Original research · Electrification
The homes with no central air, and what adding it costs the panel
One in three US homes has no central air conditioning. Retrofitting one of them is not just an equipment purchase — the National Electrical Code has a specific rule for adding cooling to a panel that was never sized to carry it, and that rule gives the new load no discount at all.
Written by HyreElectrical Research Desk Primary-source research and fact checking
The finding
Read this before the numbers: what this page counts, and does not
The homes-without-central-AC study counts installed cooling equipment, not comfort, cost, or intent to retrofit. RECS 2020 records what a home uses to cool itself today. It does not say why a given home lacks central air, whether its occupants want it, or what installing it would cost that specific property — no national survey asks those questions. The 40.84 million figure is a population count: homes without a central system today. It is not a forecast of how many will add one, and not a per-home cost estimate.
Division is the finest geography this page can honestly publish, and it does not isolate the Pacific Northwest. RECS 2020’s public tables report the nine Census divisions. "Pacific" combines California, Oregon, Washington, Alaska and Hawaii into one figure — there is no published split that separates Oregon and Washington from California. Where this page reports a Pacific-division finding, that is the scope it describes; it is not a Pacific-Northwest-specific number, and this page does not claim it is.
The NEC 220.83 mechanism below is a paraphrase, not a quoted reproduction of the code. NFPA’s text is not freely reproducible online. The numbers here are corroborated across two independent NEC-education sources and are presented as HyreElectrical’s description of the mechanism, not as code language. Confirm the adopted edition and the exact figures with a licensed electrician before relying on them for anything more than understanding the shape of the rule.
Nothing here rests on HyreElectrical’s own contractor store. That store is five states and 74% Florida; it has no bearing on a national federal survey or a national-lab panel study, and neither is used anywhere on this page.
How the US cools its homes, 2020
Source fact. Of 123.53 million US housing units, the EIA’s 2020 Residential Energy Consumption Survey found 82.69 million (66.9%) use central air conditioning, including central heat pumps, as their main cooling equipment. HyreElectrical calculation: that leaves 40.84 million homes (33.1%) — one in three — without a central system: 21.43 million on window or wall units, 2.68 million on portable units, 1.76 million on ductless mini-splits, 0.95 million on evaporative coolers, and 14.02 million running no cooling equipment at all.
| Main cooling equipment | Homes (millions) | Share of US homes |
|---|---|---|
| Central air conditioning (incl. central heat pump) | 82.69 | 66.9% |
| Window or wall air conditioner | 21.43 | 17.3% |
| Portable air conditioner | 2.68 | 2.2% |
| Ductless heat pump (mini-split) | 1.76 | 1.4% |
| Evaporative or swamp cooler | 0.95 | 0.8% |
| No cooling equipment of any kind | 14.02 | 11.3% |
| No central air (all non-central rows summed) | 40.84 | 33.1% |
Rows may not sum to exactly 123.53 million because of independent rounding to the hundredth of a million in EIA’s own table. "No central air" is HyreElectrical’s own subtraction (total minus the central-AC row); every other figure on this row is read directly from the source table.
Why a central retrofit is an electrical question a room unit never was
A window, wall or portable air conditioner plugs into a receptacle the home already has — almost always a standard 120-volt outlet, occasionally a 240-volt one for the largest room units. Adding one, or even several, rarely means opening the panel: it is a load on a circuit that already exists, sized for something else, with headroom the appliance either has or does not.
A central air conditioning system is a different electrical animal. The outdoor condensing unit’s compressor and fan motor are, in nearly every residential split system sold in the US, fed by a dedicated 240-volt branch circuit sized to the equipment’s own nameplate — its minimum circuit ampacity, set by the manufacturer and required under NEC Article 440, the article governing air-conditioning and refrigerating equipment. HyreElectrical analysis. Installing central air where none existed is therefore, categorically, new work at the panel: a new breaker, new conductors, and — this is the part a marketing brochure skips — a fresh look at whether the service has room for what that breaker is about to carry.
The calculation: what NEC 220.83 says a retrofit adds, and to what it gets added
HyreElectrical’s panel-upgrade page already covers the optional calculation for a new dwelling (220.82) and the measured-demand method (220.87). Neither is the section written for adding central air to a home that has never had it. That is 220.83, "Existing Dwelling Unit" — and its air-conditioning subsection is the mechanism this page is actually about.
