The city recorded 2,669 net additional housing units, but understanding that figure requires separating new construction, demolitions, alterations, and future development.
Data snapshot: September 18, 2026. Source: the 2025 San Francisco Housing Inventory, published in April 2026.
San Francisco added a net 2,669 housing units in 2025, according to the historical series in the city’s 2025 Housing Inventory. That figure is larger than the number of units completed through new construction alone because the net change also accounts for demolitions and alterations to existing housing. Understanding those components is the first step toward reading a housing production headline accurately. View Homzora’s published extract.
The same report records 2,406 units completed through new construction, six units demolished, and a net gain of 269 units through alterations. Adding the new construction figure and the net alteration gain, then subtracting demolitions, produces the reported net addition of 2,669 units.
Those numbers describe completed changes to the housing stock. They do not mean that 2,669 apartments were simultaneously advertised for rent, that every added home was affordable to the same household, or that all proposed developments were completed. Each of those claims would require information beyond the citywide production table.
Begin with the accounting identity
The calculation is easiest to understand when each component remains separate. New construction adds units through newly built housing. Demolitions remove units. Alterations can add or remove units within the existing stock, so the published alteration figure represents the net effect of those changes.
| Component | Housing units |
|---|---|
| Completed through new construction | 2,406 |
| Demolished | 6 |
| Net units gained through alterations | 269 |
| Net change in housing units | 2,669 |
Source: San Francisco Planning Department, 2025 Housing Inventory, Table 2, printed page 17. The net total equals 2,406 plus 269 minus six.
Separating the components prevents a common error: reporting the new construction count as though it represented every change in the housing stock. It also makes the calculation auditable. A reader can reproduce the result rather than relying on a headline whose meaning is unclear.
In this case, the net alteration gain more than offsets the demolition count. The net addition is therefore greater than new construction completions alone. That relationship could look different in another year if removals or alteration losses were larger. The same accounting framework allows those differences to be examined consistently.
Production increased compared with 2024
The report’s historical series shows 1,603 net additional units in 2024 and 2,669 in 2025. The increase was 1,066 units, or approximately 66.5 percent. These calculations use the figures for both years as presented in the same 2025 report, which avoids mixing different report versions. Download the complete historical table.
New construction completions rose from 1,457 to 2,406 units, an increase of 949 units, or approximately 65.1 percent. Demolitions fell from 132 to six, while net additions through alterations declined from 278 to 269. The changes in those three components together explain the 1,066 unit increase in the net total.
A large percentage increase should always be read beside its starting value. The 2025 result represents a substantial increase over the 2024 figure in this series. It does not establish that the city reached an exceptional level across every possible historical comparison, nor does it establish that a similar percentage increase will occur again.
The recent history provides useful perspective
The five most recent years in the published table show why a single annual comparison is not enough. The 2025 net addition was above the 2024 and 2023 figures, but below the 2022 and 2021 figures. The sequence describes variation in completed production rather than an uninterrupted upward trend.
| Year | New construction completions | Net additions |
|---|---|---|
| 2021 | 4,165 | 4,621 |
| 2022 | 2,244 | 2,781 |
| 2023 | 2,059 | 2,544 |
| 2024 | 1,457 | 1,603 |
| 2025 | 2,406 | 2,669 |
Source: San Francisco Planning Department, 2025 Housing Inventory, Table 2. The same source vintage is used for every year.
Across the twenty years from 2006 through 2025, the table records 56,720 net additional units. Dividing that total by twenty produces an annual arithmetic average of 2,836 units. The 2025 result is below that descriptive average, even though it increased considerably from the preceding year.
That average is context, not a housing target or an estimate of what the city needs. It gives every year equal weight and says nothing about changes in population, household formation, policy goals, or the composition of the housing stock. Those questions require additional measures and should not be inferred from the average alone.
Completed housing is different from authorized housing
The same historical table includes net units authorized for construction. For 2025, that figure is 1,900. It should not be added to the 2,669 net completed additions to create a larger production total. Authorization and completion describe different stages, and the units counted at each stage need not belong to the same group of projects.
A project can receive authorization during one year and reach completion in a later year. That means the number authorized in 2025 is not a denominator for calculating the share of 2025 projects completed. The annual totals describe events occurring during the year, not a linked project cohort moving through identical stages together.
