THE BUILD · a SIREN sub-index · does not feed the main number
Where are the datacentres, and what is the power and water around them doing?
Every datacentre OpenStreetMap has mapped in the United States, plus the ones it has mapped as under construction or proposed, joined to the drought in their county, the nearest river gauge /watts reads, and the grid they sit on.
This is a map of buildings, not of electricity. 1,894 datacentres have been mapped in the United States by volunteers; 1,050 of them — 55% — stand in a county the US Drought Monitor placed in drought (D1 or worse) this week, and 376 sit on one of the three grids that publish a demand number without an account. What any of these buildings actually draws is not published by anyone.
Nothing here measures a datacentre’s power or water draw. No public feed publishes either, for any site, at any cadence. What is measured is where the buildings are and what the substrate around them is doing — which is a smaller claim, and a true one.
A state with no pins is a state nobody has mapped. It is not a state with no datacentres. OpenStreetMap coverage is volunteer-maintained and uneven. Virginia has 411 pins and Wyoming has 23; some of that is the build-out and some of it is who maps.
Capacity is tagged on 51 of 1,894 sites, totalling 2,830.1 MW of tagged IT power. Any sum over this dataset is a sum over two per cent of it, which is why no total megawatt figure is printed anywhere on this page.
“Operating” means the building exists. It is not a claim that there are servers in it, that they are switched on, or that they are training anything. A decommissioned datacentre keeps its tag until somebody visits. A map is extremely persuasive, and that is exactly why it has to say all of this before it draws anything.
3 of 11 balancing authorities publish a number without an account. These are the only figures on this page that can differ from the ones you saw last time.
There is no new-sites line here, and that is deliberate.
first_seen_at spans 2026-09-30 10:25:08 UTC to 2026-10-15 05:25:11 UTC in this file.
The pins move when OpenStreetMap moves, which is on volunteer time, not on a cron.
Three controls sit on top of this picture and none of them needs JavaScript: shape, colour, and one of 49 jurisdictions. They dim the map rather than cutting it, the count beside every chip is the count you get if you press it, and the figure in the readout is already in this HTML for all 16 combinations — so it cannot lag the control and cannot be wrong in a screenshot.
1,887 of 1,887 pins 1.0×
Drag or swipe to pan, wheel or pinch to zoom, hover or tap a marker for its record. Keyboard: tab to the map, then arrows pan, + and − zoom, 0 resets; tab once more to step marker to marker with the arrows, Enter for the record and Esc to leave. The table below carries every row whether or not any of this works.
Distinguishable in greyscale and in a screenshot. These three are the map's small-scale form of the same three statuses the rest of this page marks with the site's own icons: a datacentre pin is under five pixels across at national scale, which is below the size a drawn footprint survives at. Status is read from the OpenStreetMap tag that set it, and every pin carries that tag in the JSON.
US Drought Monitor, week of 2026-10-06. Every count here is also printed in words, because a reader who cannot separate two of these hues still gets the number. 1,343 of 1,894 pins sit in a county carrying a category. Colour is the drought in that county.
Size is capacity, where it is published. A pin is 2.4 units across unless OpenStreetMap carries an IT-power tag for it, in which case its radius is 3 + 0.55 × √MW, capped at 10, and it is drawn half-transparent with a ring. No pin is ever sized by a guess: a site with no capacity tag gets the fixed dot.
matching the filter 1,8941,34365071,7781,26761709562290211447 of 1,894 mapped
Pick a state above, or press a row in the table, to open its dossier here: the status split, the drought under its pins, every balancing authority its pins sit on with that authority's live number or the named reason there is not one, and whether anybody tagged a megawatt figure. Every mapped site in this build fell inside the drawn frame, so nothing is listed away from the map.
matching the filter 4113811903703501903525006600 of 411 in this state
51 of 411 sites carry an IT-power tag, totalling 2,830.1 MW. That is a sum over 12.4% of this state and is not the state's datacentre load.