The mechanism, paraphrased — not quoted code language
Source fact, paraphrased. NEC 220.83 gives an optional, shortcut way to check whether an existing home’s service is big enough for new load, without redoing a full 220.82 calculation from scratch. Where no new air conditioning or electric heat is being added, subsection (A) counts the first 8 kVA of existing-plus-new general load at 100% and the remainder at 40% — a lower threshold than 220.82’s 10 kVA, because this is the article for a house that already exists, not one being designed from a blank sheet.
Where central air conditioning is the load being added — subsection (B) — the mechanism changes. The existing-plus-new general load is calculated the same way as (A), and then the new air-conditioning system’s load is added on top, at 100% of its nameplate rating. No 40% credit applies to it. That is not a HyreElectrical interpretation of what a fair rule should be — it is how the two subsections are structured to differ, corroborated across two independent NEC-education sources this desk checked rather than one.
The code is about to change this, which is itself confirmation of the current rule. The 2026 NEC renumbers this section (220.83 becomes 120.83) and removes the split between "no AC added" and "AC added" entirely, moving to one unified table where new loads generally get a 50% factor. A rule only gets simplified away once it has been doing something different — the fact that the 2026 edition explicitly drops "differential treatment for additional HVAC loads" confirms that the 2020 and 2023 editions, which is what nearly every jurisdiction has adopted today, do treat an added air conditioner more strictly than other added load.
What that 100% actually costs, at three illustrative equipment sizes
HyreElectrical calculation. A central air conditioner’s nameplate load varies by equipment size and manufacturer — this page does not have, and does not invent, a single number for "a central AC." Instead, three round, illustrative nameplate loads are run against the three panel ratings this desk already documents on its panel-upgrade page: 100 A, 150 A and 200 A. The middle value, 6,000 VA (25 A at 240 V), is this site’s own already-published load-calculator default HVAC nameplate — carried over here rather than invented fresh, for consistency across the site.
At 100% of nameplate, a small system (3,600 VA) consumes 15.0% of a 100 A panel’s total rated capacity by itself; the mid-size illustration (6,000 VA) consumes 25.0%; a large system (8,400 VA) consumes 35.0%. On a 200 A panel the same three systems consume 7.5%, 12.5% and 17.5% respectively — see the table below for every combination. The share does not shrink under 220.83(B) the way every other category of load does. That is the entire point of this section: whatever the rest of the house is already asking of the panel, a retrofit central air conditioner adds its full amount on top, undiscounted.
What this figure is not
It is not a claim that any specific panel will fail its calculation — that depends on every other load already on it, which varies house to house and which no national dataset can honestly generalise. It is not a claim that a 100 A panel cannot take a central air conditioner — the panel-upgrade page’s own worked example shows a 100 A service with real headroom to spare. And it is not a substitute for a licensed electrician’s stamped calculation, run against the equipment’s actual nameplate and the jurisdiction’s adopted edition — this page runs the addition in isolation because that is the part a general calculation can honestly generalise; what a specific panel is already carrying is a nameplate-by-nameplate job for someone standing in front of it.
The addition, at three panel sizes and three equipment sizes
| Panel rating | Illustrative AC nameplate | Added load (VA) | Share of total panel capacity |
|---|---|---|---|
| 100 A | Small system | 3,600 | 15.0% |
| 100 A | Mid-size system (this site’s own load-calculator default) | 6,000 | 25.0% |
| 100 A | Large system | 8,400 | 35.0% |
| 150 A | Small system | 3,600 | 10.0% |
| 150 A | Mid-size system (this site’s own load-calculator default) | 6,000 | 16.7% |
| 150 A | Large system | 8,400 | 23.3% |
| 200 A | Small system | 3,600 | 7.5% |
| 200 A | Mid-size system (this site’s own load-calculator default) | 6,000 | 12.5% |
| 200 A | Large system | 8,400 | 17.5% |
HyreElectrical calculation. Panel capacity is amps × 240 V. AC nameplate values are round, illustrative figures for this worked example, not a specific product’s rating; a real installation is sized from its own data plate. The 6,000 VA mid-size value matches this site’s own load-calculator default HVAC nameplate. Under NEC 220.83(B) this share is added at 100% — no demand factor reduces it — on top of whatever the rest of the home’s calculated load already is.
