2026 Housing Market Forecast by Metro

Across the 13 curated launch metros, the Verus-AI forecast for the 2025–2029 window points to meaningful but uneven appreciation, with metro-level mean five-year forecasts ranging from 28.59% in Houston to 35.32% in Phoenix. The distribution is compressed at the upper end by a model forecast cap of 37.01%, which binds broadly enough that the 75th-percentile forecast across the 7,233 scoreable tracts in those 13 metros already sits at the cap.

Phoenix leads the 13 curated launch metros on both mean Verus score, at 64.02, and mean five-year forecast, at 35.32%. No other metro in the group ranks first on both dimensions simultaneously. At the other end of the distribution, Houston carries a mean Verus score of 51.18 and a mean five-year forecast of 28.59%, the widest gap from the group leader on both measures.

The Verus-AI forecast is built on a history window spanning 2014 through 2024 and projects forward across a five-year horizon covering 2025 through 2029. Scores are expressed on a scale with a maximum of 100. Confidence bands are 80% intervals, meaning the model assigns an 80% probability that realized values will fall within the reported lower and upper bounds, and should not be read as guarantees. The forecast cap of 37.01% is a model-level constraint applied uniformly; any tract whose unconstrained signal would exceed that figure is reported at the cap.

Coverage is broad: the model scores 82,179 U.S. census tracts, representing 84.09% of all 3,234 U.S. counties. The 13 curated launch metros account for 7,233 of those scoreable tracts. Of the 82,179 scoreable tracts nationally, 23,955 are reported at the forecast cap, indicating that a meaningful share of the scored universe is constrained rather than freely estimated at the top of the distribution. Readers should treat cap-bound forecasts as a floor signal, the model's confidence that appreciation will be at least substantial, rather than a precise point estimate.

The score distribution across the 82,179 scoreable U.S. census tracts runs from a 10th percentile of 37.0 to a 90th percentile of 68.0, with a median of 54.0. The forecast distribution across the same population shows a 10th percentile of 15.7% and a 50th percentile of 30.48%, while both the 75th and 90th percentiles sit at the cap of 37.01%.

Among the 13 curated launch metros, Phoenix ranks first on mean five-year forecast at 35.32%, followed by Nashville at 34.79% and Raleigh at 34.05%. Houston ranks last among the 13 curated launch metros at 28.59%, a gap of 6.73 percentage points from Phoenix. The ranked table below presents the full ordering; the discussion here focuses on what the spread implies rather than the sequence of cells.

The most structurally notable feature of the ranking, across the 13 curated launch metros, is the prevalence of the forecast cap. In Phoenix, the 25th-percentile forecast is already at 37.01%, meaning at least three-quarters of the metro's 998 scoreable tracts are at or above the cap. Nashville's 25th percentile sits at 37.0%, effectively at the cap as well.

Phoenix also leads the 13 curated launch metros on mean Verus score at 64.02, with a median of 65.0. Nashville follows at 61.29, and Las Vegas at 60.23. Atlanta and Houston trail the group, with mean Verus scores of 52.65 and 51.18, respectively. The score gap between Phoenix and Houston, 12.84 points on the mean, is the widest pairwise spread among the 13 curated launch metros and suggests the divergence in forecast outcomes is not purely a function of the cap but reflects underlying compositional differences in tract quality across the two markets.

Jacksonville, with 218 scoreable tracts, is the smallest metro in the group by tract count and carries a mean Verus score of 54.32 and a mean forecast of 32.98%. Its weighted median value of $281,200 is the second-lowest among the 13 curated launch metros, above only San Antonio at $244,000. Dallas-Fort Worth, with 1,481 scoreable tracts, is the largest metro in the group by tract count and sits near the middle of the ranking on both score and forecast, with a mean Verus score of 58.26 and a mean five-year forecast of 32.93%.

The most structurally interesting feature of the data, across the 13 curated launch metros, is the prevalence of the forecast cap and what it obscures. When the 25th-percentile forecast for a metro is already at 37.01%, as it is in Phoenix and Nashville, the mean forecast figure compresses the true dispersion of the underlying signal.

The contrast with Houston is instructive. Houston's forecast distribution has a 10th percentile of 14.0% and a median of 32.21%, with the cap not binding until the 75th percentile. That shape, wide, with a long left tail, suggests genuine heterogeneity within the metro: some tracts are forecast to appreciate modestly while others reach the cap. The interquartile range for Houston, from 23.22% at the 25th percentile to 37.01% at the 75th, spans nearly 14 percentage points, the widest among the 13 curated launch metros. Atlanta shows a similar pattern, with a 10th percentile of 17.32% and a 25th percentile of 26.43%, indicating that the lower quartile of Atlanta tracts carries materially weaker appreciation signals than the lower quartile in Phoenix or Nashville.