matching the filter 2362352191219219203016161601001 of 236 in this state
No site in Texas carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 13513594012612690011008840 of 135 in this state
No site in Oregon carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 1172600116260010000000 of 117 in this state
No site in California carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 935009050030000000 of 93 in this state
No site in Ohio carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 757511071718044300000 of 75 in this state
No site in Washington carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 7210055100170000000 of 72 in this state
No site in Illinois carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 6464630646463000000000 of 64 in this state
No site in Arizona carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 63634057574066000000 of 63 in this state
No site in Georgia carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 615006150000000000 of 61 in this state
No site in New Jersey carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 600006000000000000 of 60 in this state
No site in Iowa carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 464004640000000000 of 46 in this state
No site in Pennsylvania carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 409043690000004004 of 40 in this state
No site in New York carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 3939230383822011100000 of 39 in this state
No site in Nevada carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 2828240282824000000000 of 28 in this state
No site in Tennessee carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 2727270272727000000000 of 27 in this state
No site in Colorado carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 26269026269000000000 of 26 in this state
No site in North Carolina carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 251002510000000000 of 25 in this state
No site in Nebraska carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 2323230232323000000000 of 23 in this state
No site in Wyoming carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 2221211212121000001001 of 22 in this state
No site in Utah carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 1818150151512033300000 of 18 in this state
No site in South Carolina carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 1717150171715000000000 of 17 in this state
No site in New Mexico carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 1616120161612000000000 of 16 in this state
No site in Florida carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 160001500010000000 of 16 in this state
No site in Indiana carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 1616110161611000000000 of 16 in this state
No site in Minnesota carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 164001640000000000 of 16 in this state
No site in Missouri carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 1414110131310011100000 of 14 in this state
No site in Alabama carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 13130013130000000000 of 13 in this state
No site in Maryland carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 139011290000001001 of 13 in this state
No site in Michigan carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 1313120131312000000000 of 13 in this state
No site in North Dakota carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 131101211010000000 of 13 in this state
No site in Wisconsin carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 1111110111111000000000 of 11 in this state
No site in Oklahoma carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 1010100555055500000 of 10 in this state
No site in Mississippi carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 7210721000000000 of 7 in this state
No site in Kansas carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 5100510000000000 of 5 in this state
No site in Connecticut carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 5400540000000000 of 5 in this state
No site in Kentucky carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 4440444000000000 of 4 in this state
No site in Idaho carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 4440444000000000 of 4 in this state
No site in Louisiana carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 3300330000000000 of 3 in this state
No site in Massachusetts carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 3330333000000000 of 3 in this state
No site in Maine carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 3210321000000000 of 3 in this state
No site in Montana carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 3300330000000000 of 3 in this state
No site in New Hampshire carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 2220222000000000 of 2 in this state
No site in Arkansas carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 2200220000000000 of 2 in this state
No site in South Dakota carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 2100210000000000 of 2 in this state
No site in West Virginia carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 1100110000000000 of 1 in this state
No site in District of Columbia carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 1100110000000000 of 1 in this state
No site in Vermont carries an IT-power tag, so no megawatt figure is printed for it. That is a statement about OpenStreetMap's tagging, not about the buildings.
matching the filter 0000000000000000 of 0 in this state
A state with no pins is a state nobody has mapped. It is not a state with no datacentres.
matching the filter 0000000000000000 of 0 in this state
A state with no pins is a state nobody has mapped. It is not a state with no datacentres.
This screen is narrower than 760 pixels, so the map is not drawn. 1,887 pins on a 343-pixel map is a smudge, and a smudge that cannot be read still looks authoritative. The controls above and the table below carry every channel the map encodes except position.
Every state in the contiguous frame, ranked by mapped sites. The states at the bottom with a zero are the point of this table.
