The US data-center build-out, mapped and modelled

Where AI data centers are being built, how much power they need, who is building them, and how much capacity comes online if the pipeline converts.

Operating49 GW3,172 facilities

Under construction70 GW400 projects

Planned276 GW1,199 projects

Construction spending$85B/yrAug 2026 · +73.2% y/y

Companies building1,725807 hyperscale sites

Pipeline (GW)0 · Circles: the ten largest markets, sized by pipeline. Click a state for detail.

The model

How much capacity comes online?

Move the sliders to set your own assumptions. Defaults are our base case.

Share of under-construction capacity finished by end of 2028 Share of planned projects that actually get built Years for planned projects to come online Average utilisation (load factor)

Starts from capacity with published power figures today (49 GW operating, 70 GW building, 276 GW planned). Projects under construction come online over the next two years and planned projects from year three. Only about 44% of pipeline projects publish a power figure, so these are lower-bound scales, not forecasts.

The build-out is still mostly ahead of us

About 49 GW of US data centers with published power figures are running today, and 70 GW is already under construction. If what is being built gets finished, published operating capacity grows about 2.4 times. A further 276 GW is only planned.

Power, not demand, decides how much gets built

The planned pipeline (276 GW) is roughly 4 times what is under construction. Most planned projects are waiting for grid connections, turbines or transformers. That is why the suppliers of power equipment have multi-year backlogs, and why many projects on this list will be late or never built.

A few states carry the build-out

Texas, Virginia and Pennsylvania hold about 43% of the published pipeline. Local power prices, permitting and water decide where projects land, so utilities and contractors in these states see the orders first.

New names are building the biggest campuses

Several of the largest pipelines belong to companies with little or no operating capacity today (Tract, Fermi America, SB Energy, Creekstone Energy). Many are power-first developers. Their projects depend on financing, so watch credit conditions as closely as AI demand.

What it means for investors

More construction supports electrical equipment makers (transformers, switchgear, turbines), cooling suppliers and the contractors who build the halls. The main risks are projects that never convert from planned to built, rising power prices that bring political pushback, and developers that can't fund their plans.

Data center construction spending ($B, annual rate)

Money spent each month building US data centers (buildings, power and cooling, not servers), annualised. Source: US Census Bureau, Construction Spending (Value of Private Construction Put in Place, "Data center").

Power construction spending ($B, annual rate)

Power plants, transmission and substations: the grid the data centers need. Source: US Census Bureau, Total Construction Spending: Power (TLPWRCONS), via FRED.

Who is building the most

Pipeline GW

Company	Pipeline	Building	Operating	Sites
Tract	19.4	2.0	0.0	29
Amazon AMZN	18.5	5.7	5.3	263
Fermi America	17.0	0.0	0.0	1
SB Energy	11.4	2.2	0.0	3
Creekstone Energy	10.0	0.0	0.0	1
Meta META	9.0	5.9	5.6	48
Nscale Ltd.	8.0	0.0	0.0	1
New Era Energy & Digital	7.0	0.0	0.0	1
PowerHouse Data Centers	6.9	1.7	0.0	9
Google GOOGL	6.6	5.2	0.9	103
Vantage Data Centers	5.4	3.6	0.7	53
Microsoft MSFT	5.2	1.6	3.8	117
Energy Abundance	5.0	0.0	0.0	1
Sailfish Digital Ventures	5.0	0.0	0.0	1
Roswell Information Park LLC	5.0	0.0	0.0	1
TransGas Development Systems LLC	4.8	0.0	0.0	2
Homer City Redevelopment	4.5	4.5	0.0	1
Crusoe	4.3	2.8	0.2	9
QTS	4.1	0.6	2.0	99
Joule Capital Partners	4.0	0.0	0.0	1

Can be the operator, owner, developer or tenant. Companies without a ticker are private.

The big markets

Pipeline GW

Market	Pipeline	Building	Operating	Sites
Northern Virginia VA	11.0	4.2	7.8	461
Dallas–Fort Worth TX	8.2	1.8	2.6	197
Phoenix AZ	8.0	0.8	1.8	160
Atlanta GA	7.4	0.6	1.7	146
Chicago IL	5.8	0.7	1.1	139
Columbus OH	2.3	1.3	0.7	99
Salt Lake City UT	2.3	0.2	0.7	53
Silicon Valley CA	1.3	0.1	1.0	146
Las Vegas NV	0.6	0.0	0.2	39
Portland–Hillsboro OR	0.5	0.0	0.7	57

Top counties

County	Pipeline	Operating	Sites
Pecos County, TX	7.8	0.5	7
Maricopa County, AZ	6.5	1.8	152
Hood County, TX	6.5	0.3	5
Storey County, NV	5.4	0.2	14
Caldwell County, TX	5.1	0.0	5
Webb County, TX	5.0	0.0	2
Louisa County, VA	3.8	0.0	4
Lackawanna County, PA	3.3	0.0	8
Loudoun County, VA	3.2	5.4	244
Coweta County, GA	3.1	0.0	3
Lake County, IN	3.0	0.0	12
Dallas County, TX	2.9	0.5	100

