The HVAC Recall Registry: 95 Recalls and 18.1 Million Units Since 2015

Last reviewed: September 2026 · By the HVAC Brief Editorial Team · Our sourcing and editorial standards

Checking whether a specific unit has been recalled is a solved problem: the model number goes into the CPSC database and an answer comes back. What has never been available is the shape of the thing. How often does HVAC equipment get recalled, for what, and at what scale.

This page is both. The aggregate is below, then the full registry of 95 recalls with a link to the CPSC announcement for each.

The scale

Between 1 January 2015 and 23 September 2026, CPSC announced 100 recalls covering equipment in scope for this registry. 5 of those were re-announcements or expansions of an earlier recall, which describe units already counted, so the working total is 95 recalls covering 18,094,672 units.

The recall count per year is remarkably stable: between four and thirteen, with no visible trend across twelve years. The unit count is not stable at all, because a single large recall of a mass-market consumer product moves the annual figure by millions. That is the most important thing to understand about this data before quoting it.

What gets recalled

Two patterns sit in that table, and they run in opposite directions.

Dehumidifiers produce 38% of all recalled units from 3 recalls. Boilers produce 19 recalls, the most of any class, and 1.1% of the units. The difference is not safety performance, it is distribution: a dehumidifier is a retail consumer product sold by the million, and a boiler is a specified appliance installed one at a time.

Which means recall counts and recall volumes answer different questions. Counting recalls tells you which product types generate regulatory action. Counting units tells you how many households were exposed.

The hazards

Fire dominates, in 58 of 95 recalls and 82% of units. Carbon monoxide is the hazard most associated with heating equipment in public discussion and it appears in 21 recalls, 22% of the total, but only 1.3% of units. CO recalls concentrate in installed fuel-burning equipment, which ships in far smaller volumes than retail appliances.

CPSC describes at least one reported incident in 70 of the 95 recalls and states that no injuries were reported in 43. No aggregate incident or injury total is published here, and the methodology below explains why.

The registry

All 95 recalls, most recent first. Re-announcements and expansions are excluded. Each title links to the CPSC announcement, which carries the affected model numbers and the remedy.

Frequently asked questions

Methodology and limitations

• Source: US Consumer Product Safety Commission public recall database, retrieved through the agency’s REST service on 23 September 2026 for recall dates from 1 January 2015 to 23 September 2026. 3,690 recalls were retrieved across all product categories.
• Scope: recalls are included where the recalled product is residential or light commercial heating, cooling, ventilation, dehumidification, humidification or water heating equipment, or a control for such equipment. Classification is by the product name CPSC records, checked against the recall title. Products excluded by keyword include hair appliances, cooking appliances, heated blankets and pads, vehicle and engine equipment.
• Space and portable heaters are counted as a separate class so a reader who considers them a different category can subtract them.
• Units are the US figure. CPSC states Canadian and Mexican volumes in a parenthetical (“In addition, about 126 were sold in Canada”) and those are excluded. CPSC writes counts in three formats, including scaled forms such as “About 1.56 million”, all of which are parsed.
• Re-announcements: CPSC re-issues some recalls, sometimes with an expanded scope. 5 entries are flagged as re-announcements or expansions and excluded from every total, because their units were already counted. They remain visible in the underlying data.
• No aggregate incident or injury total is published. CPSC states incident and injury counts in free text, and sometimes as words rather than digits (“two reports”, “one report of smoke inhalation”), so a machine-readable total cannot be produced without under-counting. Presence of an incident description is reported instead.
• Hazard classification is derived from the hazard text CPSC publishes, by keyword. CPSC’s own structured hazard-type field is empty on most records in this set. A recall can carry more than one hazard, so hazard shares sum above 100%.
• A unit count is units distributed, not units that failed. It is the measure of how many households were exposed, and it is not a defect rate. No defect rate can be computed from this data because CPSC does not publish failures per unit.
• Recall counts cannot be normalised by units sold, because no public source gives shipments by the same product classes. For that reason no manufacturer is ranked here: a firm with more recalls may simply sell more equipment.
• Jurisdiction: CPSC covers consumer products. Vehicle equipment sits with NHTSA and is not here. Equipment recalled directly by a manufacturer without a CPSC announcement is not captured.
• Refresh: this registry is rebuilt monthly against the CPSC service.

Sources

1. US Consumer Product Safety Commission, Recalls database, and the agency’s public recall REST service at saferproducts.gov.
2. US Consumer Product Safety Commission, Furnaces and Air Conditioners product recall categories.

The HVAC Brief is an independent trade publication. We do not sell equipment, take manufacturer advertising, or accept payment for placement in our research. Corrections welcome.