| State | Sites | Filter | Worst drought under a pin | Grid most pins sit on | Op | Bld | Ann |
|---|---|---|---|---|---|---|---|
| 411 | 4113811903703501903525006600 | D3 extreme381 of 411 in a county with a category | PJMno keyless demand feed | 370 | 35 | 6 | |
| 236 | 2362352191219219203016161601001 | D4 exceptional235 of 236 in a county with a category | ERCOTlive demand feed · 2 grids in state | 219 | 16 | 1 | |
| 135 | 13513594012612690011008840 | D3 extreme135 of 135 in a county with a category | BPAno keyless demand feed | 126 | 1 | 8 | |
| 117 | 1172600116260010000000 | D1 moderate26 of 117 in a county with a category | CAISOlive demand feed · 2 grids in state | 116 | 1 | 0 | |
| 93 | 935009050030000000 | D0 abnormally dry5 of 93 in a county with a category | PJMno keyless demand feed | 90 | 3 | 0 | |
| 75 | 757511071718044300000 | D3 extreme75 of 75 in a county with a category | BPAno keyless demand feed | 71 | 4 | 0 | |
| 72 | 7210055100170000000 | D1 moderate1 of 72 in a county with a category | MISO / SPPno keyless demand feed | 55 | 17 | 0 | |
| 64 | 6464630646463000000000 | D2 severe64 of 64 in a county with a category | WECC — vertically integrated utilitiesno keyless demand feed | 64 | 0 | 0 | |
| 63 | 63634057574066000000 | D2 severe63 of 63 in a county with a category | Southern Co. / TVA / vertically integratedno keyless demand feed | 57 | 6 | 0 | |
| 61 | 615006150000000000 | D0 abnormally dry5 of 61 in a county with a category | PJMno keyless demand feed | 61 | 0 | 0 | |
| 60 | 600006000000000000 | none0 of 60 in a county with a category | MISO / SPPno keyless demand feed | 60 | 0 | 0 | |
| 46 | 464004640000000000 | D0 abnormally dry4 of 46 in a county with a category | PJMno keyless demand feed | 46 | 0 | 0 | |
| 40 | 409043690000004004 | D0 abnormally dry9 of 40 in a county with a category | NYISOlive demand feed | 36 | 0 | 4 | |
| 39 | 3939230383822011100000 | D2 severe39 of 39 in a county with a category | WECC — vertically integrated utilitiesno keyless demand feed | 38 | 1 | 0 | |
| 28 | 2828240282824000000000 | D3 extreme28 of 28 in a county with a category | Southern Co. / TVA / vertically integratedno keyless demand feed | 28 | 0 | 0 | |
| 27 | 2727270272727000000000 | D4 exceptional27 of 27 in a county with a category | WECC — vertically integrated utilitiesno keyless demand feed | 27 | 0 | 0 | |
| 26 | 26269026269000000000 | D2 severe26 of 26 in a county with a category | PJMno keyless demand feed | 26 | 0 | 0 | |
| 25 | 251002510000000000 | D0 abnormally dry1 of 25 in a county with a category | MISO / SPPno keyless demand feed | 25 | 0 | 0 | |
| 23 | 2323230232323000000000 | D3 extreme23 of 23 in a county with a category | WECC — vertically integrated utilitiesno keyless demand feed | 23 | 0 | 0 | |
| 22 | 2221211212121000001001 | D3 extreme21 of 22 in a county with a category | WECC — vertically integrated utilitiesno keyless demand feed | 21 | 0 | 1 | |
| 18 | 1818150151512033300000 | D2 severe18 of 18 in a county with a category | Southern Co. / TVA / vertically integratedno keyless demand feed | 15 | 3 | 0 | |
| 17 | 1717150171715000000000 | D3 extreme17 of 17 in a county with a category | WECC — vertically integrated utilitiesno keyless demand feed | 17 | 0 | 0 | |
| 16 | 1616120161612000000000 | D3 extreme16 of 16 in a county with a category | Southern Co. / TVA / vertically integratedno keyless demand feed | 16 | 0 | 0 | |
| 16 | 160001500010000000 | none0 of 16 in a county with a category | PJMno keyless demand feed | 15 | 1 | 0 | |
| 16 | 1616110161611000000000 | D3 extreme16 of 16 in a county with a category | MISO / SPPno keyless demand feed | 16 | 0 | 0 | |
| 16 | 164001640000000000 | D1 moderate4 of 16 in a county with a category | MISO / SPPno keyless demand feed | 16 | 0 | 0 | |
| 14 | 1414110131310011100000 | D3 extreme14 of 14 in a county with a category | Southern Co. / TVA / vertically integratedno keyless demand feed | 13 | 1 | 0 | |
| 13 | 13130013130000000000 | D1 moderate13 of 13 in a county with a category | PJMno keyless demand feed | 13 | 0 | 0 | |
| 13 | 139011290000001001 | D0 abnormally dry9 of 13 in a county with a category | PJMno keyless demand feed | 12 | 0 | 1 | |
