America Built 1.4 Million Apartments in Three Years. The Housing Shortage Still Got Worse.

4.03 million. That is the number of homes America is short, as of the end of 2025, according to Realtor.com's 2026 Housing Supply Gap Report. One year earlier, it was 3.80 million. The gap grew by 230,000 homes in a single year.

This happened during the biggest apartment construction boom in four decades. Between 2023 and 2025, developers completed 1.42 million apartments, a pace not seen since the mid-1980s. Multifamily completions peaked at 584,000 units in 2024, running 92% above the pre-pandemic average of 303,000 per year.

And the shortage still got worse.

Now the boom is over. Apartment starts have fallen 55% from their mid-2022 peak of 580,000 to roughly 260,000 in 2025. Units under construction dropped from 1.1 million to 534,000. RealPage projects multifamily completions will fall below 300,000 in 2026. Continental Properties estimates 316,000. Either way, by 2027, apartment deliveries will hit their lowest level in a decade. That is the supply cliff.

The Impossible Math

Here is why the boom failed to close the gap. I built a simple deficit model using public data. The inputs are transparent so you can argue with them:

Housing starts, 2025: 1.359 million (Census Bureau)
Household formation, 2025: 1.41 million (Realtor.com, up from 999,000 in 2024)
Demolitions + natural disasters: ~300,000-400,000 units/year (estimated from Census AHS data)
Housing completions, 2025: 1.498 million (Census)

Run the numbers. Completions (1.498M) minus household formation (1.41M) equals a net surplus of just 88,000 units. But subtract demolitions (conservatively 300,000) and the net is negative 212,000. The gap widens every year, even during a boom.

Now look at the pipeline going forward. Current starts (1.359M) are already below household formation (1.41M) by 51,000, before demolitions even enter the equation. Starting in 2026, completions will plunge as the collapse in starts works through the 18-24 month construction pipeline. RealPage estimates multifamily completions follow this trajectory:

Year	Multifamily Completions	Change vs. Peak (2024)
2024	584,000	Baseline
2025	~410,000	-30%
2026	<300,000	-49%
2027 (projected)	~200,000	-66%

By 2027, the apartment pipeline will deliver fewer than 200,000 units per year. That is 34% below the pre-pandemic average. Meanwhile, household formation shows no sign of slowing. Millennials, the largest generation in U.S. history, are in their late 30s and early 40s. They are forming households. They need places to live. Household formation jumped 41% in a single year (999K to 1.41M) despite a weak job market that lost 173,000 payroll jobs in October 2025 alone.

The Tariff Tax on Every New Home

Building materials got more expensive while housing got scarcer. According to the National Association of Home Builders, tariffs now add $10,900 to the cost of each new home. Building materials costs have risen 40% since December 2020, and over 60% of builders surveyed report higher costs due to tariffs.

Lumber is the biggest hit. Canadian softwood lumber duties rose from 14.5% to 35% (antidumping and countervailing), with an additional 10% Section 232 tariff, for a combined 45% jump. Canada supplies 85% of U.S. softwood lumber imports, roughly 25% of total U.S. supply. Steel and aluminum face a 50% Section 232 tariff.

At current production rates, the tariff burden across all housing starts: $10,900 per home multiplied by 1.359 million starts equals $14.8 billion per year in tariff costs passed to buyers and renters. That is not a rounding error. It is a $14.8 billion annual tax on new housing supply.

The Per-City Exposure Map

The national numbers hide a cruel pattern. RealPage analyst Carl Whitaker published metro-level recovery tranches in February 2026 that divide major apartment markets into three groups based on how quickly they will absorb the boom's oversupply and feel the cliff's undersupply:

Tranche 1 (faster recovery): Atlanta, Nashville. Supply is already pulling back and demand remains solid. These markets will feel the cliff first. Rents start rising in late 2026.

Tranche 2 (uncertain): Orlando, Tampa, Dallas-Fort Worth. Demand is there but supply is still washing through the system. Could go either way depending on the job market.

Tranche 3 (deep laggards): Austin, Denver, San Antonio, Phoenix. These metros built the most during the boom and are the furthest from absorbing it. Austin and Phoenix have strong demand but got overwhelmed by supply. Denver and San Antonio have weaker demand on top of persistent supply pressure. They need a year or more to reach equilibrium.

Here is the paradox: the cities that tried hardest to build are in the deepest hole. Austin permitted more apartments per capita than almost any major metro. It now has the weakest rent growth and highest concession rates in the country. Landlords there offer an average of 26 days free rent on new leases, the most generous concessions since 2011-2012. National occupancy sits at 94.6%, with effective rent growth at just 0.4% year-over-year as of December 2025.

But that is the snapshot, not the trend. RealPage forecasts national rents rising 1.9% by year-end 2026. The cliff has not hit yet. When completions drop below 300,000 in 2026 and below 200,000 in 2027, the cities in Tranche 1 will see rent spikes first. Tranche 3 cities, currently drowning in supply, will feel relief by 2027 and then face their own shortages by 2028 as the construction pipeline runs dry.

What Minneapolis Tells Us About the Zoning Fix

The standard policy prescription is zoning reform: let people build more densely. Minneapolis tried it. The Minneapolis 2040 Plan, approved in October 2019, allowed duplexes and triplexes citywide, eliminating single-family-only zoning.

