The apartment supply wave crested in 2024. The bottleneck moves to the planning department next.

608,000. That is how many multifamily units were completed in the United States in 2024, the most in any single year since 1986, according to NAHB's tabulation of Census Bureau construction data (see NAHB's February 2026 multifamily outlook). National vacancy ran up to a record 7.3% by December 2025 as that supply kept working through the market, per Cotality economist Molly Boesel. Rents fell roughly 1% year over year nationally. A record delivery year produced a landlord's market's worth of empty units, and the reason is not complicated once you look at when those 608,000 apartments actually got their debt.

They were financed in 2021 and 2022, when the ten-year Treasury sat below 2% and construction lenders were still writing checks on assumptions from a demand boom that had already peaked. Multifamily starts hit 547,000 units in 2022, the top of the cycle. Then the Fed did what it did, and starts fell to 355,000 by 2024, a 35% drop in two years. Everything that broke ground in the good years landed on the market together in 2024 and early 2025, into a demand environment the loans were never underwritten for. That is the whole autopsy. Oversupply in 2024 was not a forecasting failure. It was a three-year-old rate decision arriving on schedule.

Why did the pipeline collapse so fast, and where did it collapse worst?

Wherever it had been fullest. RealPage's third-quarter 2025 data update shows the deepest rent cuts concentrated in exactly the metros that built the most: Denver and Austin down roughly 8% year over year, Phoenix and San Antonio down about 5% (see RealPage's Q3 2025 analytics). Nationwide, effective asking rents fell 0.3% in the quarter itself, the first mid-year rent cut RealPage has recorded since 2009, with occupancy slipping 30 basis points to 95.4% after five straight quarters of gains. Nearly 22% of apartments nationwide were running concessions that quarter, averaging 6.2% off face rent.

Meanwhile the opposite metros were fine. San Francisco, New York, San Jose, and a cluster of Midwest markets posted rent growth in the same quarter RealPage recorded the first national mid-year decline in sixteen years. Same country, same interest-rate environment, opposite outcomes, and the variable that explains the split is not demand. Household formation did not diverge that sharply between Austin and New York. Supply did. Multifamily Dive's reporting on the 2025 cycle put coastal deliveries at roughly 1.4% of existing stock for 2025, against a Sun Belt pipeline running a full percentage point higher, which is exactly the gap you would expect given which markets absorbed the hangover and which never got the wave at all.

Coastal markets never got the wave in the first place, mostly because they cannot.

The forward numbers make the point even harder to argue with. Austin, the poster child for the Sun Belt building boom, is on pace to deliver as few as 4,600 new units in 2026 by CoStar's count, a roughly 74% collapse from the roughly 17,500 units it delivered in 2025 (see Pew's coverage of the Austin cycle). RealPage and the local apartment association both put the 2026 number somewhat higher, in the 10,000-to-13,000 range, but every estimate agrees on direction and rough magnitude: the pipeline that pushed Austin inventory up 32% since 2023 is emptying out fast. Charlotte grew 27% over the same stretch, Nashville 24%, Phoenix 23%, and Yardi Matrix is projecting Phoenix, Denver, Austin, Dallas, and Orlando as the weakest rent-growth markets again in 2026, per GlobeSt's June 2026 reporting. The same five metros that overbuilt from 2022 through 2024 are the five metros still working off the hangover two years later. Oversupply does not clear on a calendar. It clears unit by unit, lease by lease, and the markets that built the most take the longest.

Is the starts data even telling the truth?

Depends whose desk you are reading it from, and that disagreement is itself a signal worth sitting with. Census's own July 2025 print showed multifamily starts at a seasonally adjusted annual rate of 470,000 units, up sharply month over month. John Burns Research & Consulting's Chris Nebenzahl put private client and survey data 8% to 10% lower year over year in the same window, and said flatly that "the Census data doesn't represent what we are seeing from our clients, third-party data providers that we subscribe to, or what our surveys are saying" (via Multifamily Dive's reporting on the discrepancy). NAHB's Robert Dietz offered a defense: Census may be catching smaller-market, lower-density starts that private trackers under-sample. Maybe. But an analyst underwriting a national portfolio off a single government print in 2026 is making a bet on which survey methodology is wrong, whether they know it or not.