| 13 | 1313120131312000000000 | D3 extreme13 of 13 in a county with a category | MISO / SPPno keyless demand feed | 13 | 0 | 0 | |
| 13 | 131101211010000000 | D2 severe1 of 13 in a county with a category | MISO / SPPno keyless demand feed | 12 | 1 | 0 | |
| 11 | 1111110111111000000000 | D4 exceptional11 of 11 in a county with a category | MISO / SPPno keyless demand feed | 11 | 0 | 0 | |
| 10 | 1010100555055500000 | D2 severe10 of 10 in a county with a category | MISO / SPPno keyless demand feed | 5 | 5 | 0 | |
| 7 | 7210721000000000 | D2 severe2 of 7 in a county with a category | MISO / SPPno keyless demand feed | 7 | 0 | 0 | |
| 5 | 5100510000000000 | D0 abnormally dry1 of 5 in a county with a category | ISO New Englandno keyless demand feed | 5 | 0 | 0 | |
| 5 | 5400540000000000 | D1 moderate4 of 5 in a county with a category | PJMno keyless demand feed | 5 | 0 | 0 | |
| 4 | 4440444000000000 | D3 extreme4 of 4 in a county with a category | WECC — vertically integrated utilitiesno keyless demand feed | 4 | 0 | 0 | |
| 4 | 4440444000000000 | D3 extreme4 of 4 in a county with a category | MISO / SPPno keyless demand feed | 4 | 0 | 0 | |
| 3 | 3300330000000000 | D1 moderate3 of 3 in a county with a category | ISO New Englandno keyless demand feed | 3 | 0 | 0 | |
| 3 | 3330333000000000 | D3 extreme3 of 3 in a county with a category | ISO New Englandno keyless demand feed | 3 | 0 | 0 | |
| 3 | 3210321000000000 | D3 extreme2 of 3 in a county with a category | MISO / SPPno keyless demand feed | 3 | 0 | 0 | |
| 3 | 3300330000000000 | D1 moderate3 of 3 in a county with a category | ISO New Englandno keyless demand feed | 3 | 0 | 0 | |
| 2 | 2220222000000000 | D3 extreme2 of 2 in a county with a category | MISO / SPPno keyless demand feed | 2 | 0 | 0 | |
| 2 | 2200220000000000 | D1 moderate2 of 2 in a county with a category | MISO / SPPno keyless demand feed | 2 | 0 | 0 | |
| 2 | 2100210000000000 | D0 abnormally dry1 of 2 in a county with a category | PJMno keyless demand feed | 2 | 0 | 0 | |
| 1 | 1100110000000000 | D0 abnormally dry1 of 1 in a county with a category | PJMno keyless demand feed | 1 | 0 | 0 | |
| 1 | 1100110000000000 | D0 abnormally dry1 of 1 in a county with a category | ISO New Englandno keyless demand feed | 1 | 0 | 0 | |
| 0 | 0000000000000000 | A state with no pins is a state nobody has mapped. It is not a state with no datacentres. | 0 | 0 | 0 | ||
| 0 | 0000000000000000 | A state with no pins is a state nobody has mapped. It is not a state with no datacentres. | 0 | 0 | 0 | ||
Every mapped site in this build fell inside the drawn frame. A site outside it would be listed here in words rather than dragged to the edge of the picture.
A dot is not an argument. This is what is behind one, for each of the three grids that publish a number without an account, plus the grid carrying the most pins on this map — which publishes nothing. Each line is the site's own stable sentence, then the live grid reading for the grid it sits on, carrying that reading's state. A number awaiting its baseline is printed and labelled as one; it is never printed as a scored number.
ERCOT · 235 pins
Digital Realty Dallas DFW35 Digital Realty
Dallas County, TX · ERCOT · 68.1% of the county in D3 · nearest gauge Trinity Rv at Dallas, 23.6 km, ERCOT headroom used 75.80% of committed capacity AWAITING BASELINE
CAISO · 101 pins
EdgeConneX San Diego EdgeConneX
San Diego County, CA · CAISO · 65.5% of the county in D0 · nearest gauge Colorado Rv at Lees Ferry, 677.6 km, CAISO overnight floor 20,351 MW AWAITING BASELINE
NYISO · 40 pins
SU Green Data Center Syracuse University
Onondaga County, NY · NYISO · 83.1% of the county in D0 · nearest gauge Potomac Rv at Little Falls, 459.6 km, NYISO overnight floor 13,172 MW AWAITING BASELINE
PJM · 687 pins
Brandaleone Lab for Data and Visualization Duke Libraries
Durham County, NC · PJM · 48.2% of the county in D2 · nearest gauge Goose Creek near Leesburg, 356.2 km NO READING
PJM publishes real-time demand only through Data Miner 2, which requires a registered subscription key. Keys are free; secrets in this repo are not permitted.
