U.S. data center market: state profiles, operators & grid queue
All 50 states and D.C. — facility counts, power draw, operator ownership, the ERCOT large-load funnel, SEC filings, and how the grid is responding to the demand wave.
On this page
- All 50 states: facility count & power draw
- Owners, operators & shell LLCs
- ERCOT large-load interconnection queue
- SEC 10-K competitor analysis
- The demand wave — ERCOT & PJM
- How the grid operators & FERC are responding
All 50 states and D.C.: data center facility count & power draw
| State | Abbrev | Facilities | TWh/yr | Major hubs | Upcoming |
|---|---|---|---|---|---|
| Alabama | AL | 35 | 0.6 | Birmingham, Huntsville | — |
| Alaska | AK | 5 | 0.1 | Anchorage | — |
| Arizona | AZ | 190 | 10.5 | Phoenix (Mesa, Goodyear, Chandler) — Microsoft, Apple, Vantage, QTS, Iron Mountain | ✓ |
| Arkansas | AR | 15 | 0.3 | West Memphis — Google | ✓ |
| California | CA | 321 | 11.0 | Silicon Valley, Los Angeles — Equinix, Digital Realty, CoreSite, Vantage | — |
| Colorado | CO | 70 | 1.8 | Denver, Colorado Springs — Flexential, Viawest | — |
| Connecticut | CT | 60 | 1.0 | Stamford, Hartford | — |
| Delaware | DE | 22 | 0.4 | Wilmington | — |
| Dist. of Columbia | DC | 12 | 0.3 | Washington D.C. | — |
| Florida | FL | 135 | 4.2 | Miami, Tampa, Jacksonville — Equinix, Digital Realty, CyrusOne | — |
| Georgia | GA | 162 | 9.0 | Atlanta (Douglasville, Palmetto) — Switch, Google, Microsoft, QTS | ✓ |
| Hawaii | HI | 8 | 0.1 | Honolulu | — |
| Idaho | ID | 12 | 0.2 | Boise | — |
| Illinois | IL | 244 | 12.0 | Chicago (Elk Grove Village, Aurora) — Microsoft, Equinix, Digital Realty, Meta | — |
| Indiana | IN | 85 | 2.0 | Lebanon, La Porte, Indianapolis — Amazon ($15B), Meta ($10B), Microsoft ($1B) | ✓ |
| Iowa | IA | 115 | 3.8 | Des Moines, Altoona — Apple, Google, Meta | — |
| Kansas | KS | 22 | 0.4 | Kansas City metro, Wichita | — |
| Kentucky | KY | 40 | 0.6 | Louisville, Lexington | — |
| Louisiana | LA | 30 | 0.5 | Richland Parish, Caddo & Bossier Parishes — Meta ($27B), Amazon ($12B), Hut 8 | ✓ |
| Maine | ME | 8 | 0.1 | Portland | — |
| Maryland | MD | 48 | 0.9 | Baltimore, Frederick | — |
| Massachusetts | MA | 58 | 1.1 | Boston, Holyoke, Cambridge | — |
| Michigan | MI | 62 | 1.1 | Detroit, Grand Rapids | — |
| Minnesota | MN | 78 | 1.5 | Minneapolis, Shakopee | — |
| Mississippi | MS | 20 | 0.3 | Madison County, Meridian — Amazon ($16B+), Compass ($10B) | ✓ |
| Missouri | MO | 55 | 1.2 | Kansas City, St. Louis — Google ($2B) | ✓ |
| Montana | MT | 12 | 0.2 | Billings, Missoula | — |
| Nebraska | NE | 42 | 2.1 | Omaha, Lincoln — Google, Meta | — |
| Nevada | NV | 72 | 2.8 | Las Vegas (Henderson), Reno — Switch (SUPERNAP), Google, Apple | — |
| New Hampshire | NH | 14 | 0.2 | Manchester | — |
| New Jersey | NJ | 95 | 2.2 | Secaucus, Piscataway (NYC Metro) — Equinix, Digital Realty | — |
| New Mexico | NM | 28 | 1.6 | Albuquerque, Los Lunas — Meta | — |
