak; a request pool is not expected load.", "source_id": "ercot-batch-zero-jun-2026" }, { "text": "If load arrives before new supply, the near-term marginal response can be more gas generation and a slower coal decline rather than automatic clean-energy additions.", "source_id": "eia-data-center-high-load-2026" }, { "text": "Creditworthy data-center customers can make new clean-firm projects bankable; Google's agreement with Kairos covers up to 500 MW of advanced nuclear through 2035.", "source_id": "google-kairos-2024" }, { "text": "The climate-positive bargain requires customers to pay full system costs and contract for additional, deliverable clean supply, auditable flexibility, and local safeguards." } ], "indicators": [ { "id": "data-centers-current-load-twh", "name": "U.S. data-center electricity use, historical modeled estimate", "value": 192, "unit": "TWh/yr", "goal": null, "goal_year": null, "as_of": "2024-12-31", "staleness_window_days": 400, "stale": true, "methodology": "display: 192 TWh | 21.9 GW average; 4.7% of projected U.S. electricity on Berkeley Lab denominator.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-current-load-twh", "source_id": "iea-energy-ai-2025" }, { "id": "data-centers-2030-load-twh", "name": "U.S. data-center electricity use, 2030 reference case", "value": 649, "unit": "TWh/yr", "goal": 649, "goal_unit": "TWh/yr", "goal_year": 2030, "target_year": 2030, "projection": true, "as_of": "2024-12-01", "staleness_window_days": 540, "stale": true, "methodology": "display: 649 TWh | 74.1 GW average; 11.8% of projected U.S. electricity; 33% of projected U.S. load growth, 2024–2030.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-2030-load-twh", "source_id": "lbnl-dc-energy-2024" }, { "id": "data-centers-2030-load-low-twh", "name": "U.S. data-center electricity use, 2030 low case", "value": 521, "unit": "TWh/yr", "goal": 521, "goal_unit": "TWh/yr", "goal_year": 2030, "target_year": 2030, "projection": true, "as_of": "2026-06-01", "staleness_window_days": 540, "stale": false, "methodology": "display: 521 TWh | 59.5 GW average; 9.5% of projected U.S. electricity.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-2030-load-low-twh", "source_id": "lbnl-dc-energy-2025-update" }, { "id": "data-centers-2030-load-high-twh", "name": "U.S. data-center electricity use, 2030 high case", "value": 843, "unit": "TWh/yr", "goal": 843, "goal_unit": "TWh/yr", "goal_year": 2030, "target_year": 2030, "projection": true, "as_of": "2026-06-01", "staleness_window_days": 540, "stale": false, "methodology": "display: 843 TWh | 96.2 GW average; 15.3% of projected U.S. electricity.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-2030-load-high-twh", "source_id": "lbnl-dc-energy-2025-update" }, { "id": "data-centers-us-grid-2024-twh", "name": "U.S. net electricity generation, present-grid scale reference", "value": 4308.60846, "unit": "TWh/yr", "goal": null, "goal_year": null, "as_of": "2024-12-31", "staleness_window_days": 400, "stale": true, "methodology": "display: 4,308.6 TWh | U.S. net generation in 2024.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-us-grid-2024-twh", "source_id": "eia-us-electricity-profile-2024" }, { "id": "data-centers-tx-energized-gw", "name": "ERCOT observed energized large-load projects", "value": 5.778, "unit": "GW", "goal": null, "goal_year": null, "as_of": "2026-03-26", "staleness_window_days": 400, "stale": false, "methodology": "display: 5.778 GW | All large-load projects observed energized.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-tx-energized-gw", "source_id": "ercot-large-load-apr-2026" }, { "id": "data-centers-tx-approved-gw", "name": "ERCOT large loads approved to energize", "value": 3.241, "unit": "GW", "goal": null, "goal_year": null, "as_of": "2026-03-26", "staleness_window_days": 400, "stale": false, "methodology": "display: 3.241 GW | All large loads approved to energize.