Skip to content
Get my AI grade

Data center map

Where America's AI data centers are being built, mapped

Every reviewed planned AI data center in the US, summed by state, with its owner, the year it is due and the price of power where it is going. Epoch AI and EIA data, analysed and recorded by Rank.ai.

Updated Oct 20, 2026Sources and methodDownload CSVCite this map

Recorded Oct 20, 2026

33.7GW

of AI data center capacity is planned across 20 US states, at 39 reviewed sites.

Source: Epoch AI (CC BY 4.0), summed by state in Rank.ai's planned capacity board.

Planned facility power by state

  • Under 500 MW
  • 500 MW – 1 GW
  • 1–2 GW
  • 2–4 GW
  • 4 GW +
  • None published

Point at a state for its record; click to open its evidence page.Tap a state for its record.

Key findings

5 findings from the map, each ready to share

Every chart has its own link, embed code and PNG, and says whose data it is. The numbers move when the data does.

Rank.ai original data#

1 of 24answers naming Oracle Cloud Infrastructure

Oracle plans 8.3 GW of US AI data center capacity, more than any other cloud. ChatGPT, Claude and Gemini named Oracle Cloud Infrastructure in 1 of 24 GPU cloud recommendations.

Capacity by owner from Epoch AI, Oct 20, 2026. Recommendations from the 24 answers Rank.ai recorded to GPU cloud buying questions on Sep 17, 2026. The cloud named most was Runpod, in 14 of 24.

  1. OracleOracle Cloud Infrastructure8.3 GW1 of 24
  2. AmazonAmazon EC24.3 GW1 of 24
  3. GoogleGoogle Cloud4.1 GW10 of 24
  4. MicrosoftMicrosoft Azure3.7 GW9 of 24
  5. CoreWeave2.2 GW10 of 24

Owners are as Epoch AI labels them. A recommendation is an answer that names the cloud when asked which GPU cloud to use.

Source: Epoch AI, Data on AI data centers (CC BY 4.0); Rank.ai, Cloud GPU platform rank (recorded AI answers). Data as of Sep 17, 2026.

Rank.ai original data#

+3.5 GWsince Jul 25, 2026

Planned US AI data center capacity rose from 30.1 GW to 33.7 GW since Rank.ai began recording it on Jul 25, 2026

Each point is one Rank.ai publication of Epoch AI's planned facilities, 49 so far. Capacity leaves the total when a project is cancelled or switched on.

30.1 GW33.7 GW

Source: Rank.ai, Planned AI data center capacity ledger; Epoch AI, Data on AI data centers (CC BY 4.0). Data as of Oct 20, 2026.

Epoch AI data, Rank.ai analysis#

42.5%of 33.7 GW in 3 states

Texas, Wisconsin and Indiana hold 42.5% of planned US AI data center capacity

33.7 GW is planned across 20 states at 39 reviewed sites, as of Oct 20, 2026.

  1. Texas8 sites6.9 GW
  2. Wisconsin2 sites4.0 GW
  3. Indiana2 sites3.3 GW
  4. New Mexico2 sites2.7 GW
  5. Ohio4 sites2.7 GW
  6. Louisiana1 site2.2 GW
  7. Tennessee1 site2.1 GW
  8. Iowa2 sites1.9 GW
  9. Mississippi2 sites1.8 GW
  10. Michigan1 site1.4 GW

Planned facility power is estimated capacity. Planned does not mean permitted, financed, built or switched on.

Source: Epoch AI, Data on AI data centers (CC BY 4.0); Rank.ai, Planned AI data center capacity by state. Data as of Oct 20, 2026.

EIA and Epoch AI data, Rank.ai analysis#

8.25¢per kWh, vs 9.71¢ median

Planned AI data centers sit where industrial power averages 8.25¢ per kWh, 15.0% below the median state

Each state's industrial rate weighted by its planned megawatts, against 9.71¢ for the median state. Rates are EIA statewide averages published Aug 31, 2026.

  1. Texas6.9 GW7.21¢
  2. Wisconsin4.0 GW10.88¢
  3. Indiana3.3 GW9.28¢
  4. New Mexico2.7 GW5.08¢
  5. Ohio2.7 GW9.75¢
  6. Louisiana2.2 GW6.10¢
  7. Tennessee2.1 GW6.92¢
  8. Iowa1.9 GW9.31¢
  9. Mississippi1.8 GW7.57¢
  10. Michigan1.4 GW8.99¢

A statewide industrial average is a public benchmark, not a negotiated data center tariff. The ten states with the most planned capacity shown.

Source: US Energy Information Administration, Electric Power Monthly table 5.6.A; Epoch AI, Data on AI data centers (CC BY 4.0); Rank.ai, State power economics. Data as of Aug 31, 2026.

Epoch AI data, Rank.ai analysis#

60.3%due in 2028

60.3% of planned US AI data center capacity is scheduled for 2028

Each facility counts in the year of the milestone its planned power is estimated for. 20.3 GW of 33.7 GW falls in 2028.