| New York | NY | 155 | 4.5 | NYC Metro, Buffalo — Equinix, DataBank, Digital Realty | — |
| North Carolina | NC | 115 | 4.5 | Charlotte, Research Triangle, Richmond County — Apple, Google, Amazon ($10B) | ✓ |
| North Dakota | ND | 18 | 0.3 | Fargo | — |
| Ohio | OH | 203 | 7.5 | Columbus (New Albany), Hebron — Google, Meta, Amazon ($10B), Microsoft | ✓ |
| Oklahoma | OK | 38 | 0.6 | Oklahoma City, Tulsa | — |
| Oregon | OR | 148 | 6.5 | Hillsboro, Prineville, The Dalles — Apple, Meta, Google | — |
| Pennsylvania | PA | 98 | 2.5 | Salem Twp., Fairless Hills, Lancaster — Amazon ($20B), CoreWeave ($6B) | ✓ |
| Rhode Island | RI | 8 | 0.2 | Providence | — |
| South Carolina | SC | 40 | 0.6 | Aiken County, Charleston — Meta ($800M) | ✓ |
| South Dakota | SD | 10 | 0.1 | Sioux Falls | — |
| Tennessee | TN | 68 | 1.8 | Nashville, Memphis — Oracle ($1.5B), xAI ($1B+) | — |
| Texas | TX | 413 | 17.0 | Dallas–Fort Worth, Austin, San Antonio, West Texas — Google ($40B), Meta, Stargate | ✓ |
| Utah | UT | 48 | 2.6 | Salt Lake City, Bluffdale — Meta, Google | — |
| Vermont | VT | 5 | 0.1 | Burlington | — |
| Virginia | VA | 665 | 24.0 | Ashburn ('Data Center Alley'), Loudoun Co., Prince William — AWS, Microsoft, Google, Equinix | ✓ |
| Washington | WA | 142 | 4.2 | Quincy, Seattle, Moses Lake — Microsoft, Amazon | — |
| West Virginia | WV | 8 | 0.1 | Charleston | — |
| Wisconsin | WI | 52 | 0.9 | Mount Pleasant, Milwaukee — Microsoft ($3.3B) | ✓ |
| Wyoming | WY | 18 | 0.5 | Cheyenne — Microsoft | — |
50 states and D.C.
Top 15 states by annual power draw
As of July 2026 · changes over months · Source: U.S. Data Center Power Map by State — ElectricChoice.com compilation (updated July 2026; CC-BY 4.0). Underlying load research: LBNL 2024 US Data Center Energy Usage Report; rates: EIA
Facility counts vary by directory because each one draws the boundary differently — DataCenterMap lists 44 in Alabama where this table says 35, since it counts small network rooms alongside hyperscale campuses. Cite the TWh figure rather than the facility count when you can; it is the number that speaks to grid impact. Underlying load research is LBNL-2001637 (Dec 2024).
Sources: Lawrence Berkeley National Lab, 2024 US Data Center Energy Usage Report — 325–580 TWh (6.7–12% of US electricity) by 2028, up from 176 TWh (4.4%) in 2023 · EIA — State Electricity Profiles (generation, prices, consumption, emissions per state; U.S. Energy Information Administration) · U.S. Data Center Power Map by State — ElectricChoice.com compilation (updated July 2026; CC-BY 4.0). Underlying load research: LBNL 2024 US Data Center Energy Usage Report; rates: EIA
Owners, operators & shell LLCs
A campus has up to three separate parties — the operator (who runs it), the owner (often a PE fund), and the tenant (who actually consumes the power). Land is bought through single-purpose shell LLCs, the join key back to county deed records.