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-tx-approved-gw", "source_id": "ercot-large-load-apr-2026" }, { "id": "data-centers-tx-2030-forecast-gw", "name": "ERCOT data-center contribution to adjusted 2030 coincident peak forecast", "value": 22.175, "unit": "GW", "goal": 22.175, "goal_unit": "GW", "goal_year": 2030, "target_year": 2030, "projection": true, "as_of": "2025-04-01", "staleness_window_days": 540, "stale": true, "methodology": "display: 22.175 GW | Data-center coincident summer-peak forecast.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-tx-2030-forecast-gw", "source_id": "ercot-long-term-load-forecast-2025" }, { "id": "data-centers-tx-requests-gw", "name": "ERCOT tracked large-load requests", "value": 438, "unit": "GW", "goal": null, "goal_year": null, "as_of": "2026-06-18", "staleness_window_days": 400, "stale": false, "methodology": "display: 438 GW | Tracked large-load requests; 89% data centers.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-tx-requests-gw", "source_id": "ercot-batch-zero-jun-2026" }, { "id": "data-centers-az-operating-gw", "name": "Arizona operating data-center facility capacity", "value": 2, "unit": "GW", "goal": null, "goal_year": null, "as_of": "2026-04-20", "staleness_window_days": 400, "stale": false, "methodology": "display: >2.0 GW | Operating facility capacity statewide.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-az-operating-gw", "source_id": "azcc-data-center-workshop-2026" }, { "id": "data-centers-az-contracted-gw", "name": "TEP contracted data-center load", "value": 0.286, "unit": "GW", "goal": null, "goal_year": null, "as_of": "2026-04-20", "staleness_window_days": 400, "stale": false, "methodology": "display: 0.286 GW | TEP contracted load.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-az-contracted-gw", "source_id": "azcc-data-center-workshop-2026" }, { "id": "data-centers-az-system-forecast-gw", "name": "APS total-system peak forecast", "value": 12, "unit": "GW", "goal": 12, "goal_unit": "GW", "goal_year": 2035, "target_year": 2035, "projection": true, "as_of": "2026-06-30", "staleness_window_days": 540, "stale": false, "methodology": "display: 8.7→12.0 GW | APS total-system peak, 2025–2035.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-az-system-forecast-gw", "source_id": "aps-data-centers-2026" }, { "id": "data-centers-az-requests-gw", "name": "APS uncommitted extra-large-load requests", "value": 19, "unit": "GW", "goal": null, "goal_year": null, "as_of": "2026-04-20", "staleness_window_days": 400, "stale": false, "methodology": "display: 19 GW | APS uncommitted request pool.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-az-requests-gw", "source_id": "aps-data-centers-2026" }, { "id": "data-centers-ut-operating-gw", "name": "Utah operational data-center facility and IT capacity", "value": 0.92, "unit": "GW", "goal": null, "goal_year": null, "as_of": "2026-08-01", "staleness_window_days": 400, "stale": false, "methodology": "display: >0.920 GW | Operational facility / IT nameplate.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-ut-operating-gw", "source_id": "utah-data-centers-aug-2026" }, { "id": "data-centers-ut-construction-gw", "name": "Utah data-center facility capacity under construction", "value": 2.602, "unit": "GW", "goal": null, "goal_year": null, "as_of": "2026-08-01", "staleness_window_days": 400, "stale": false, "methodology": "display: 2.602 GW | Facility / IT nameplate under construction.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-ut-construction-gw", "source_id": "utah-data-centers-aug-2026" }, { "id": "data-centers-ut-requests-gw", "name": "Rocky Mountain Power large-load requests evaluated in six months", "value": 3, "unit": "GW", "goal": null, "goal_year": null, "as_of": "2026-08-01", "staleness_window_days": 400, "stale": false, "methodology": "display: ≈3 GW | Requests evaluated in six months.