Where Could a Data Center Still Be Built?
On 25 August 2026 the City Council declined a moratorium on data centres and sent staff to rewrite the zoning code instead. The Zoning Commission finalises in September and Council votes in October. This page answers the question that decides the vote and that nobody had published: under the proposed rules, how much land is left, and where is it?
Not adopted. DSD staff proposal, still in stakeholder feedback. NOT adopted. Zoning Commission September 2026; Council vote expected mid-October 2026. Everything below is “if this draft passes unchanged”, scored against the proposal as of 2026-08-26.
What each rule does on its own
The proposal is four rules at once. Applied in order, this is where the land goes. The three setback figures overlap and must not be added together. A parcel near both a school and a house is inside both.
| Step | Acres left | Removed |
|---|---|---|
| Allowed today | 81,746 | – |
| Commercial + industrial only (drops office and downtown) | 80,063 | 1,684 |
| Minus the Edwards Aquifer recharge zone and the CBD | 73,712 | 6,351 |
| Minus everything within 500 ft of a home, school, hospital, nursing home or day care | 47,276 | 26,436 |
The setback is the number being negotiated
Development Services said at the stakeholder meeting that the 500-foot setback is still taking feedback: supporters asked for less, critics for more. So it is scored at four distances rather than one. Read the range, not the headline.
| Setback | Eligible acres | % of today | |
|---|---|---|---|
| 300 ft | 56,568 | 69.2% | |
| 500 ft proposed | 47,276 | 57.8% | |
| 750 ft | 37,816 | 46.3% | |
| 1,000 ft | 32,352 | 39.6% |
Which zoning list is adopted matters more than the setback
The public record contains two drafts, and they disagree about the thing that decides the most land. The proposal reported in August confines data centres to “commercial and industrial districts”, the broad list scored above. The staff briefing given to Council on 4 March confined them to C-3 and I-1 only, and used a 1,000 ft buffer that also covered parks. Under the narrow list the starting land is less than half, so the zoning list becomes the dominant rule and the setback the secondary one. That is the reverse of the broad reading. Neither draft is adopted. Do not quote one without the other.
| Permitted districts | Starting land | Removed by the zoning list | Eligible at 500 ft |
|---|---|---|---|
| Commercial + industrial as reported August 2026 | 73,712 | 1,684 | 47,276 |
| C-3 and I-1 only City and utility staff briefing, 2026-03-04 | 32,721 | 49,025 | 21,325 |
The March briefing also counted parks as an incompatible use, which the August reporting does not. Adding parks to the broad reading removes a further 6,569 acres, leaving 44,017. It is scored on its own line rather than folded into the headline, because the current draft does not name it.
By council district
The citywide figure is not what any one council office is voting on. This is the same calculation, district by district. District 0 does not appear because it is not a council district, and land outside the city limits is outside the City’s zoning reach entirely.
| District | Allowed today | Eligible at 500 ft | % of today | Removed by homes | |
|---|---|---|---|---|---|
| District 1 | 3,904 | 1,007 | 25.8% | 2,175 | |
| District 2 | 14,014 | 10,043 | 71.7% | 3,707 | |
| District 3 | 14,973 | 11,382 | 76% | 3,432 | |
| District 4 | 15,159 | 11,879 | 78.4% | 3,195 | |
| District 5 | 3,034 | 1,002 | 33% | 1,853 | |
| District 6 | 8,243 | 5,259 | 63.8% | 2,846 | |
| District 7 | 3,119 | 907 | 29.1% | 1,990 | |
| District 8 | 6,561 | 2,260 | 34.4% | 2,407 | |
| District 9 | 5,619 | 596 | 10.6% | 1,109 | |
| District 10 | 7,122 | 2,941 | 41.3% | 3,012 |
allowed today still eligible under the proposal
Where that land is
Every shaded block is land that survives all four rules at 500 ft. Blocks are 1,200 ft square and shaded by how much of the block is eligible, because a block is rarely eligible end to end. Eligible is not available: nothing here knows who owns a parcel, whether it is for sale, whether it is already built on, or whether CPS Energy and SAWS can serve it.
1,881 shaded runs drawn, in three density shades. Measured on a 100-foot raster over the 324,428-acre city limits, so a boundary is resolved to within 100 feet. This is not a survey and no decision about an individual parcel should be taken from it.
What was scored, and what had to be assumed
Permitted districts under the proposal: C-1, C-2, C-2NA, C-2P, C-2PNA, C-3, C-3NA, C-3R, NC, I-1, I-2, L, MI-1, MI-2.
Counted as allowed today but not under the proposal: O-1, O-1.5, O-2, D , because a data centre is currently permitted as “office data processing and management” in most nonresidential districts.