Selection rule, so this is a sample and not a curation: one site per grid, the named site with a named operator whose county carries the worst drought category, ties broken by the stable site id. Nothing here is chosen for effect.
Drought is the resource this map can actually carry, because the US Drought Monitor publishes it by county and every pin has a county. It is the colour of every dot above. 1,343 of 1,894 pins stand in a county carrying a category on the map dated 2026-10-06, and 650 of those are in D2 or worse.
Drought figures are the share of a county’s AREA in a category. A county 20% in D3 does not say which 20%, and the Drought Monitor is a weekly expert judgement, not an instrument reading.
The same nine USGS gauges What does the SIREN infrastructure index read? reads, each one beside a named datacentre cluster, measured against that gauge's own long-run median for this day of the year. 633 of 1,894 pins have one within 50 km; 321 are regional and 933 are distant. A gauge 800 km away describes a different watershed, and every pin carries its distance for that reason.
| Gauge | % of median | Reading | State |
|---|---|---|---|
| Goose Creek near LeesburgLoudoun County · Data Center Alley itself; a small local stream | 1,307.5% | 1,207.5% above the median | LIVEday 2026-10-13 |
| Potomac Rv at Little FallsNorthern Virginia · the regional supply gauge for the DC metro | 516.4% | 416.4% above the median | LIVEday 2026-10-13 |
| Chattahoochee Rv near NorcrossAtlanta metro · metro Atlanta water supply | 81.1% | 18.9% below the median | LIVEday 2026-10-13 |
| Scioto Rv at ColumbusCentral Ohio · New Albany and the Columbus corridor | 123.5% | 23.5% above the median | LIVEday 2026-10-13 |
| Missouri River at Omaha, NEno cluster named | — | no reading on this build | LIVEday 2026-10-13 |
| Trinity Rv at DallasDallas–Fort Worth · ERCOT North; the densest Texas cluster | 136.1% | 36.1% above the median | LIVEday 2026-10-13 |
| Colorado Rv at AustinAustin–San Antonio · central Texas build-out; ERCOT South | 50.0% | 50.0% below the median | LIVEday 2026-10-13 |
| Colorado Rv at Lees FerryPhoenix metro · the accounting point for Lower Basin deliveries, which is how Central Arizona Project water reaches Phoenix; it is 300 miles upstream of the datacentres and is a supply gauge, not a local one | 74.2% | 25.8% below the median | LIVEday 2026-10-13 |
| Columbia Rv at The DallesThe Dalles · the Oregon cluster sits on this river | 81.3% | 18.7% below the median | LIVEday 2026-10-13 |
Flow is a statement about rainfall and reservoir operation, not about consumption. Nothing in this table is a datacentre's water draw, and no public feed publishes one.
Every pin is assigned to the balancing authority that runs its state's grid, with named county exceptions where that is wrong. Three of those authorities publish a real-time number without an account, and they cover 376 of 1,894 pins. The single largest block on this map — 687 pins on PJM — is not one of them.
A blank cell reads as an oversight. A named wall reads as a boundary of the instrument, which is what it is. Every grid below with no number says why in its own words, and those words are the same ones What does the SIREN infrastructure index read? prints for the same grid.
75.80 % of committed capacity
ERCOT headroom used · observed 2026-10-15 05:20:00 UTC
AWAITING BASELINEreal number, not yet scored20,351 MW
CAISO overnight floor · observed 2026-10-15 05:24:57 UTC
AWAITING BASELINEreal number, not yet scored13,172 MW
NYISO overnight floor · observed 2026-10-14 08:00:00 UTC
AWAITING BASELINEreal number, not yet scoredPJM publishes real-time demand only through Data Miner 2, which requires a registered subscription key. Keys are free; secrets in this repo are not permitted.
DARKno keyless feedMISO’s public data broker answered {"error": "no data"} on every message type tested on 2026-09-23. SPP has no equivalent keyless endpoint.
DARKno keyless feedBonneville publishes balancing-authority totals on a different cadence and format; not yet read by this collector.
DARKno keyless feedVertically integrated utilities, no organised market, no public demand feed.
DARKno keyless feedThe Southeast is not an organised market. There is no independent system operator here and therefore no public real-time demand feed.
DARKno keyless feedthis California load pocket is balanced by a municipal utility, not by CAISO
DARKno keyless feedISO-NE publishes real-time demand through a web-services account; no keyless endpoint was found.