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-ut-requests-gw", "source_id": "utah-data-centers-aug-2026" }, { "id": "data-centers-va-metered-gw", "name": "Dominion Virginia metered coincident data-center demand", "value": 3.127, "unit": "GW", "goal": null, "goal_year": null, "as_of": "2023-12-31", "staleness_window_days": 400, "stale": true, "methodology": "display: 3.127 GW | Metered coincident demand in 2023.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-va-metered-gw", "source_id": "dominion-virginia-irp-2024" }, { "id": "data-centers-va-esa-gw", "name": "Dominion Virginia data-center load under electric service agreements", "value": 8.012, "unit": "GW", "goal": null, "goal_year": null, "as_of": "2024-12-31", "staleness_window_days": 400, "stale": true, "methodology": "display: 8.012 GW | Electric service agreement stage.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-va-esa-gw", "source_id": "dominion-virginia-irp-2024" }, { "id": "data-centers-va-2032-forecast-gw", "name": "Dominion Virginia data-center coincident demand forecast", "value": 7.686, "unit": "GW", "goal": 7.686, "goal_unit": "GW", "goal_year": 2032, "target_year": 2032, "projection": true, "as_of": "2024-06-30", "staleness_window_days": 540, "stale": true, "methodology": "display: 7.686 GW | Data-center coincident forecast for 2032.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-va-2032-forecast-gw", "source_id": "dominion-virginia-irp-2024" }, { "id": "data-centers-ga-contracted-gw", "name": "Georgia Power contracted OpenAI load", "value": 3.2, "unit": "GW", "goal": null, "goal_year": null, "as_of": "2026-08-26", "staleness_window_days": 400, "stale": false, "methodology": "display: 3.2 GW | Approved OpenAI contract.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-ga-contracted-gw", "source_id": "georgia-power-openai-2026" }, { "id": "data-centers-ga-system-forecast-gw", "name": "Georgia Power total-system winter peak forecast", "value": 24.5, "unit": "GW", "goal": 24.5, "goal_unit": "GW", "goal_year": 2030, "target_year": 2030, "projection": true, "as_of": "2026-06-30", "staleness_window_days": 540, "stale": false, "methodology": "display: 16.3→24.5 GW | Total winter peak, 2025–2030.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-ga-system-forecast-gw", "source_id": "georgia-power-openai-2026" }, { "id": "data-centers-ga-pipeline-gw", "name": "Georgia Power large-load economic-development pipeline", "value": 73.1, "unit": "GW", "goal": null, "goal_year": null, "as_of": "2026-08-26", "staleness_window_days": 400, "stale": false, "methodology": "display: 73.1 GW | Large-load development pipeline.", "anchor_url": "https://moreenergycleanplanet.com/scorecard/energy-production/data-centers-ga-pipeline-gw", "source_id": "georgia-power-openai-2026" } ], "cost_curve": null, "companies": [ { "id": "amazon-web-services", "name": "Amazon Web Services", "stage": "commercial", "one_line_description": "Largest US cloud + AI-infrastructure operator (subsidiary of Amazon NASDAQ:AMZN). 2026 capex >$125B (Morningstar 2025). Talen / Vistra nuclear PPAs anchor 24/7 carbon-free-energy posture; multi-GW data-center campuses across PJM + ERCOT + SPP.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/amazon-web-services" }, { "id": "apple", "name": "Apple", "stage": "commercial", "one_line_description": "Consumer-device and cloud operator with a material U.S. data-center footprint and long-running renewable-electricity procurement, but a smaller causal role in the current AI power buildout than the leading cloud and model companies.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/apple" }, { "id": "bloom-energy", "name": "Bloom Energy", "stage": "public", "one_line_description": "US solid-oxide fuel-cell OEM (NYSE: BE). BTM on-site-power deployments at AWS / Equinix / Oracle / Walmart at hyperscaler + colo scale. Hydrogen-ready stack; multi-MW deployments.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/bloom-energy" }, { "id": "constellation-energy", "name": "Constellation Energy", "stage": "commercial", "one_line_description": "Largest US nuclear-generation operator (NYSE: CEG). 