| Operator | Category | Owner / PE parent | Model | Tenant disclosed | Filing LLCs |
|---|---|---|---|---|---|
| Hyperscaler (owns & consumes) | Public (Alphabet, NASDAQ: GOOGL) | Own | Self-consumed | Land-acquisition shells: Jet Stream LLC, Sharka LLC | |
| Meta | Hyperscaler (owns & consumes) | Public (NASDAQ: META) | Own | Self-consumed | Codename shells: Greater Kudu, Raven Northbrook, Stadion, Wobniar, Pinnacle Mountain |
| Microsoft | Hyperscaler (owns & consumes) | Public (NASDAQ: MSFT) | Mixed | Self-consumed | Project-name / generic 'Holdings LLC'; less consistently codenamed |
| Amazon (AWS) | Hyperscaler (owns & consumes) | Public (NASDAQ: AMZN) | Own | Self-consumed | Vadata, Inc. — AWS's build/operating entity on many permits |
| xAI (Colossus) | AI / neocloud | Private (xAI) | Mixed | Sometimes | Site/retrofit LLCs (e.g. former Electrolux plant, Memphis) |
| OpenAI · Oracle (Stargate) | AI / neocloud | JV: OpenAI · Oracle · SoftBank; sites developed by Oracle/Crusoe/Vantage | Mixed | Sometimes | Per-site developer LLCs (Crusoe/Lancium at Abilene, etc.) |
| CoreWeave | AI / neocloud | Public (NASDAQ: CRWV, IPO Mar 2025) | Lease | Rarely | Mostly TENANT in partners' buildings: Core Scientific, Chirisa/Bulk, Flexential, Lincoln, TierPoint, Digital Realty, DataBank, Switch, Galaxy, Applied Digital |
| Digital Realty | Colocation / wholesale REIT | Public REIT (NYSE: DLR) | Mixed | Rarely | 'Digital [City]/[Address] LLC'; legacy Telx entities; Ascenty (LatAm) |
| QTS | Colocation / wholesale REIT | Blackstone take-private ~$10B, closed 2021 | Own | Rarely | 'QTS ... LLC' + Blackstone holding entities; early generic project LLCs |
| Vantage | Colocation / wholesale REIT | $9.2B equity round closed Jun 2024 | Own | Rarely | Per-campus 'Vantage ... LLC' site entities |
| CyrusOne | Colocation / wholesale REIT | ~$15B take-private, closed 2022 | Own | Rarely | 'CyrusOne LLC' subsidiaries |
| Aligned | Colocation / wholesale REIT | ~$40B acquisition announced Oct 2025, close H1 2026 (press) | Own | Rarely | Per-site LLCs |
| Switch | Colocation / wholesale REIT | ~$11B EV take-private, closed 2022 | Own | Rarely | 'Switch ... LLC'; campuses branded 'Primes' |
| Stack Infrastructure | Colocation / wholesale REIT | Blue Owl acquired IPI Partners, 2025 | Own | Rarely | Per-site LLCs |
| EdgeConneX | Colocation / wholesale REIT | EQT-backed | Own | Rarely | Per-site LLCs |
| Equinix | Colocation / wholesale REIT | Public REIT (NASDAQ: EQIX) | Mixed | Rarely | IBX-branded site entities |
| Core Scientific | Colocation / wholesale REIT | Public (NASDAQ: CORZ); CoreWeave acquisition proposed Jul 2025 — outcome uncertain, verify | Own | Sometimes | Site LLCs; ~10 US facilities (~840 MW HPC + ~500 MW crypto; ~1.3 GW gross) |
⏳ No verification date recorded · changes over months · Source: Operator ownership & filing-LLC naming patterns — compiled from operator releases and county filings, then checked against SEC EDGAR full-text search. Composite, not a single citation: look the entity up yourself before naming it in a filing
Each row carries its own attribution level — first-party rows come from the operator's own site, but colocation and REIT ownership was compiled from trade press and never passed an independent fact-check. Check the attribution column before citing a row.
LLC resolution: county assessor / GIS parcel → grantee LLC on the deed → state Secretary-of-State business database → registered agent & principals.