Every state

Sortable

State	Pipeline GW	Building GW	Operating GW	Current sites	Hyperscale	Power ¢/kWh	Water stress
Texas	93.1	19.5	10.5	561	71	6.1	High
Virginia	31.6	8.3	9.0	591	194	9	Low
Pennsylvania	23.1	8.1	0.3	132	9	7.9	Medium-High
Georgia	19.8	1.2	2.5	232	21	7.2	Medium-High
Utah	19.3	0.2	0.7	69	20	8.2	High
Ohio	17.6	3.8	1.2	238	54	7.1	Medium-High
New Mexico	15.7	2.5	0.4	28	8	5.4	High
West Virginia	13.4	0.0	0.0	27	8	7.8	Medium-High
Arizona	10.6	1.4	1.8	176	29	7.9	Extremely High
Illinois	10.6	1.9	1.3	164	8	8.8	High
Indiana	8.6	4.1	1.6	106	39	8.1	Medium-High
Kentucky	7.9	0.9	0.3	45	11	6.5	Low
Nevada	7.7	2.8	0.7	73	14	8.6	Low
New York	6.1	0.4	0.6	168	18	9.2	Low
Louisiana	5.8	2.8	0.0	28	8	5.6	Low
Wyoming	5.1	1.9	0.3	39	22	8	Extremely High
Missouri	4.6	1.9	0.2	88	8	7.9	Low
Iowa	4.4	1.4	1.8	67	22	6.8	Low
Oklahoma	3.9	0.3	0.5	60	18	5.8	Medium-High
North Dakota	3.4	1.0	1.5	37	7	7.2	Low
Alabama	2.4	0.4	0.2	31	8	7.3	Low-Medium
California	2.2	0.2	1.4	343	14	21.5	Low-Medium
Michigan	2.2	1.0	0.4	75	15	8.3	Low
South Carolina	2.2	0.0	0.1	32	10	6.9	High
Kansas	2.2	0.0	0.0	36	0	7.7	Low-Medium
North Carolina	2.1	0.6	0.8	99	10	7.8	High
Nebraska	2.1	0.0	0.5	50	9	7.7	Low
Arkansas	2.1	0.4	0.0	22	4	6.6	Low
Maine	2.0	0.0	0.0	15	3	12.1	Low
Colorado	1.9	0.0	0.2	87	10	8.6	Extremely High
Maryland	1.5	0.4	0.3	62	12	9.9	Low-Medium
Washington	1.4	0.0	1.7	117	16	6.6	Low-Medium
Minnesota	1.4	0.4	0.1	89	11	9.1	Low
Wisconsin	1.3	1.1	1.5	54	15	8.5	Medium-High
Montana	1.3	0.0	0.1	19	0	7.6	Medium-High
Oregon	1.2	0.1	2.3	101	30	8.1	Low
Alaska	1.1	0.0	0.0	20	0	19.3	Low
Florida	0.8	0.0	0.1	141	6	8.5	High
New Jersey	0.6	0.3	0.6	103	4	11.8	Medium-High
South Dakota	0.6	0.0	0.2	17	4	8.3	High
Mississippi	0.4	0.2	1.1	36	15	6.8	Medium-High
Tennessee	0.3	0.1	1.4	74	9	6.2	Low
Idaho	0.1	0.0	0.2	22	3	7.7	Extremely High
Connecticut	0.1	0.0	0.2	31	6	17.1	Low
Massachusetts	0.0	0.0	0.1	56	2	18.2	Medium-High
Puerto Rico	0.0	0.0	0.1	11	0	23.7	Low
Hawaii	0.0	0.0	0.0	23	0	34.1	—
Delaware	0.0	0.0	0.0	16	1	9	Medium-High
New Hampshire	0.0	0.0	0.0	24	0	16.2	Low-Medium
Vermont	0.0	0.0	0.0	9	0	11.6	Low
District of Columbia	0.0	0.0	0.0	13	0	10.8	Low
Rhode Island	0.0	0.0	0.0	14	1	19.7	Medium-High

Stocks most exposed to the build-out

VRT+3.2%ETN+2.4%GEV+5.0%PWR+4.1%EME+2.8%FIX+8.1%STRL+3.6%CEG+0.5%VST+2.1%EQIX+1.2%DLR+0.3%IREN+5.0%APLD+1.0%CRWV+4.5%

Sources

Facilities, pipeline, states, markets, counties and companies: US Data Center Map (dcmap.us), data as of 2026-10-12. Copied to our database once a month. Methodology.
Data center construction spending: US Census Bureau, Construction Spending, Value of Private Construction Put in Place, seasonally adjusted annual rate, "Data center" line.
Power construction spending: US Census Bureau, Total Construction Spending: Power (TLPWRCONS), via FRED, Federal Reserve Bank of St. Louis.
County locations: US Census Bureau Gazetteer files (2024). State map: US Census Bureau cartographic boundaries.
US electricity use (model comparison): about 4,100 TWh in 2024, US Energy Information Administration.

Power totals combine published or announced figures, which use different measures (IT load, utility capacity, campus total) and count shared campus figures once. The model, map and commentary are by Singularity Research. Not investment advice.