Homebuilders cut specs—and now unsold completed inventory has pulled back from recent highs

Today’s letter is brought to you by the Lennar Investor Marketplace!

Speaking at the Bank of America Housing Symposium in June 2025, Toll Brothers CEO Doug Yearley—who has since stepped down—acknowledged that parts of Arizona, Florida, and Texas were dealing with spec inventory “overhangs” that he said would eventually “clean up [over time] because the builders are starting fewer spec homes in the softer market, and I think that will naturally work its way out.”

At the height of the Pandemic Housing Boom, when nearly everything homebuilders were building was flying off the shelves, there were only 32,000 unsold completed new-build homes in March 2022. Once the boom fizzled out, that figure quickly began to rebound—especially in Sun Belt boomtowns—reaching a high of 134,000 unsold completed new-build homes by December 2025.

However, data published this week shows that the number of unsold completed new-build homes has, at least for now, fallen to 119,000 as of March 2026. While the count of unsold completed new-build homes is still up year-over-year (there were 113,000 unsold completed in March 2025), the decline over the past few months has been larger than seasonality alone would suggest.

To put the number of unsold completed new single-family homes into better historic context, we have the ResiClub Finished Unsold New Homes Supply Index. It accounts for unsold completed inventory relative to new home sales. A higher index score indicates a softer national new construction market with greater supply slack, while a lower index score signifies a tighter new construction market with less supply slack. Over the past few months, that reading has almost drifted back down into the “historically normal” range.

After experiencing a softer 2025 than expected—and greater-than-expected margin compression—many giant homebuilders told analysts heading into 2026 that they’d pivot toward fewer spec builds and more build-to-order homes. The reason was simple: build-to-order margins are materially higher. Built-to-order homes tend to generate higher margins because they’re sold before construction begins, reducing inventory carrying costs and the risk of having to deploy larger incentives to sell them.

Doing fewer specs and starts in softer pockets of the Sun Belt, has already helped some of the builders reduce their count of unsold completed homes. Just look at America’s largest homebuilder D.R. Horton.

While the U.S. Census Bureau doesn't give us a greater market-by-market breakdown on these unsold completed new-builds, we have a good idea where they are based on total active inventory homes for sale (including existing)—likely much of it is in the Mountain West and Sun Belt, particularly around the Gulf.

We should point out that while many markets in Texas and Florida experienced significant post–Pandemic Housing Boom inventory bounce back, that inventory growth has decelerated in recent months. In fact, many parts of Florida are now seeing year-over-year active inventory for sale declines. The heavy discounting by homebuilders in weaker pockets of Texas and Florida to move unsold inventory—combined with reduced housing starts and fewer spec builds in those pockets heading into 2026—has, in part, contributed to that slowdown in inventory growth.

Unlike the existing-home market—where U.S. existing-home sales are still -23.6% below pre-pandemic 2019 levels—U.S. new-home sales are essentially on par with pre-pandemic 2019 levels right now 👇

Why haven’t U.S. new home sales come down more given the affordability picture and what’s happened in the existing-home market?

A lot of it boils down to the fact that many homebuilders since the Pandemic Housing Boom fizzled out have done larger affordability adjustments—including everything from bigger buydowns, more money back at close, and even outright price cuts—in order to keep moving product when they run into softness in a given neighborhood. The most aggressive homebuilder on the incentive front is Lennar. Last quarter, Lennar spent the equivalent of 14% of the final sales price on sales incentives. For a $400,000 home, that translates to $56,000 in incentives. Lennar’s cycle low in Q2 2022, when it spent 1.5% of the final sales price on sales incentives.

In order to do bigger incentives—and pay for sticky land prices—homebuilders have been compressing margins. Indeed, all 11 of the major publicly traded U.S. homebuilders that ResiClub tracks the most closely have seen year-over-year gross margin compression.

So, in other words, big homebuilders have been willing to adjust prices and incentives in order to maintain sales volume, while existing home sellers, in aggregate, have fought harder against price adjustments—at the expense of speed of sale and turnover. Another factor is that homebuilders willingness to sell isn’t impacted by so-called affordability “lock-in.” Ever since mortgage rates spiked, high switching costs have left many homeowners either unwilling or unable to sell and buy at today’s prices and rates, further suppressing existing-home turnover.

Before we conclude today’s new construction report, here’s a historic look at nationally aggregated permits:

PulteGroup—America's third-largest homebuilder and No. 229 on the Fortune 500—is partnering with San Francisco-based smart panel maker SPAN on a product that could give builders a new value proposition: XFRA, a distributed data center solution that puts AI compute nodes inside residential homes.

PulteGroup believes that the solution could help reduce homeowner costs.

How does it work?

SPAN will install its smart electrical panel plus a liquid-cooled compute node loaded with NVIDIA GPUs in some new PulteGroup homes. The panel intelligently manages the home's spare electrical capacity and routes it to power the compute node, which then runs AI inference tasks—essentially doing the behind-the-scenes computation when someone uses an AI application. The homeowner gets a smarter panel and battery backup; SPAN and its cloud partners get a slice of distributed computing power inside the home.

Based on the press release, the only financial benefit mentioned is discounted electricity and internet rates. But it doesn't say how much the savings would likely be. The press release doesn’t mention anything about the homeowner getting a hosting fee. The value exchange seems to be: The homeowner lets them put this compute node in their home, and in return get the SPAN panel, battery backup, and lower utility/internet rates.

Where does PutleGroup build single-family homes? Pulling from the ResiClub Terminal, we made the map below last year:

Zillow economists use an economic model known as the Zillow Market Heat Index to gauge the competitiveness of housing markets across the country. 

This model looks at key indicators—including home price changes, inventory levels, and days on market—to generate a score showing whether a market favors sellers or buyers. 

Higher scores point to hotter, seller-friendly metro housing markets. Lower scores signal cooler markets where buyers hold more negotiating power.

According to Zillow: 

Nationally, Zillow rates the U.S. housing market at 55 spring 2026.

That said, Zillow’s reading varies significantly across the county.

Click here to view an interactive version of the map below

Among the 250 largest metro area housing markets, these 20 are the HOTTEST markets, where sellers have the most power:

Among the 250 largest metro area housing markets, these 20 are the COLDEST markets, where buyers have the most power:

Does ResiClub agree with Zillow’s assessment?

Directionally, ResiClub agrees with Zillow’s identification of regional housing markets where buyers have gained the most leverage—particularly along the Gulf Coast—as well as markets where sellers have retained relatively more control, including parts of the Northeast and Midwest. That said, while many Northeast and Midwest markets still tilt toward sellers compared to the rest of the country, ResiClub believes Zillow somewhat overstates the degree of seller strength based on other market indicators.

Based on ResiClub’s housing analysis, SWFL Florida and chunks of Central Texas (particularly areas with a lot of new single-family home construction outside Austin) are the softest chunk of the U.S. housing market where buyers have the most power. Not too far behind are certain pockets of Colorado, Georgia, and Arizona markets where homebuilders in certain areas have been clearing some built up unsold spec inventory. 

What did this Zillow analysis look like back in spring 2021 during the Pandemic Housing Boom? Below is Zillow’s March 2021 reading—published in April 2021.