21 reactors / ~33 GW. Three Mile Island Unit 1 restart agreed with Microsoft (Sep 2024, 20-year PPA, targeted 2028 ops). Coal divestiture complete pre-spinoff. Primary sector: nuclear-fission; cross-tagged here for managed-decline-of-coal context.", "primary_sector": "coal", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/constellation-energy" }, { "id": "coreweave", "name": "CoreWeave", "stage": "scaling", "one_line_description": "GPU-cloud operator leasing and developing high-density AI capacity across U.S. markets; a material demand counterparty whose public power disclosures remain less complete than the largest hyperscalers.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/coreweave" }, { "id": "crusoe-energy", "name": "Crusoe Energy", "stage": "scaling", "one_line_description": "AI-data-center developer + BTM-generation operator. Originally stranded-gas-to-Bitcoin Digital Flare Mitigation; now AI-infrastructure-as-a-service with Oracle anchor + Stargate JV. Multiple GW-scale projects in pipeline. Status: scaling-to-hyperscaler.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/crusoe-energy" }, { "id": "cyrusone", "name": "CyrusOne", "stage": "commercial", "one_line_description": "Data-center developer and operator serving hyperscale and enterprise customers; landlord capacity is tracked separately from tenant demand to avoid double counting.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/cyrusone" }, { "id": "digital-realty", "name": "Digital Realty Trust", "stage": "commercial", "one_line_description": "US data-center REIT (NYSE: DLR). ~300 facilities globally; hyperscaler build-to-suit + colocation. 100% renewable-electricity for US colo footprint; long-term green-bond financing.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/digital-realty" }, { "id": "entergy-louisiana", "name": "Entergy Louisiana", "stage": "commercial", "one_line_description": "Regulated utility building the initial generation and grid package for Meta Hyperion, including about 2,262 MW of gas generation and procurement for up to 1,500 MW of solar.", "primary_sector": "the-grid", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/entergy-louisiana" }, { "id": "equinix", "name": "Equinix", "stage": "commercial", "one_line_description": "Largest US carrier-neutral colocation operator. NASDAQ: EQIX. ~260 data centers in 70 markets globally. 100% renewable-electricity commitment by 2030; PPAs across US + Europe; Bloom Energy fuel-cell deployments at multiple sites.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/equinix" }, { "id": "fervo-energy", "name": "Fervo Energy", "stage": "public", "one_line_description": "Enhanced geothermal developer using horizontal drilling and AI-enabled subsurface analytics to deliver 24/7 carbon-free baseload power.", "primary_sector": "geothermal", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/fervo-energy" }, { "id": "google", "name": "Google (Alphabet)", "stage": "commercial", "one_line_description": "Google Cloud + Search + Workspace + AI hyperscaler (subsidiary of Alphabet NASDAQ:GOOGL). 2026 capex ~$75B (Morningstar 2025). 24/7 carbon-free-energy commitment (originator of the framework); Fervo geothermal PPA; multi-GW data-center campuses across US.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/google" }, { "id": "kairos-power", "name": "Kairos Power", "stage": "growth", "one_line_description": "Developer of fluoride salt-cooled high-temperature pebble-bed nuclear reactors using TRISO fuel, with the Hermes demonstration reactor under construction in Oak Ridge and a 500 MW master agreement with Google.", "primary_sector": "nuclear-fission", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/kairos-power" }, { "id": "mainspring-energy", "name": "Mainspring Energy", "stage": "commercial", "one_line_description": "US BTM on-site-power OEM; proprietary linear generator runs on natural gas, hydrogen, biogas, or ammonia. Deployments at hyperscaler + colo sites (AWS, Walmart, Lineage); positioned as fast-deploy bridge generation pending grid energization.