The Federal Reserve Bank of Minneapolis built a data dashboard tracking 39 indicators with a synthetic control comparison. After four-plus years, the result: no statistically significant impact on most housing metrics. Only 1 of 5 measurable indicators showed a significant difference (housing choice). The Fed's conclusion was direct. "The dashboard does not yet show changes in housing affordability beyond what might have occurred without the plan."

That does not mean zoning reform is pointless. It means zoning reform alone, even aggressive zoning reform in a willing city, operates on a timeline measured in decades, not years. Minneapolis changed the rules in 2019. By 2026, the measurable impact on affordability was effectively zero. The housing deficit grew by 230,000 homes last year alone. The math does not wait for zoning reform to work.

The Affordable Housing Crater

The supply gap is not evenly distributed across income levels. The NLIHC's "The Gap 2026" report, published March 5, 2026, found a shortage of 7.2 million affordable and available rental homes for extremely low-income renters. Only 35 affordable units exist for every 100 extremely low-income renter households. Seventy-four percent of 11 million extremely low-income renters spend more than half their income on rent.

The geographic distribution is stark. Nevada has the worst ratio: 16 affordable units per 100 extremely low-income households. South Dakota has the best at 73 per 100. California's shortage alone exceeds 1 million units. Thirteen of the 50 largest metros have shortages exceeding 100,000 units each. Only 1 in 4 qualifying households receive housing assistance.

The Strongest Case That This Is Not a Crisis

Maybe the supply cliff is demand-matched. If the economy weakens enough, fewer households form, and the housing gap stabilizes or shrinks on its own. The 2025 job market showed genuine cracks: negative 173,000 payroll jobs in October, followed by anemic gains of 56,000 and 50,000 in November and December. For comparison, December 2024 alone added 256,000 jobs. Immigration curtailment further reduces housing demand while simultaneously reducing the construction labor supply needed to build new homes.

This is the strongest version of the "market self-correction" argument. It is also the argument that the housing crisis is best solved by making people poorer.

The data so far contradicts it. Household formation surged to 1.41 million in 2025 despite the weak labor market. Demographic forces are driving household formation more than the job market. Millennials are aging into their household-forming years regardless of payroll numbers. Even if a recession slows formation to 2024 levels (999,000), current starts (1.359M) minus demolitions (~300K) still only produce a net surplus of roughly 60,000 units per year. At that rate, closing a 4.03-million-unit gap would take 67 years.

What We Do Not Know

Different organizations produce significantly different gap estimates. Realtor.com says 4.03 million. Freddie Mac estimated roughly 3.8 million. Moody's Analytics put it at approximately 1.5 million. The methodology differences are real: Realtor.com includes household formation trends, demolition estimates, and vacancy rates; Moody's uses a narrower supply-demand framework. The 4.03 million figure used throughout this analysis is the highest credible estimate, not the consensus.

The NAHB's $10,900 tariff cost per home comes from a builder survey, not a rigorous input-output economic model. Actual cost passthrough depends on substitution effects (domestic lumber production could expand, reducing reliance on Canadian imports) and contract timing (builders who locked in lumber prices before tariff hikes face lower costs). Multifamily start and completion data from Census lags 2-3 months. The Minneapolis 2040 comparison may still be too early to judge. The per-city recovery tranches rely on RealPage's proprietary categorization methodology.

What You Can Do

If you are renting in a Tranche 1 city (Atlanta, Nashville): Lock in a long-term lease now. Rents in these markets will rise first as the supply cliff bites. Concessions (free months, reduced security deposits) are near their peak. Once completions drop below 300,000 nationally, landlords lose their incentive to discount.

If you are renting in a Tranche 3 city (Austin, Denver, Phoenix, San Antonio): You have 12-18 months of leverage. Demand concessions aggressively. The average concession nationally is 7% off asking rent, or about 26 free days. In oversupplied markets, you can push for more. That window closes in 2027.

If you are considering buying: Tariff costs ($10,900 per home) are baked into new-construction pricing now. Existing homes avoid the tariff premium but face limited inventory. Monitor the 10-year Treasury yield, which drives mortgage rates more directly than the Fed funds rate. Every 50-basis-point drop in the 10-year translates to roughly 0.4% off mortgage rates and adds approximately 3% to purchasing power. The spread between mortgage rates and the 10-year is currently elevated at roughly 2.5% vs. a historical average of 1.7%. If that normalizes, mortgage rates drop without the Fed doing anything.

If you vote on local housing policy: Zoning reform is necessary but insufficient on its own. Minneapolis rezoned the entire city. After four years, the Fed found no measurable affordability impact. The constraint is not permission to build. It is the economics of building: construction costs, labor shortages, and tariffs. Policy that addresses only zoning without addressing construction costs will produce the same result Minneapolis got.

The Bottom Line

The biggest apartment boom in 40 years could not close the housing gap. It grew by 230,000 homes in 2025 to 4.03 million, even as 1.498 million homes were completed. Now apartment starts have collapsed 55%. Completions will fall from 584,000 (2024) to under 200,000 (projected 2027). Tariffs add $10,900 per new home. Household formation is accelerating, not slowing. The cities that built the most during the boom are currently the most oversupplied, and the cities that built the least will be the first to feel the cliff. There is no plausible scenario in which the housing gap shrinks before 2030.