Where the homes without central air are: all nine Census divisions
Source fact, HyreElectrical calculation. RECS 2020 reports cooling equipment for each of the nine Census divisions. New England and Middle Atlantic — the two Northeast divisions the contract for this page named — are genuinely among the lowest in central-AC saturation, and Pacific joins them. That corroborates the broad geography; it does not confirm the Pacific Northwest specifically, for the reason given above.
| Census division | Homes (millions) | No central air | Share no central | No cooling at all |
|---|---|---|---|---|
| New England | 5.88 | 4.24 | 72.1% | 15.3% |
| Middle Atlantic | 16.04 | 9.11 | 56.8% | 9.4% |
| Pacific | 18.51 | 9.81 | 53.0% | 31.7% |
| Mountain | 9.22 | 2.88 | 31.2% | 13.1% |
| East North Central | 18.55 | 5.16 | 27.8% | 7.3% |
| West North Central | 8.50 | 1.87 | 22.0% | 5.3% |
| West South Central | 14.62 | 2.76 | 18.9% | 5.3% |
| East South Central | 7.38 | 1.29 | 17.5% | 6.2% |
| South Atlantic | 24.84 | 3.71 | 14.9% | 6.0% |
EIA RECS 2020, Tables HC7.7 (Northeast, Midwest) and HC7.8 (South, West), retrieved 6 September 2026. "Mountain" is EIA’s own published combined column for Mountain North and Mountain South — unlike this site’s water-heating study, this table set publishes that combination directly.
Summing the nine divisions’ no-central-air counts gives 40.83 million against the national total of 40.84 million, and the no-cooling-at-all counts sum to 14.02 million against a national 14.02 million — both within EIA’s own independent rounding, not a suppressed-cell gap. Every figure in this table is a directly published, unsuppressed EIA cell or a subtraction of two such cells.
Two divisions lack central air for two different reasons
HyreElectrical analysis. New England is the least centrally cooled division in the country at 27.9% central air, with Middle Atlantic close behind at 43.2%. Neither division runs high on "no cooling at all," though: New England’s share without any cooling equipment is 15.3%, and Middle Atlantic’s is 9.4%. Most of their gap to the national average is made up of homes running window, wall or portable units instead of a central system — cooling exists in most of these homes; it typically is not central.
Pacific tells a different story. Its central-AC share, 47.0%, sits close to New England’s, but its share with no cooling equipment at all — 31.7% — is more than double any other division’s and nearly six times the national rate. 5.87 million Pacific-division homes, out of 14.02 million nationally, simply do not cool the house at all. California’s mild coastal climate — the division’s largest state by population — is the plausible driver; RECS does not publish a state-level breakdown that would let this page test that directly, so it is stated as a pattern, not a proven cause.
We have not attempted to join this division cut to a division-level year-built table — RECS does not publish one for air conditioning, and this desk will not manufacture a precision the source does not have by relabelling the era table below as though it described these same divisions. What the era table does support, on its own, is a plausible structural reason central air lags in older housing generally: it shows below that central-AC saturation climbs steadily as construction gets newer, which is consistent with — but does not prove — older regional housing stock carrying a disproportionate share of the homes still without it.
By year built: central air tracks the age of the house closely
Source fact, HyreElectrical calculation. Splitting the national count by year of construction shows a much cleaner pattern than the water-heating fuel study on this site found for its own question: the share of homes without central air falls in almost every successive construction band, from 57.0% in homes built before 1950 to 15.6% in homes built 2016–2020.
| Year built | Homes (millions) | No central air (millions) | Share no central air |
|---|---|---|---|
| Before 1950 | 20.26 | 11.55 | 57.0% |
| 1950–1959 | 12.48 | 5.01 | 40.1% |
| 1960–1969 | 12.76 | 4.75 | 37.2% |
| 1970–1979 | 18.34 | 6.57 | 35.8% |
| 1980–1989 | 16.30 | 4.84 | 29.7% |
| 1990–1999 | 17.16 | 3.87 | 22.6% |
| 2000–2009 | 16.16 | 2.71 | 16.8% |
| 2010–2015 | 5.53 | 0.85 | 15.4% |
| 2016–2020 | 4.56 | 0.71 | 15.6% |
EIA RECS 2020, Table HC7.3, retrieved 6 September 2026.
80.1% of all no-central-air homes (32.72 of 40.84 million) are in homes built before 1990. Even the newest construction band, 2016–2020, is 15.6% without central air — lower than every older band, but not zero. Summing the nine era bands gives 40.86 million against the national 40.84 million, a gap under 0.05%, within EIA’s own independent rounding.
Federal money treats a heat pump’s cooling and the panel as separate line items
Source fact. The Home Electrification and Appliance Rebates programme (HEAR), the federal rebate this site’s electrification-incentives guide tracks in detail, lists an electric heat pump for space heating and cooling as its own measure, capped at $8,000, separate from an electric load service center — the panel — capped at $4,000, and electric wiring capped at $2,500. HyreElectrical analysis. That statutory separation matches this page’s finding: the programme was written by people who expected the cooling equipment and the panel work behind it to be two different line items on a real number of jobs, not an edge case. Full eligibility rules, income tiers and the duplication rule against the HOMES programme are on our electrification incentives guide — this page does not repeat that detail. Who is already struggling to pay for the electricity is home energy insecurity, and the four-end-use count that cooling sits inside is how many US homes are all-electric.