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/mainspring-energy" }, { "id": "meta", "name": "Meta Platforms", "stage": "commercial", "one_line_description": "Hyperscaler + AI-infrastructure operator (Facebook, Instagram, WhatsApp, AI). NASDAQ: META. 2026 capex ~$60-65B (Morningstar 2025). 100% renewable-energy match commitment achieved 2020; net-zero value-chain by 2030; multi-GW data-center campuses across SE US.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/meta" }, { "id": "microsoft", "name": "Microsoft", "stage": "commercial", "one_line_description": "Hyperscaler + AI-infrastructure operator. NASDAQ: MSFT. 2026 capex ~$90B+ (Morningstar 2025). Three Mile Island Unit 1 restart PPA with Constellation (20-year, Sep 2024); Helion fusion PPA (2028 target); 100% renewable-energy target by 2025; net-zero by 2030.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/microsoft" }, { "id": "oklo", "name": "Oklo", "stage": "public", "one_line_description": "Public advanced fission developer building the Aurora Powerhouse, a 15-75 MW liquid-metal-cooled fast reactor with a ~14 GW non-binding pipeline targeting AI data centers, with Atomic Alchemy subsidiary targeting first criticality by July 4 2026.", "primary_sector": "nuclear-fission", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/oklo" }, { "id": "openai-stargate", "name": "OpenAI / Stargate", "stage": "scaling", "one_line_description": "AI developer and Stargate demand counterparty behind multiple gigawatt-scale U.S. campuses, including an approved 3.2 GW Georgia Power service contract.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/openai-stargate" }, { "id": "oracle-cloud", "name": "Oracle Cloud Infrastructure", "stage": "commercial", "one_line_description": "Cloud + AI-infrastructure operator (subsidiary of Oracle NYSE:ORCL). 2026 capex ramping (~$25B+ AI infrastructure spend); anchor tenant in Crusoe AI-data-center pipeline (Stargate JV with OpenAI + SoftBank). Aggressive nuclear + behind-the-meter strategy.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/oracle-cloud" }, { "id": "qts-data-centers", "name": "QTS Data Centers / Blackstone", "stage": "commercial", "one_line_description": "Large U.S. data-center developer and operator owned by Blackstone, serving hyperscale and enterprise customers across major power markets.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/qts-data-centers" }, { "id": "sage-geosystems", "name": "Sage Geosystems", "stage": "foak", "one_line_description": "US enhanced + geopressured-geothermal developer. Geopressured Geothermal System (GGS) couples energy generation with subsurface storage. San Miguel TX pilot operational. Meta data-center PPA signed Aug 2024.", "primary_sector": "geothermal", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/sage-geosystems" }, { "id": "schneider-electric-data-centers", "name": "Schneider Electric (Data Centers)", "stage": "commercial", "one_line_description": "Global power-management + cooling OEM (Euronext: SU). Schneider Electric's data-center BU is the largest single market; APC UPS line + EcoStruxure microgrid software + Galaxy power-distribution.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/schneider-electric-data-centers" }, { "id": "stack-infrastructure", "name": "STACK Infrastructure", "stage": "commercial", "one_line_description": "Hyperscale data-center developer and operator with large U.S. campus portfolios; landlord capacity may house demand disclosed by a tenant and is not additive to it.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/stack-infrastructure" }, { "id": "submer", "name": "Submer", "stage": "commercial", "one_line_description": "Spanish immersion-cooling OEM for high-density GPU + AI racks. SmartPod + SmartPod GO products at hyperscaler + colo deployments. Backed by Mubadala, Cisco Ventures, M&G.