DARKno keyless feedthis Texas county is outside ERCOT — El Paso is WECC, the far Panhandle is SPP, the far east is MISO
DARKno keyless feedA balancing authority is not a state, and this table treats it as one. a balancing authority is not a state; outside the named exceptions this assignment is an approximation, and a site near a seam may be on a different grid than the one printed. El Paso and Hudspeth, the far Panhandle, east Texas, and the Los Angeles, Sacramento and Imperial load pockets are excluded by county FIPS and carry the reason instead of a wrong answer.
The 12 biggest of the 51 sites carrying an IT-power tag — two per cent of the map. This is not a ranking of the largest datacentres in the United States. It is a ranking of the ones a volunteer typed a megawatt figure into. Each line is the site's own stable sentence, then the live grid reading for the grid it sits on, carrying that reading's state.
CloudHQ LC4 CloudHQ 180 MW
Loudoun County, VA · PJM · 36.6% of the county in D1 · nearest gauge Goose Creek near Leesburg, 9.3 km
operating · confidence high · way/1381994242
STACK’s Loudoun County campus STACK Infrastructure 180 MW
Loudoun County, VA · PJM · 36.6% of the county in D1 · nearest gauge Goose Creek near Leesburg, 7.9 km
announced · confidence high · node/14265747424
CloudHQ LC1 CloudHQ 144 MW
Loudoun County, VA · PJM · 36.6% of the county in D1 · nearest gauge Goose Creek near Leesburg, 10 km
operating · confidence high · way/794147654
CloudHQ LC12 CloudHQ 108 MW
Loudoun County, VA · PJM · 36.6% of the county in D1 · nearest gauge Goose Creek near Leesburg, 8.7 km
under construction · confidence high · way/1560822519
CloudHQ LC9 CloudHQ 108 MW
Loudoun County, VA · PJM · 36.6% of the county in D1 · nearest gauge Goose Creek near Leesburg, 9.4 km
announced · confidence high · node/14201428242
CloudHQ LC10 CloudHQ 108 MW
Loudoun County, VA · PJM · 36.6% of the county in D1 · nearest gauge Goose Creek near Leesburg, 9.2 km
announced · confidence high · node/14201428241
CloudHQ LC11 CloudHQ 108 MW
Loudoun County, VA · PJM · 36.6% of the county in D1 · nearest gauge Goose Creek near Leesburg, 8.7 km
under construction · confidence high · way/1560822517
CloudHQ LC13 CloudHQ 108 MW
Loudoun County, VA · PJM · 36.6% of the county in D1 · nearest gauge Goose Creek near Leesburg, 8.9 km
announced · confidence high · node/14201428243
CloudHQ LC14 CloudHQ 108 MW
Loudoun County, VA · PJM · 36.6% of the county in D1 · nearest gauge Goose Creek near Leesburg, 8.9 km
under construction · confidence high · way/1560822518
CloudHQ LC8 CloudHQ 108 MW
Loudoun County, VA · PJM · 36.6% of the county in D1 · nearest gauge Goose Creek near Leesburg, 9.6 km
under construction · confidence high · way/1560822523
CyrusOne NVA5 CyrusOne 105 MW
Loudoun County, VA · PJM · 36.6% of the county in D1 · nearest gauge Goose Creek near Leesburg, 12.4 km
operating · confidence high · way/435817145
Digital Realty IAD39 Digital Realty 96 MW
Loudoun County, VA · PJM · 36.6% of the county in D1 · nearest gauge Goose Creek near Leesburg, 9.6 km
operating · confidence high · way/701923203
Confidence is a statement about the record, not about the building.
| Source | Sites | State | Error |
|---|---|---|---|
| OpenStreetMap datacentre tags, via Overpassoperating and under-construction sites, with surveyed coordinates | 1,968 | LIVEvia cache | no error |
| Announcements extracted from the SIREN newsroomannounced sites, state-level, traceable to a scored news item id | 7 | LIVEvia local | no error |
| Federal Register documents naming a datacentre projectannounced sites named in a federal notice, state-level | 0 | LIVEvia network | no error |
| SEC filings saying "data centers", joined to operatorsoperator-level corroboration; no pins, no coordinates | 0 | LIVEvia network | no error |
| Census TIGERweb county boundaries | — | LIVEvia cache | no error |
| USGS site service — coordinates for the nine /watts gauges | 9 | LIVEvia cache | no error |
| US Drought Monitor, county statistics | — | LIVE | no error |
Two of the sources share one host, and the whole existing half of this map depends on a single volunteer-run Overpass instance. Three mirrors and a seven-day disk cache are the only redundancy there is. Sites geocode against Census county polygons simplified to about 550 metres, so a site on a county line can land in the neighbouring county; 66 pins carry that flag.