This distinction is particularly important when a report contains several large numbers. Adding authorizations, proposals, and completions can count different stages of the same development process as though each represented an independent addition to the city. A useful article identifies the stage before interpreting the total.
Temporary occupancy can be relevant to the completion measure
The Housing Inventory’s definitions include applicable units that have received temporary certificates of occupancy when counting completed housing. Readers should therefore use the report’s stated definition rather than assume that every completed unit must have reached exactly the same administrative milestone. See the report’s introductory definitions.
The practical implication is about consistency. If one dataset includes certain units at a temporary occupancy stage and another uses a different milestone, their totals may differ without either calculation being arithmetically wrong. Before comparing them, a reader needs to determine whether they are counting the same event.
The Homzora extract preserves the report’s figures rather than creating a new completion definition. That makes the source easier to trace and allows readers who need more detail to return to the original publication. The extract is a readable entry point, not a replacement for every technical note in the report.
Net additions are not a count of available rentals
The housing production table covers housing units rather than only apartments currently available for rent. A completed home can have a different tenure, occupancy status, price, or eligibility arrangement from the property a particular renter is seeking. The net addition figure does not describe those characteristics for each unit.
It is therefore inaccurate to turn the citywide total into a statement that a renter suddenly has 2,669 additional suitable choices. Some units may not be offered in the rental market at all. Others may not match the household’s needs or search timing. The dataset cannot resolve those differences.
For an active housing search, production statistics provide background about changes in the stock. Current availability must be established through actual listings and the relevant housing providers. Homzora’s neighborhood shortlist guide can help readers organize those property comparisons without confusing a broad supply measure with a list of homes they can secure.
A citywide total cannot describe every neighborhood
The table used in this article reports citywide figures. It does not show how the 2025 net addition was distributed among neighborhoods, how close the units were to particular destinations, or which areas experienced the largest change relative to their existing housing stock.
Those questions are worth asking, but the answers should come from appropriately defined geographic data. A neighborhood comparison also needs consistent boundaries. An informal area name used in a property advertisement may not correspond exactly to the geographic unit used by a public agency.
It would also be a mistake to assume that every neighborhood experienced the citywide rate of change. A total summarizes the combined result across its coverage area. It can conceal very different local experiences. A future neighborhood analysis should make those differences visible rather than distribute the citywide result evenly by assumption.
More production does not establish an immediate rent effect
The production table does not estimate how the added units affected rents. It records changes in housing stock. Explaining a rent movement would require attention to other conditions and to the timing, location, type, and occupancy of the units involved.
For example, placing an annual completion figure beside a monthly rent index does not show that one caused the other. Even if both moved during a similar period, the comparison alone would not isolate the effect of construction. It also would not establish what rents would have been without those additions.
Readers can explore Homzora’s rent series and production series together while preserving that distinction. The San Francisco housing data library presents them as separate datasets because they answer separate questions. One describes a market rent measure over time. The other describes completed changes in the number of housing units.
Use one historical version when making comparisons
SF Planning advises that its housing data has undergone corrections and that the latest report contains the most accurate available historical information. Figures for the same year may therefore differ across reports. Read the Housing Inventory source page.
This article uses the 2025 report’s historical series throughout. It does not take one year from an older publication and another from a newer one. That approach makes the comparisons internally consistent and reduces the risk that a reported change partly reflects a revision rather than the relationship the article intends to examine.
A later Homzora update should identify the new source version and review whether earlier figures changed. Keeping the snapshot date visible gives readers a way to reconcile this article with a future download. It also makes corrections more transparent when a publisher improves its historical records.
Questions to carry into the next release
The next useful questions extend beyond whether the headline total rose or fell. How much of the change came from new construction rather than alterations? Did demolitions affect the net result? Are the figures based on the same definitions and report vintage? Does the available evidence support a citywide conclusion or only a narrower one?
For renters, another question remains separate: which homes are actually available, suitable, and financially workable now? Production statistics can inform that search, but they cannot answer it alone. A household still needs current property information and a complete comparison of recurring expenses and upfront cash requirements.
San Francisco’s 2,669 net additions in 2025 are a concrete result that can be traced to a published source and reproduced from its components. Keeping those components, dates, and definitions visible makes the number more useful. Readers can download the full series and consult Homzora’s data methodology before making their own comparisons.