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/submer" }, { "id": "talen-energy", "name": "Talen Energy", "stage": "commercial", "one_line_description": "US IPP. Divested entire coal generation portfolio 2023 (Brunner Island, Montour, etc.); now owns Susquehanna nuclear (2.5 GW boiling-water reactor). $650M PPA to AWS for Cumulus data-center campus. Emerged from Chapter 11 May 2023 with clean balance sheet. The cleanest single-name coal-to-nuclear-IPP pivot on record.", "primary_sector": "coal", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/talen-energy" }, { "id": "terrapower", "name": "TerraPower", "stage": "foak", "one_line_description": "US advanced-reactor developer (Bill Gates + Khosla Ventures backed). Natrium sodium-cooled fast reactor with integrated molten-salt storage; 345 MWe + 500 MWe peaking. Kemmerer WY first-of-a-kind under construction since 2024 (non-nuclear systems first; targeting 2030 ops). DOE ARDP $2B cost-share.", "primary_sector": "nuclear-fission", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/terrapower" }, { "id": "vantage-data-centers", "name": "Vantage Data Centers", "stage": "commercial", "one_line_description": "Hyperscale campus developer and operator; its 1.4 GW Frontier campus in Texas pairs Oracle demand with a large onsite natural-gas microgrid.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/vantage-data-centers" }, { "id": "vertiv", "name": "Vertiv", "stage": "commercial", "one_line_description": "Data-center power + cooling OEM (NYSE: VRT). Cold-plate + immersion liquid-cooling for GPU racks (the H100/B200 generation); modular UPS + power-distribution; the BoS exemplar for high-density AI infrastructure.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/vertiv" }, { "id": "vistra-energy", "name": "Vistra Energy", "stage": "commercial", "one_line_description": "US IPP (NYSE: VST). Owns Comanche Peak nuclear (2.4 GW). Coal retirements: Vermilion + Joppa + Edwards (IL) retired; published 2030 coal exit. AWS PPA at Comanche Peak (2024). Energy Harbor acquisition (Mar 2024) added Davis-Besse + Perry + Beaver Valley nuclear (Sustainable Energy segment).", "primary_sector": "coal", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/vistra-energy" }, { "id": "voltagrid", "name": "VoltaGrid", "stage": "scaling", "one_line_description": "Onsite natural-gas generation and microgrid provider serving fast-to-power industrial and AI data-center projects.", "primary_sector": "natural-gas", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/voltagrid" }, { "id": "x-energy", "name": "X-energy", "stage": "public", "one_line_description": "Developer of high-temperature gas-cooled small modular reactors using TRISO ceramic fuel pebbles, with first commercial deployment at Dow's Seadrift Texas petrochemical site under DOE's Advanced Reactor Demonstration Program.", "primary_sector": "nuclear-fission", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/x-energy" }, { "id": "xai-spacexai", "name": "xAI / SpaceXAI", "stage": "scaling", "one_line_description": "AI infrastructure operator using grid service and dedicated onsite generation; its Southaven project is a clear speed-to-power case built around permanent natural-gas turbines.", "ecosystem_url": "https://moreenergycleanplanet.com/ecosystem#company/xai-spacexai" } ], "faqs": [ { "question": "How much electricity do U.S. data centers use?", "answer_md": "Berkeley Lab estimates **192 TWh in 2024**, equal to 21.9 GW of average demand. Its 2030 reference case is **649 TWh**, or 74.1 GW average and 11.8% of projected U.S. electricity. The modeled 2030 range is **521–843 TWh**. These are modeled estimates, not a national meter total or contracted load.", "source_id": "lbnl-dc-energy-2025-update" }, { "question": "Does Texas really have 438 GW of data centers coming?", "answer_md": "No. ERCOT's **438 GW** figure was a June 