Sources: Operator ownership & filing-LLC naming patterns — compiled from operator releases and county filings, then checked against SEC EDGAR full-text search. Composite, not a single citation: look the entity up yourself before naming it in a filing · Vantage Data Centers — $9.2B equity investment led by DigitalBridge & Silver Lake (closed Jun 2024) · Switch, Inc. — DigitalBridge & IFM Investors complete $11B take-private, Dec 6 2022 (first-party) · CoreWeave to acquire Core Scientific — all-stock ~$9B (announced Jul 2025)
ERCOT large-load interconnection queue
ERCOT is the only ISO that publishes an aggregate large-load (not generation) interconnection picture and breaks out the data-center share. The story is the funnel: what’s requested dwarfs what’s approved, which dwarfs what’s actually running. Data as of March 26, 2026.
- Seeking interconnection (410,000 MW) — Total large-load capacity requesting to connect. ~87% is data centers. Mostly speculative — many projects never energize. ERCOT — Large Load Update to Senate Committee on Business & Commerce (Apr 1 2026; data as of Mar 26 2026)
- Approved to energize (9,042 MW) — Received Approval to Energize from ERCOT Operations — cleared to turn on. ERCOT — Large Load Interconnection status, March TAC report (Mar 13 2026)
- Actually operational (3,801 MW) — Simultaneous monthly peak ERCOT has actually had to serve — real load on the grid right now (non-simultaneous observed peak: 3,883 MW). ERCOT — Large Load Interconnection status, March TAC report (Mar 13 2026)
The bar chart uses a linear scale; the “seeking interconnection” bar dwarfs the others because most of that capacity is speculative. Many projects never energize.
Sources: ERCOT — Large Load Update to Senate Committee on Business & Commerce (Apr 1 2026; data as of Mar 26 2026) · ERCOT — Large Load Interconnection status, March TAC report (Mar 13 2026) · ERCOT — Large Load Integration (forms, reports & Batch Zero)
SEC 10-K competitor analysis
What the six largest cloud / AI filers tell the SEC about who they compete with — a useful cross-reference of how these companies see each other.
| Filer | Names rivals? | 10-K language | Named | Editorial cross-ref |
|---|---|---|---|---|
| Alphabet (Google) | No | “developers and providers of AI products and services”; also “providers of enterprise cloud services” and consumer-hardware makers. | — | OpenAI, Anthropic, Microsoft, Amazon (AWS), Meta, xAI |
| Meta | No | “companies in the development and application of AI, particularly with respect to the development of frontier AI models.” | — | OpenAI, Google (Alphabet), Anthropic, xAI, Microsoft |
| Microsoft | No | “AI-first application companies”; Azure’s “AI offerings compete with AI products from hyperscalers … and … open source offerings.” | — | Google, Amazon (AWS), OpenAI, Anthropic, Meta, Oracle, Salesforce |
| Amazon (AWS) | No | competition intensified by “practical applications of artificial intelligence and machine learning” in “web and infrastructure computing services.” | — | Microsoft (Azure), Google Cloud, Oracle (OCI), CoreWeave, NVIDIA, OpenAI, Anthropic |
| Oracle | Yes | cloud/software/hardware offerings “compete directly with … Alphabet Inc., Amazon.com, Inc., … IBM … Microsoft Corporation, Salesforce, Inc. and SAP SE …” | Adobe, Alphabet, Amazon.com, Cisco, Intel, IBM, Microsoft, Salesforce, SAP, HPE, Workday | Amazon (AWS), Microsoft (Azure), Alphabet (Google Cloud), IBM, CoreWeave |
| CoreWeave | No | “We primarily compete with hyperscalers … several of which are also customers … We also compete with smaller cloud service providers.” | — | Amazon (AWS), Microsoft (Azure), Google Cloud, Oracle (OCI), Lambda, Nebius |
Sources: Alphabet Inc. — Form 10-K FY2024, Item 1 “Competition” (SEC EDGAR) · Meta Platforms, Inc. — Form 10-K FY2025, Item 1 “Competition” (SEC EDGAR) · Microsoft Corp. — Form 10-K FY2025, Item 1 “Competition” (SEC EDGAR) · Amazon.com, Inc. — Form 10-K FY2025, Item 1 “Competition” (SEC EDGAR) · Oracle Corp. — Form 10-K FY2026, Item 1 “Competition” (SEC EDGAR) · CoreWeave, Inc. — Form 10-K FY2025, Item 1 “Competition” (SEC EDGAR)
EIA pilot survey: In March 2026 the EIA launched its first-ever pilot survey of energy use at data centers — the federal government’s first attempt to measure an industry that currently self-reports nothing. EIA — Pilot survey on energy use at data centers (Mar 2026) · EIA-930 — hourly net generation by fuel type (API v2, free key)
The demand wave — ERCOT & PJM
The two ISOs with the most data-center load growth tell opposite stories about how fast the grid can absorb it.