Protected residential zoning: R-1, R-2, R-3, R-4, R-5, R-6, R-20, RE, MH, RM-4, RM-5, RM-6, MF-18, MF-25, MF-33, MF-40, MF-50, MF-65.
- Measurement: Measured from the protected parcel or zone boundary, not centre to centre and not structure to structure. The proposal does not yet say which it means, and this is the stricter reading.
- Residential: Residential ZONING, not dwelling units. Vacant residentially-zoned land therefore counts as protected.
- Schools: Proxied by the City school-zone polygons, which surround campuses. Generous toward exclusion: the zone is larger than the school.
- Baseline: Today's rule is taken as commercial + industrial + office + downtown, because a data centre is currently permitted as "office data processing and management" in most nonresidential districts.
What this cannot tell you
- ⛔ THESE RULES ARE NOT LAW. They are the Development Services staff proposal as reported on 2026-08-26, still in stakeholder feedback. The Zoning Commission finalises in September and Council votes in October. Every figure on this page is "if this draft passes unchanged", and the draft will change.
- ⛔⛔ THE PUBLIC RECORD CONTAINS TWO DIFFERENT DRAFTS, and which one is adopted matters more than the setback distance. The August 2026 proposal is reported as "commercial and industrial districts", the broad list scored here. But the March 2026 staff briefing to Council limited data centres to C-3 and I-1 ONLY, and used a 1,000 foot buffer that also covered parks. Under the narrow list the starting land is less than half, so the zoning list becomes the dominant rule and the setback the secondary one, which is the reverse of the broad reading. Both are published above. Do not quote one without the other.
- The 500-foot setback is the most contested number in the proposal, with supporters asking for less and critics for more, which is why every figure here is also published at 300, 750 and 1000 feet. The March briefing used 1,000 feet. Read the sweep, not just the headline.
- Parks were named as an incompatible use in the March briefing and are NOT in the August reporting, so they are scored on their own line and never folded into the headline.
- ⛔ The proposal does not yet say whether 500 feet is measured from the property line or from the structure. This page measures from the protected parcel or zone boundary, the stricter reading. A structure-based rule would leave MORE land eligible than shown here.
- "Residential" here means residential ZONING, not dwelling units, because zoning is the complete dataset. Vacant residentially-zoned land counts as protected, which again is the stricter reading.
- Schools are proxied by the City school-zone polygons, which surround a campus and are larger than it. That over-excludes slightly, and it is the only school layer the City publishes in this org.
- ⛔⛔ THE CITY ZONING LAYER OVERLAPS ITSELF, so the obvious way to do this calculation is wrong. Adding up the layer’s own Shape__Area for the commercial and industrial districts returns about 114,989 acres, but those polygons cover the same ground more than once, so counting each piece of land once gives about 81,000 instead. Two independent checks: asking the City’s own service which parcels genuinely OVERLAP a sampled parcel, which excludes parcels that merely share a boundary, returned overlaps on 11 of 12 sampled; and the overlap factor measured here is 1.42 whether the geometry is fetched simplified or raw, so it is in the data rather than in our method. Anyone who sums the area column, which is the first thing a spreadsheet does, overstates the land available by roughly 40%.
- ⛔ Everything is measured on a 100-foot raster, so a boundary is resolved to within 100 feet. That is adequate against a setback being negotiated between 300 and 1000 feet. IT IS NOT A SURVEY, and no decision about an individual parcel should be taken from it.
- Eligible land is not available land. Nothing here knows who owns a parcel, whether it is for sale, whether it is already built on, or whether CPS Energy and SAWS can serve it. Zoning permission is a floor under the question, not an answer to it.
- The "removed by" figures for residential, schools and medical OVERLAP and must never be summed. A parcel near both a school and a house is counted in both.
- Office districts (O-1, O-1.5, O-2) are NOT counted as eligible under the proposal, but they ARE counted in the "allowed today" baseline, because a data centre is currently permitted as "office data processing and management" in most nonresidential districts. That single choice moves the headline percentage, so it is stated rather than buried.
- Only land inside the city limits is scored. The ETJ and unincorporated Bexar County are outside the City's zoning reach, and a data centre built there is unaffected by any of this.
Take the data. Download the CSV · Use the API · rebuilt weekly, generated 2026-08-29.
Which rule does the work depends on which draft passes
Two drafts are on the public record, and they disagree about the thing that decides the most land.
Under the August draft, the setback is almost the whole story. That version confines data centres to “commercial and industrial districts”. Measured against the City’s own zoning layer, that list removes 1,684 acres, close to a rounding error, because a data centre is already permitted today as “office data processing and management” in most nonresidential districts. Keeping them out of the Edwards Aquifer recharge zone and the central business district removes 6,351 more. Then the 500-foot setback removes 25,725 acres on the residential rule alone, 15.3 times what the zoning list does.