2026 pool of large-load requests, 89% associated with data centers. Its finalized, probability-adjusted 2030 forecast assigns data centers **22.175 GW** of coincident summer peak. Requests include projects that move, duplicate sites, miss dates, or withdraw.", "source_id": "ercot-batch-zero-jun-2026" }, { "question": "Will data centers run on clean energy?", "answer_md": "Some will help build new clean supply; others are already pulling new gas plants and onsite turbines into the market. The project, fuel, status, grid relation, and additivity matter more than an annual renewable-energy claim. Demand creates leverage. The power contract decides what it builds.", "source_id": "eia-data-center-high-load-2026" }, { "question": "How can the data-center buildout help with climate?", "answer_md": "Large, creditworthy customers can serve as anchor buyers for first-of-a-kind and grid-scale projects. Google / Kairos, Amazon / X-energy, Microsoft / Crane, and Google / Fervo show the mechanism. The climate value appears when the contract brings forward additional, deliverable clean power rather than merely reallocating existing attributes.", "source_id": "google-kairos-2024" }, { "question": "Who should pay for the new power and grid infrastructure?", "answer_md": "The customer creating the load should pay the full cost of generation, transmission, interconnection, standby service, and any stranded assets. Minimum bills, collateral, direct upgrade charges, staged commitments, and exit fees keep speculative projects out of utility forecasts and protect other ratepayers.", "source_id": null } ], "related_sectors": [ { "slug": "bioenergy", "title": "Bioenergy" }, { "slug": "coal", "title": "Coal" }, { "slug": "fusion", "title": "Fusion" }, { "slug": "geothermal", "title": "Geothermal" }, { "slug": "hydro", "title": "Hydro" }, { "slug": "natural-gas", "title": "Natural Gas" }, { "slug": "nuclear-fission", "title": "Nuclear Fission" }, { "slug": "oil", "title": "Oil" }, { "slug": "solar", "title": "Solar" }, { "slug": "storage", "title": "Energy Storage" }, { "slug": "the-grid", "title": "The Grid" }, { "slug": "wind", "title": "Wind" } ], "embeds": [ { "id": "data-centers-load-scale", "title": "The next grid is arriving inside this one.", "permalink": "https://moreenergycleanplanet.com/embed/data-centers-load-scale", "svg": "https://moreenergycleanplanet.com/embed/data-centers-load-scale.svg", "png": "https://moreenergycleanplanet.com/embed/data-centers-load-scale.png" }, { "id": "data-centers-state-status", "title": "A request is not a forecast.", "permalink": "https://moreenergycleanplanet.com/embed/data-centers-state-status", "svg": "https://moreenergycleanplanet.com/embed/data-centers-state-status.svg", "png": "https://moreenergycleanplanet.com/embed/data-centers-state-status.png" } ], "sources": [ { "id": "iea-energy-ai-2025", "title": "Energy and AI", "url": "https://www.iea.org/reports/energy-and-ai", "publisher_tier": "primary", "source_type": "report", "note": "Canonical source for data-center electricity demand growth: 415 TWh global 2024 → 945 TWh by 2030 (Base Case, +2.28x over 6 years); US specifically 183 TWh → 426 TWh (+2.33x). AI-focused load triples. The \"↑ more than double by 2030\" marker on the Sankey's data-centers node sources here." }, { "id": "lbnl-dc-energy-2024", "title": "2024 United States Data Center Energy Usage Report", "url": "https://eta-publications.lbl.gov/sites/default/files/2024-12/lbnl-2024-united-states-data-center-energy-usage-report_1.pdf", "publisher_tier": "primary", "source_type": "report", "note": "325-580 TWh by 2028 range. The bracket IS the editorial." }, { "id": "lbnl-dc-energy-2025-update", "title": "United States Data Center Energy Usage Report: 2025 Update", "url": "https://escholarship.org/content/qt33m6w3x0/qt33m6w3x0.pdf", "publisher_tier": "primary", "note": "Bottom-up modeled estimate and projections. Excludes cryptocurrency mining. Table 2 and Power