Take the forecast with that caveat attached. NAHB's own model still expects starts to rise to 413,000 in 2025, a 16% jump, before falling back to 392,000 in 2026 and 367,000 in 2027 as the current financing environment works through the pipeline. Directionally, every source agrees on the shape: a short-lived bounce off the 2024 trough, then a longer grind lower. Nobody credible is calling for a return to 2022's 547,000-unit pace anytime soon.

What does the country actually need, and who is positioned to build it?

More than the current pipeline is going to produce. The National Multifamily Housing Council and National Apartment Association's widely cited study puts the shortfall at 4.3 million apartments by 2035, made up of a roughly 600,000-unit existing deficit plus 3.7 million units of unmet future demand, with Texas, Florida, and California alone accounting for 40% of that need (see the NMHC/NAA demand study, via Connect CRE). That study is three years old now, and the last two years of underbuilding have not made the gap smaller. A country that just delivered its biggest apartment year since 1986 is still, on the Council's own math, behind schedule.

Here is where entitlement stops being background noise and becomes the actual constraint. NAHB and NMHC's joint 2022 regulatory-cost survey found that government regulation across every level accounts for 40.6% of the total cost of a multifamily development, with changes to building codes over the prior decade alone eating 11.1% and site-work-related requirements another 8.5% (see Eye on Housing's summary of the NAHB/NMHC study). Almost half the developers surveyed, 47.9%, said they simply avoid jurisdictions with inclusionary zoning. A staggering 87.5% avoid jurisdictions with rent control entirely. That is not a rounding error on a pro forma. That is a national map of where capital will not go, drawn by planning departments rather than by demand.

Forty cents of every construction dollar is a decision made in a zoning file, not a lumber yard.

Every recovery cycle after a starts collapse runs into the same wall: the parcels that are actually entitled, financeable, and shovel-ready right now are the ones an underwriter should be paying up for, and there are fewer of them than there were in 2021. States and cities know this. Massachusetts's Chapter 40B safe-harbor mechanism, which we have covered in Braintree, exists specifically because ordinary local zoning was not going to produce enough affordable-adjacent supply on its own, and even that safe harbor turns on an arithmetic fight over a single percentage point. California's Builder's Remedy, tested hard in Beverly Hills, exists because a state legislature decided the local entitlement process itself was the housing shortage's root cause, not a neutral referee over it. Both mechanisms are proof, at the level of an actual project file, that the industry has stopped pretending entitlement is a formality standing between a good site and a shovel.

What does this mean for anyone underwriting the next cycle?

Stop screening for land. Start screening for the zoning file. A parcel with an approved site plan, a resolved traffic study, and a closed public-comment period is worth materially more in 2026 than the same dirt with none of that done, because the entitlement queue itself has become the scarce resource, not construction capacity or even debt. Lenders have already priced this; that is why nearly 40% of quick-service franchise operators are turning to reverse sale-leasebacks to raise cash rather than wait on new construction loans, a pattern from an adjacent asset class that tells you the same thing multifamily lenders are quietly pricing: hold costs on unentitled land are no longer a rounding error. Our own multifamily coverage reads jurisdiction by jurisdiction for exactly this reason, because a national supply forecast tells you nothing about whether the specific parcel in front of you clears its specific planning commission before the debt comes due.

My call, and it is checkable against next year's data: the 2026-2027 rent recovery will not be broad. It will show up first, and most sharply, in the handful of coastal and inland-Northeast metros that already under-delivered through the 2024 wave, while several Sun Belt metros post flat-to-negative rent growth again in 2026 even as national vacancy improves. If Austin, Phoenix, and Denver post rent growth above the national average before 2027, I was wrong about how sticky this oversupply is. Either way, the next apartment cycle is not going to be won on cap rate. It is going to be won by whoever already has the entitlement done.

This analysis is a source-cited research summary drawn from public records and industry reporting, not legal or financial advice. It can contain errors and should be verified independently before any investment decision.