The Federal Register names no datacentre projects: of 437 documents whose full text carries the phrase over eighteen months, four have it in the title and not one of them is a project — a flight-data web portal, a paediatric emergency data repository, a permitting executive order and a listening session. SEC full-text search returns the filer's registered head office rather than the site, so it adds no pins and is used only to corroborate an operator. Full OpenStreetMap geometry would give a footprint area at roughly ten times the payload and a materially higher chance of the query timing out on a server that was already refusing requests; that trade was made in favour of the query completing, which is why there is no area field anywhere on this page.
Compiled 2026-10-15 05:25:11 UTC from
the same JSON this page reads.
Infrastructure readings from the 2026-10-15 05:24:56 UTC Watts run.
Drought map dated 2026-10-06, LIVE.
Streamflow day 2026-10-13, LIVE.
This does not feed the SIREN index and has no route into it:
collector/engine.mjs does not read this file, and none of these sources sits in the frozen
reference distribution the composite is scored against.
1,894 datacentres have been mapped in the United States by OpenStreetMap volunteers; 1,050 of them — 55% — stand in a county the US Drought Monitor placed in drought (D1 or worse) this week. This is a map of buildings, not of electricity.
Nothing here measures a datacentre's power or water draw. Capacity is tagged on a small share of sites, so no total megawatt figure is printed. "Operating" means the building exists — not that servers are in it.
Three controls sit on top of this picture and none of them needs JavaScript: shape, colour, and jurisdiction. 1,887 pins are on the map.
Drought is the resource this map can actually carry, because the US Drought Monitor publishes it by county and every pin has a county. 1,343 of 1,894 pins stand in a county carrying a drought category.
Every pin is assigned to the balancing authority that runs its state's grid. Three of those authorities publish a real-time number without an account.
A dot is not an argument. This is what is behind one, for each of the three grids that publish a number without an account, plus the grid carrying the most pins on this map.
A ranking of sites carrying an IT-power tag — a sample of a sample of the map, not a ranking of the largest datacentres in the United States.
Status and confidence are stated in words. Every limitation is published in full. This does not feed the main SIREN number.
The same nine USGS gauges the infrastructure index reads, each beside a named datacentre cluster, measured against that gauge's own long-run median for this day of the year. Flow is a statement about rainfall and reservoir operation, not about consumption.
Status marks on this map are stated in words in the method section — operating, planned, and the other recorded states — as labels for mapped buildings, not for server load.
Confidence is a statement about the record, not about the building. High, medium and low are defined in the method section.
The method section lists every source this map reads, including OpenStreetMap Overpass and the drought and streamflow joins.
Every limitation is published in the method section, including Overpass dependency, county-line geocoding risk, and that this page does not feed the main index.
Federal Register full-text search, SEC address geocoding as pins, and full OpenStreetMap geometry were tried and dropped for the reasons stated in the method section.
Shape is status. Distinguishable in greyscale and in a screenshot. Status is read from the OpenStreetMap tag that set it, and every pin carries that tag in the JSON.
Colour is the drought in that county. US Drought Monitor categories are printed in words beside the ramp so a reader who cannot separate two hues still gets the number.
Size is capacity, where it is published. A fixed dot means known site with capacity unpublished; a ringed and scaled pin means an IT-power tag exists.
No. Nothing here measures a datacentre's power or water draw, and the page says so above the map. What any of these buildings actually draws is not published by anyone.
A state with no pins is a state nobody has mapped. It is not a state with no datacentres.
No. This page is a SIREN sub-index. It draws, joins, and counts mapped sites — it does not feed the hourly activity index.
No. Activity tempo, not probability of harm. The US map counts OpenStreetMap datacentre sites joined to drought and grid data — not a probability of harm.
Every state in the contiguous frame, ranked by mapped sites. The states at the bottom with a zero are the point of this table.
Balancing authorities that publish a number without an account. These are the only figures on this page that can differ from the ones you saw last time. This page does not feed the main index.
Search-shaped sibling: How many datacentres has SIREN mapped on Earth?.
SIREN is free and carries no display ads. Keep it running: Support SIREN on Ko-fi →