Capacity Implications." }, { "id": "eia-us-electricity-profile-2024", "title": "United States Electricity Profile 2024", "url": "https://www.eia.gov/electricity/state/unitedstates/", "publisher_tier": "primary", "note": "Table 1: 4,308,608,460 MWh net generation and 1,230,391 MW net summer capacity." }, { "id": "ercot-large-load-apr-2026", "title": "Large Load Update to Texas House State Affairs", "url": "https://www.ercot.com/files/docs/2026/04/09/ERCOTLargeLoadUpdate-April9HouseStateAffairsHearing.pdf", "publisher_tier": "primary", "note": "March 26, 2026 status snapshot. Energized and approved-to-energize values cover all large loads, not data centers alone." }, { "id": "ercot-long-term-load-forecast-2025", "title": "2025 Long-Term Load Forecast", "url": "https://www.ercot.com/files/docs/2025/04/08/ERCOT-2025-Long-Term-Load-Forecast-Report.pdf", "publisher_tier": "primary", "note": "Finalized adjusted coincident summer-peak contribution by large-load class." }, { "id": "ercot-batch-zero-jun-2026", "title": "PUCT approves ERCOT large-load interconnection review", "url": "https://www.ercot.com/news/release/06182026-puct-approves-ercots", "publisher_tier": "primary", "note": "438 GW tracked large-load request pool; 89 percent associated with data centers. Requests are not a forecast." }, { "id": "azcc-data-center-workshop-2026", "title": "ACC Large Load / Data Center Workshop Highlights", "url": "https://azcc.gov/kevin-thompson/news/2026/04/20/acc-large-load-data-center-workshop-highlights", "publisher_tier": "primary", "note": "Statewide operating facility capacity and TEP contracted load." }, { "id": "aps-data-centers-2026", "title": "Data Centers: Protecting Customers While Planning for Big Energy Needs", "url": "https://www.aps.com/en/Utility/Company-wide-Events-and-Messages/Data_Centers_How_Were_Protecting_Customers_While_Planning-for_Big_Energy_Needs", "publisher_tier": "primary", "note": "APS total-system peak forecast and uncommitted extra-large-load request pool." }, { "id": "utah-data-centers-aug-2026", "title": "Data Centers in Utah", "url": "https://le.utah.gov/interim/2026/pdf/PB_DataCenters.pdf", "publisher_tier": "primary", "note": "Facility and IT nameplate values; the construction total includes about 1.7 GW described as behind meter or off grid." }, { "id": "dominion-virginia-irp-2024", "title": "2024 Integrated Resource Plan", "url": "https://www.dominionenergy.com/-/media/content/about/our-company/irp/pdfs/2024-irp-w_o-appendices.pdf", "publisher_tier": "primary", "note": "Metered coincident demand, electric-service-agreement stage, and coincident forecast. Development stages may overlap." }, { "id": "georgia-power-openai-2026", "title": "OpenAI contract approved as part of Georgia Power resource portfolio", "url": "https://www.georgiapower.com/news-hub/press-releases/contract-openai-approved-part-portfolio-delivering-950-million-annual-savings.html", "publisher_tier": "primary", "note": "Approved 3.2 GW OpenAI contract, system winter-peak forecast, and large-load development pipeline." }, { "id": "eia-data-center-high-load-2026", "title": "Fossil generation could rise with faster-than-expected growth in data center power demand", "url": "https://www.eia.gov/todayinenergy/detail.php?id=67344", "publisher_tier": "primary", "note": "High-load sensitivity holds generating capacity fixed and isolates the near-term marginal response; it is not a forecast." }, { "id": "google-kairos-2024", "title": "Google signs advanced nuclear agreement with Kairos Power", "url": "https://blog.google/company-news/outreach-and-initiatives/sustainability/google-kairos-power-nuclear-energy-agreement/", "publisher_tier": "self-reported", "note": "Master plant development agreement for up to 500 MW through 2035; development-stage, not operating supply." } ], "mcp": { "server": "https://mcp.moreenergycleanplanet.com/mcp", "tools": [ "get_sector", "get_cost_curve", "list_indicators", "get_sources" ] } } ```