ERCOT has ~233 GW of large-load interconnection requests, 72.9% data centers. Only a fraction will energize — see the funnel above.
PJM projects +32 GW of peak-load growth over the next decade, 94% from data centers. Dominion Energy Virginia’s zone alone shows 23.9 GW of load, up 23% from 2019.
Sources: ERCOT — Large Load Integration (forms, reports & Batch Zero) · PJM — 2025 Long-Term Load Forecast (data-center-driven growth) · EIA — Commercial electricity sales in Virginia driven by data centers (2025)
How the grid operators & FERC are responding
The through-line: co-location (data centers siting next to power plants to bypass grid queues) and cost allocation (who pays for the transmission upgrades these loads require).
FERC, PJM, ERCOT, MISO & SPP — what each is doing
FERC
In December 2025 FERC directed PJM to write co-location rules for data centers siting next to power plants. In June 2026 FERC issued show-cause orders to MISO, SPP, and other RTOs requiring them to adopt large-load interconnection standards. FERC — directs PJM to write co-location rules for data centers (Dec 18, 2025 fact sheet; Docket EL25-49/AD24-11) · FERC — show-cause orders to MISO, SPP & other RTOs on large-load interconnection (Jun 18, 2026)
PJM
The 2025 capacity auction cleared at record prices (~$16.4B total across 13 states), driven in large part by data-center load growth. PJM is developing new co-location rules per FERC’s order. Monitoring Analytics (PJM market monitor) — data centers drove $6.3B of the record $16.4B 2025 capacity auction (38%), and $29.4B of $63.6B across the last four auctions (46%); see the 2025/2026 RPM Base Residual Auction analyses
ERCOT / Texas SB 6
Texas SB 6 (signed 2025) authorizes ERCOT to curtail or disconnect large loads (≥75 MW) in emergencies, establishes new interconnection standards, and imposes a $50,000/MW fee on new large-load interconnections. Texas SB 6 (89R, signed June 20, 2025) — ERCOT may curtail/disconnect large loads (≥75 MW) in emergencies; new interconnection standards. Official bill text & history · Texas Legislature — SB 6 (89R), large-load grid rules, signed June 20, 2025. Official bill text & history
MISO
MISO launched its Large Load Interconnection Reliability (LLIR) process, requiring 90-day reliability studies for loads ≥100 MW before interconnection. MISO — Large Load Additions & Large Load Interconnection Reliability Requirements (2025–26)
SPP
SPP adopted its High Impact Large Load (HILL) 90-day study process for loads that would materially change system conditions. SPP — High Impact Large Load (HILL) integration, the 90-day study process from RR696; FERC accepted the tariff revisions effective Jan 15 2026 (first-party)
What to do with this
Use the state data to benchmark your community, the operator table to identify who you’re negotiating with, and the queue data to gauge how much more is coming.
Open the toolkit — meeting prep & CBA templates Your state briefing Executives & megaprojects