  1. 1.5 GW2026
  2. 11.0 GW2027
  3. 20.3 GW2028
  4. 800 MW2029

Milestone dates are Epoch AI's estimates from developer announcements and permits.

Source: Epoch AI, Data on AI data centers (CC BY 4.0). Data as of Oct 20, 2026.

The record

The largest planned sites, and every state ranked

39 reviewed planned facilities in 20 states, as of Oct 20, 2026. Select a state for its evidence page.

Open the full board
Epoch AI data, Rank.ai analysis#

2.7 GWMicrosoft Fairwater Wisconsin

Microsoft Fairwater Wisconsin is the largest planned AI data center in the US, at 2.7 GW

The ten largest of 39 reviewed planned facilities, by the facility power estimated for their next milestone.

  1. Microsoft Fairwater WisconsinMicrosoft · WI · 20282.7 GW
  2. OpenAI Stargate New MexicoOracle · NM · 20282.5 GW
  3. Anthropic-Amazon New CarlisleAmazon · IN · 20282.3 GW
  4. Meta HyperionMeta · LA · 20282.2 GW
  5. Colossus 2SpaceXAI · TN · 20272.1 GW
  6. OpenAI Stargate ShackelfordOracle · TX · 20282.0 GW
  7. QTS Cedar RapidsOwner undisclosed · IA · 20271.6 GW
  8. OpenAI Stargate MichiganOracle · MI · 20281.4 GW
  9. Meta PrometheusMeta · OH · 20281.3 GW
  10. OpenAI Stargate WisconsinOracle · WI · 20281.3 GW

Names and owners are as Epoch AI labels them.

Source: Epoch AI, Data on AI data centers (CC BY 4.0). Data as of Oct 20, 2026.

Planned capacity by state

#StatePlannedSitesNamed ownersCapex, 2025 $Industrial power
1Texas6.9 GW8CoreWeave, Google, Microsoft, Oracle$189.9B7.21¢ / kWh
2Wisconsin4.0 GW2Microsoft, Oracle$119.9B10.88¢ / kWh
3Indiana3.3 GW2Amazon, Google$105.6B9.28¢ / kWh
4New Mexico2.7 GW2Meta, Oracle$74.6B5.08¢ / kWh
5Ohio2.7 GW4Google, Meta, Softbank$77.4B9.75¢ / kWh
6Louisiana2.2 GW1Meta$61.8B6.10¢ / kWh
7Tennessee2.1 GW1SpaceXAI$58B6.92¢ / kWh
8Iowa1.9 GW2Google$52.8B9.31¢ / kWh
9Mississippi1.8 GW2Amazon$57.7B7.57¢ / kWh
10Michigan1.4 GW1Oracle$37.4B8.99¢ / kWh
11Virginia1.0 GW3CoreWeave$26.6B9.98¢ / kWh
12Nebraska813 MW2Google$25.7B8.37¢ / kWh
13North Dakota520 MW1CoreWeave$15.2B8.12¢ / kWh
14Alabama465 MW1Meta$13.6B8.11¢ / kWh
15New York441 MW1AI XPV Platform$14.3B10.54¢ / kWh
16Utah411 MW1Meta$11.5B8.56¢ / kWh
17Pennsylvania336 MW2Amazon, CoreWeave$9.9B10.47¢ / kWh
18Wyoming249 MW1Meta$7.8B8.80¢ / kWh
19Oklahoma181 MW1CoreWeave$5.8B8.16¢ / kWh
20South Carolina168 MW1Meta$4.9B7.88¢ / kWh

Planned capacity is estimated from reviewed announcements, not metered load, and planned does not mean permitted, financed, built or switched on. Industrial power is the EIA statewide average retail rate, not a negotiated data center tariff.

Sources

Sources, method and how to cite this map

Free to use with attribution. Download the facilities and the states, or copy a citation.

Sources and method

Each state's figure is the sum of Epoch AI's latest future-milestone facility power estimates for the distinct reviewed facilities it attributes to that state. Facilities without one explicit US state in their source record are left out, not geocoded. States without a reviewed planned project stay unshaded: missing evidence is unknown, never zero.

Projects leave the map when they are cancelled and when they switch on. Rank.ai keeps every publication of the board, and the capacity line above is drawn from that ledger.

Publication requires at least 50% mean confidence. At least 10 entities must clear the quality gate.

Read the full method

Download the data

  • Planned facilities (CSV)

    39 facilities: owner, state, planned MW, milestone date, capex, H100 equivalents.

  • States (CSV)

    All 50 states and DC: planned MW, sites, owners, capex, industrial power price.

Last updated October 20, 2026

Charts

Related charts, with their full history

All charts

See how often AI names you.

Enter your website. In about two minutes, Rank.ai asks ChatGPT, Claude and Gemini 12 questions your customers ask and grades how often they name you.

  • Your grade out of 100How often AI names you, cites your site, and how high it ranks you.
  • Who gets namedEvery competitor in the answers, most named first.
  • The pages AI readsThe sources behind each answer.
  • Three fixesWhat to fix first, with a brief for the first article.