Limitations
- The 40.84 million figure is an equipment count, not a retrofit forecast
The 40.84 million figure describes today’s cooling equipment. It says nothing about how many of these households intend to add central air, on what timeline, or at what rate.
- No state-level breakdown, and the Pacific division cannot isolate the Pacific Northwest
The finest published geography in this table set is the Census division. Oregon and Washington cannot be separated from California, Alaska and Hawaii inside RECS’s public "Pacific" figure. A true Pacific Northwest number would need RECS’s restricted or public-use microdata file, which this page has not processed.
- The division and era cuts are not cross-tabulated
RECS 2020’s public tables do not publish air-conditioning equipment by Census division AND year of construction at once. This page reports both cuts honestly and does not claim they describe the same households.
- The NEC 220.83 description is a paraphrase, not quoted code text
NFPA’s text is not freely reproducible online. This desk corroborated the mechanism — the 8 kVA threshold, the 40% remainder, and the 100%-of-nameplate treatment of added air conditioning — across two independent NEC-education sources rather than one, but a licensed electrician working from the actual adopted-edition text is the authority on the exact figures.
- The panel-size illustration uses national, not division-specific, panel-amperage data
The LBNL/NREL stock estimate (29% of US single-family homes at 100 A, 2% below) is a national figure carried over from this site’s panel-upgrade page. No published dataset gives panel amperage by Census division, so this page does not claim the low-central-AC divisions specifically have a higher share of 100 A panels — only that the national stock includes a meaningful share of them.
- This page does not cost a retrofit
Installed cost for a central air conditioning retrofit, including the electrical work, is a separate question this page does not answer. Panel-upgrade pricing specifically lives on our panel-upgrade cost page.
Method
Sources. US Energy Information Administration, 2020 Residential Energy Consumption Survey: Table HC7.1 (air conditioning by housing unit type, for the national total), Table HC7.3 (air conditioning by year of construction), and Tables HC7.7 and HC7.8 (air conditioning by Census division). All four opened directly from eia.gov/consumption/residential/data/2020/hc/pdf/ on 6 September 2026. NFPA 70 (National Electrical Code), Article 220, section 220.83, described as a paraphrase and corroborated against two independent NEC-education summaries the same day. LBNL and the National Renewable Energy Laboratory’ panel-stock estimate, carried over from this site’s panel-upgrade page (retrieved there 5 September 2026).
Derivation. "No central air" subtracts the published central-AC row from the published total, by nation, division and year-built band, using only cells EIA published without suppression. "No cooling at all" is EIA’s own "does not use air-conditioning equipment" row, unmodified. "Room/individual cooling" is the difference between the two. The panel-size illustration multiplies three round nameplate VA values by three panel amp ratings at 240 V; no demand factor beyond NEC 220.83(B)’s stated 100% is applied to the added AC load in that table.
What is not a measurement. The claim that low-central-AC divisions and older construction eras are more likely to carry smaller panels is HyreElectrical’s own inference from two separately published facts — the era table on this page, and the national panel-stock estimate on this site’s panel-upgrade page — not a single dataset that measures both at once.
Reproducibility. Every EIA figure on this page is read directly off a public table with no interpretation applied before the sums shown here. Anyone with the same four files can rebuild every row.
Questions
How many US homes don’t have central air conditioning?
What percentage of US homes have no air conditioning at all?
Which part of the US has the least central air conditioning?
Does this study prove the Pacific Northwest specifically has low central-AC saturation?
Why do older homes have central air conditioning less often?
What does NEC Article 220 say about adding central air to an existing panel?
Does adding central air always require a panel upgrade?
Is there federal money for adding central air along with electrical work?
Written and audited by
HyreElectrical Research Desk
Primary-source research and fact checking
We read the model code, the federal safety notice, the municipal fee sheet or the utility tariff ourselves, and we publish the figure with the document it came from and the date we retrieved it. Where a number cannot be traced to a primary source, we publish the shorter page and say what we could not verify. HyreElectrical does not perform, supervise or warrant electrical work. Authorship is organisational: this desk, not a named persona.