Aldine in Houston, for example, carries a current value of $138,521 and a five-year forecast of 37.01%. Tract 1111 in San Antonio shows a current value of $236,800 and a forecast of 37.01%. Both are at the cap, but the absolute dollar appreciation implied, and therefore the relationship between entry price and projected terminal value, differs substantially.

At the tract level, the top national list is dominated by Phoenix, which places six of the top ten entries in the composite-score ranking across the 7,233 scoreable tracts across 13 major U.S. metros. Nashville's Tract 1900 is the only non-Phoenix entry in the top five of that list, with a composite score of 85.07, a current value of $453,455, and a forecast terminal value of $621,273. The forecast chart for Tract 4204 in Phoenix illustrates the model's trajectory in detail: observed values rose from $120,400 in 2014 to $331,000 in 2024, and the Verus-AI forecast projects a terminal value of $453,499 by 2029, with an 80% confidence interval running from $389,923 to $527,441 at the terminal year. The band width of $137,518 at the terminal year reflects the model's honest acknowledgment that a ten-year extrapolation carries meaningful uncertainty.

The structured data does not supply explicit rental yield figures, so no yield calculations are presented here. However, the current-value data embedded in the tract-level tables provides the denominator for any yield analysis a reader wishes to construct independently. The dispersion in current values across the top-ranked tracts is wide: within Phoenix alone, the top ten tracts range from $330,209 at Tract 9500 to $547,700 at Tract 1036.

The more analytically relevant observation is that metros with lower weighted median values and cap-level forecasts present a different yield profile than metros with higher median values and similar forecasts. San Antonio's weighted median value of $244,000 and Houston's of $272,400 sit well below Phoenix's $427,690 and Raleigh's $446,667. If gross rental income scales with value at a roughly constant rate across markets, an assumption the data does not confirm or deny, then lower-value metros with comparable appreciation forecasts would imply higher gross yields at entry. Readers should treat this as a framing observation rather than a modeled conclusion, given the absence of rental income data in the current dataset.

Within the top national tract list, the range of current values is also notable. Tract 4408 in Jacksonville carries a current value of $796,426 alongside a Verus score of 76 and a forecast of 37.01%, while Alamo Heights in San Antonio shows a current value of $157,362 with a score of 77 and a forecast of 37.01%. Both are at the cap, but the yield and total-return arithmetic for an investor entering at $796,426 versus $157,362 differs in ways the forecast percentage alone does not capture. The ranked tables render these pairings explicitly, and the current-value column warrants as much attention as the forecast column for any yield-sensitive analysis.

The Verus-AI forecast is a model-derived estimate, not a guarantee of realized appreciation. The 80% confidence intervals reported for the leading tract series illustrate the range of outcomes the model considers plausible: for Tract 4204 in Phoenix, the terminal-year 80% interval spans from $389,923 to $527,441, a band width of $137,518. That interval widens monotonically from $47,660 in the first forecast year to $137,518 by 2029, which is the expected behavior of a well-calibrated forward model, uncertainty compounds over time. Readers should not treat the point forecast as the central scenario without also considering the width of the band.

The forecast cap of 37.01% introduces a specific interpretive risk. For the 23,955 tracts across the 82,179 scoreable U.S. census tracts that are reported at the cap, the model is communicating that the unconstrained signal is at least as strong as the cap, but not providing a precise estimate of how much stronger. In metros where the cap binds broadly, Phoenix most prominently among the 13 curated launch metros, where the 25th-percentile forecast is already at the cap, the reported mean forecast understates the dispersion of the underlying signal. Analysts comparing metros on mean forecast should be aware that a higher mean in a cap-bound metro may reflect a higher proportion of cap-bound tracts rather than a uniformly stronger signal.

The history window of 2014 through 2024 encompasses a period of unusually strong national home price appreciation, including the post-2020 acceleration. A model trained on that window may embed structural assumptions about appreciation rates that do not persist into the forecast period. The data does not provide a mechanism to adjust for this, and the analysis does not attempt to do so. Additionally, the coverage rate of 84.09% across 3,234 U.S. counties means that roughly 16% of counties are not represented in the scored universe; conclusions drawn from the 82,179 scoreable U.S. census tracts should not be extrapolated to the full national housing stock without that caveat in mind.

Climate, demographic, and macroeconomic variables that could affect realized appreciation over the 2025–2029 horizon are not individually disclosed in the structured data. The risk grades assigned to individual tracts, Low or Moderate, as shown in the top national tract list, are labels derived from the model and should be interpreted as relative designations within the scored population rather than absolute assessments of investment risk. No insurance, lending, or regulatory implications are asserted or implied by those labels.