Answer in brief
CMHC’s September release puts Canada’s six-month construction trend at 244,149 annualized units. A flat standalone August rate masks a softer pipeline and distinct local markets.
The latest release shows a softer construction trend
Canada housing starts lost momentum in the latest August dataset, released by CMHC on September 16 and available at the October 1 publication cutoff. The six-month trend fell 1.3% to an annualized 244,149 units. The standalone August rate was essentially flat at 229,046. For readers assessing future housing choice, that difference matters: one stable month can coexist with a weakening average across the recent construction pipeline.
The release concerns work beginning on new homes. It does not establish how many keys were handed over or how many existing homes changed owners. Those stages answer different questions about immediate availability, while starts offer an earlier signal of supply that still has to pass through construction.
Three measures need their own units and boundaries
CMHC publishes actual counts, a seasonally adjusted annualized rate and a six-month average of that rate. Annualization removes seasonal variation and multiplies the adjusted monthly level by twelve. The result expresses a pace that can be compared across months; it is not a count of homes built during a calendar year.
The table preserves the reporting geography and comparison period for each row. Statistics Canada’s September 17 table also publishes the six-month series, in thousands of units, with CMHC identified as the source. Its July and August entries correspond to 247,446 and 244,149 units after converting the published scale. It is a second official publication of the same survey series, rather than an independent construction sample.
| Measure | Latest reading | Comparison reading | Coverage and units |
|---|---|---|---|
| Standalone monthly rate | August 2026: 229,046 | July 2026: 229,360 | All Canada; annualized, seasonally adjusted units |
| Six-month trend | August 2026: 244,149 | July 2026: 247,446 | All Canada; mean of six monthly annualized rates |
| Actual monthly starts | August 2026: 17,691 | August 2025: 18,112 | Centres with at least 10,000 residents; actual units |
A construction beginning leaves delivery questions open
CMHC defines a start at the point when concrete has been poured for the whole footing around the structure, or an equivalent stage for a building without a basement. Counting that milestone separates physical construction from a planning announcement. It also leaves a substantial distance between a measured start and a dwelling available to a household.
A prospective buyer evaluating a particular development still needs the building’s schedule and completion evidence. The national series does not show whether an individual contractor is delayed, when a specific unit will be occupied, or how its final price will change. Those project-level facts remain unknown for an unidentified development.
Local trends diverge from the national average
The August release reports a 6% monthly increase in the six-month trend for Montréal, driven by higher starts across housing types. Vancouver also rises 6%, with multi-unit construction driving its change. Toronto’s trend is flat. These local readings use the same trend concept, but the building mix and city boundaries remain essential to the interpretation.
A national decline therefore does not imply fewer starts in every city. Apartment construction can also respond to rental demand while the supply of homes intended for ownership develops differently. A household comparing locations should pair the city’s construction mix with the type of home it needs, rather than treat every new dwelling as an interchangeable purchasing opportunity.
The supply requirement is an affordability scenario
A separate CMHC report released September 10 estimates that annual starts would need to reach 417,000–469,000 to restore pre-pandemic affordability by 2036. That range comes from a model and a stated policy horizon. It describes a construction requirement under the report’s framework, rather than a measured future outcome or a promise that those homes will be delivered.
The report says rental housing dominates new supply in major markets outside Calgary and Edmonton, while construction for ownership has weakened. This helps explain why a softer buying market does not automatically resolve future supply needs. Comparing the model’s annual range with one annualized monthly rate can show scale, but cannot produce a verified count of missing homes or a future local price.
The next release can update the pace, not delivery certainty
CMHC’s calendar schedules the next monthly construction release for October 16. At this article’s cutoff it remains future. Readers will be able to check whether the trend continues to soften and whether local movements persist. Published data should be compared on the same release basis, because the construction tables include revised series.
The current news is a measurable weakening in the six-month national trend, alongside an almost unchanged latest monthly pace. For a real housing decision, keep actual counts, annualized rates and modelled requirements separate. Then add the location, tenure and delivery evidence for the home being considered. That sequence turns a large construction number into a usable assessment of future choice.
Questions and answers
Were 229,046 homes actually started in August?
No. That is the seasonally adjusted annualized rate for all Canadian areas. The release separately counts 17,691 actual August starts in centres with at least 10,000 residents.
Why can the monthly rate be flat while the trend falls?
The trend averages six monthly seasonally adjusted annualized rates. Older months leave that rolling window as a new month enters, so its movement can differ from the latest standalone rate.
Does a housing start mean a home is ready to occupy?
CMHC counts the beginning of construction, generally when the full footing has been poured or an equivalent stage is reached. Completion and occupancy require separate evidence.