- 10
- long-form launch pages in this layout
- 5
- states with verified licence records
- 16,369
- electrical companies in those records
- 74%
- of the store is Florida — disclosed first, not footnoted
How this desk works
- Primary sources only. Code statements come from NFPA 70 as the model code. Safety statements come from CPSC. Permit fees come from the city or county fee sheet. Utility charges come from the filed tariff. We do not cite a blog that cites a source; we open the source.
- Our contractor store is five states, and 74% of it is Florida. Any figure built on that store is titled to those states and names the concentration in the first screen. It is not a national sample of electricians.
- Load calculations on this site are a published simplification in the shape of NEC 220.82 — first 10 kVA at 100%, remainder at 40%, HVAC at 100%, EVSE at 125%. They are labelled as not a stamped calculation. A licensed electrician using the adopted edition does that work.
- No national price for a panel upgrade is shipped. Labour rates, the adopted code edition, whether the utility is involved, and the state of the existing wiring move the invoice too far for a roundup to help. Cost intent lives on the cost page; this page explains the decision.
- We do not perform electrical work, and we take no payment for placement, ranking or a favourable mention. Nobody buys a position on this site.
Data as of 6 September 2026. Authorship on this site is organisational: the analysis belongs to the desk rather than to a named individual, and we do not publish credentials we do not hold. Our editorial policy sets out how we source, date and correct what we publish.
Sources & retrieval dates
- US Energy Information Administration — 2020 Residential Energy Consumption Survey, Table HC7.1 (Air conditioning in U.S. homes, by housing unit type) , The national cooling-equipment split: 82.69 million homes on central air conditioning, 21.43 million on window or wall units, 2.68 million on portable units, 1.76 million on ductless mini-splits, 0.95 million on evaporative coolers, and 14.02 million with no cooling equipment, out of 123.53 million total housing units. Retrieved 6 September 2026.
- US Energy Information Administration — 2020 RECS, Table HC7.3 (Air conditioning in U.S. homes, by year of construction) , The nine construction-era bands and their central-AC and no-AC splits used in the era table and chart. Retrieved 6 September 2026.
- US Energy Information Administration — 2020 RECS, Table HC7.7 (Air conditioning in homes in the Northeast and Midwest regions) , New England, Middle Atlantic, East North Central and West North Central division figures. Retrieved 6 September 2026.
- US Energy Information Administration — 2020 RECS, Table HC7.8 (Air conditioning in homes in the South and West regions) , South Atlantic, East South Central, West South Central, Mountain (published combined) and Pacific division figures. Retrieved 6 September 2026.
- NFPA 70, National Electrical Code (NEC) — standard development , Cited as the model code for Article 220, section 220.83 ("Existing Dwelling Unit"), the optional calculation for adding load to an existing dwelling and its air-conditioning subsection. Described here as a paraphrase, not a reproduction of the article. Confirm the edition your jurisdiction adopted. Retrieved 6 September 2026.
- up.codes — summary of NEC 220.83, "Existing Dwelling Unit" , Independent corroboration of the 220.83(A)/(B) structure used to describe the mechanism on this page. A summary, not verbatim code text. Retrieved 6 September 2026.
- ElectricalLicenseRenewal.com — NEC continuing-education page on the 2026 renumbering of 220.83 to 120.83 , Corroborates the 8 kVA / 40% mechanism and describes the 2026 NEC’s removal of differential treatment for added HVAC load, used here as confirmation that the current (2020/2023) edition treats it differently. Retrieved 6 September 2026.
- Murphy, Liu, Less et al. — Characterizing electrical panel capacity, breaker space, and loads in U.S. single-family homes , Journal of Building Engineering vol. 120 (2026), LBNL and the National Laboratory of the Rockies. National estimate that 31% of US single-family homes have panels rated 100 A or smaller (29% at 100 A, 2% below). Carried over from this site’s panel-upgrade page, retrieved there 5 September 2026; not independently re-verified for this page beyond confirming the citation. Retrieved 2026-09-05.
- HyreElectrical — Home electrification incentives (electrification-incentives guide) , HEAR programme caps: $8,000 for an electric heat pump for space heating and cooling, $4,000 for an electric load service center, $2,500 for electric wiring. Reused here, not re-derived. Retrieved 6 September 2026.
Adding central air where none existed?
Run the load before you order the equipment. Tell us where the house is and what the panel currently carries, and we will connect you with licensed electrical companies in your area.
HyreElectrical does not perform, supervise or warrant electrical work, and takes no payment for placement, ranking or favourable mention. This is research on published federal survey data and a model-code mechanism, not advice on a specific property. Every figure on this page is a population count or an illustrative calculation, not a forecast of what any individual home will need or cost. We have no commercial relationship with EIA, NFPA, LBNL, or any manufacturer named above.