Under the March draft, that reverses. The staff briefing given to Council on 4 March confined data centres to C-3 and I-1 only. That list starts from 32,721 acres rather than 73,712, so the zoning restriction alone takes about 49,025 acres off the table before any setback applies, and the setback becomes the secondary rule. The March version also used a 1,000-foot buffer, and counted parks as an incompatible use, which the August reporting does not.
So the honest answer to “which rule matters most” is that it has not been decided yet, and the choice of zoning list moves more land than the choice of setback distance does. Both are scored above. Schools account for a further 2,450 acres and hospitals, nursing homes and day care for 1,080; those overlap and cannot be added together.
Within a given list, the setback is still the number under active negotiation. Moving it from 500 to 750 feet costs another 9,460 acres; moving it down to 300 gains 9,293. Development Services said at its stakeholder meeting that it was still taking feedback on that number, with supporters asking for less and critics asking for more.
The districts are not affected anything like equally
A citywide figure is not what any one council office is voting on.
Citywide, 57.8% of the land where a data centre is allowed today would still be eligible. Almost no district is close to that average, and the spread is the widest thing on this page.
At one end, District 9 keeps 10.6% of its currently eligible land, District 1 keeps 25.8% and District 7 keeps 29.1%. These are districts whose commercial and industrial land sits threaded through neighbourhoods, so a 500-foot buffer off residential zoning consumes nearly all of it.
At the other end, District 4 keeps 78.4%, District 3 keeps 76.0% and District 2 keeps 71.7%. Their industrial land sits in larger blocks with fewer houses against it, so the same rule barely touches it.
District 6, which has eleven data centres already and whose councilman asked for the moratorium, sits in between at 63.8%, and still holds 5,259 eligible acres under the proposal: the third largest figure in the city. Whether that is the right outcome is a political question this page takes no position on. It is simply not what a reader would guess from the debate, and it is checkable in one query.
The obvious way to compute this is wrong by about 40%
Worth knowing before anyone checks the arithmetic, because the check itself has a trap in it.
The natural way to answer “how much commercial and industrial land is there” is to filter the City zoning layer and add up its own area column. Do that and you get about 114,989 acres. It is the first thing a spreadsheet does, and it is wrong.
The zoning layer overlaps itself. Its polygons cover the same ground more than once, so the area column double counts. Two independent checks say so. Ask the City’s own service which parcels genuinely overlap a given parcel, a relation that deliberately excludes parcels merely sharing a boundary, and 11 of 12 sampled commercial and industrial parcels come back with overlaps. And when this page paints those polygons onto a grid, the average covered spot is claimed about 1.42 times, whether the geometry is fetched simplified or raw, so the overlap is in the data rather than in the method used to measure it.
Counting each piece of land once gives about 80,000 acres, not 114,989. Every figure on this page counts once. It is also why the numbers here will not match a quick check done the quick way, and why the overlap factor is published in the API rather than left as something to discover.
What this is not
The limits are as load-bearing as the findings.
These rules are not law. Everything here scores a Development Services staff proposal as reported on 2026-08-26, while it was still in stakeholder feedback. The draft will change. That is exactly why the criteria are published in full above and served by the API, and why the whole thing can be re-run against a new draft in minutes.
Two things the published proposal does not say had to be assumed, and both are stated above rather than buried: whether the 500 feet is measured from the property line or from the building, and whether “residential” means residential zoning or actual dwellings. This page takes the stricter reading of both. A structure-based rule would leave more land eligible than shown here.
Eligible is not available. Nothing on this page knows who owns a parcel, whether it is for sale, whether something is already built on it, or whether CPS Energy and SAWS could serve it. Zoning permission is a floor under the question, not an answer to it. Those utility questions are the ones residents actually ask about data centres, and no zoning layer contains them.
It is measured on a 100-foot grid across the 324,428 acres inside the city limits, so a boundary is resolved to within 100 feet. That is adequate against a setback being negotiated between 300 and 1,000 feet. It is not a survey, and no decision about an individual parcel should be taken from it. There is deliberately no address lookup: whether a specific site may host a data centre is a zoning determination, which the City issues and we do not.
Check it yourself
Every layer used here is published by the City.
The zoning districts, the Edwards Aquifer zones, the school zones, the long-term care facility locations, the hospital locations and the council district boundaries all come from the City of San Antonio’s own ArcGIS services. Nothing was bought, scraped or estimated. The full criteria, the per-district table and the setback sweep are available as CSV and through the public API, which needs no key.
If the Zoning Commission changes the setback in September, the page will be re-run against the new draft.