New home construction in the United States has fallen to its lowest level since 2020. The "1996-style soft landing script" is now entering the stress test phase.

The sharp decline in apartment projects dragged down the US new residential construction to the weakest level in six years. Federal statistics released on Tuesday showed that housing starts in the US dropped sharply by 15.4% in May, translating to an annual rate of only 1.18 million units. This data was lower than all the economists' predictions covered in the survey, marking the lowest statistic since May 2020. This decline was mainly driven by a 40.2% plummet in multifamily housing starts. Single-family housing starts also unexpectedly decreased by 1.9%, with an annual rate of 882,000 units. Despite signs that the US real estate market is hinting at a soft landing for the economy, well-known Wall Street investment firm Piper Sandler insists on comparing the current cycle to the tech investment boom in 1996, rather than the imminent burst of the internet bubble in 1999. The federal statistics report on Tuesday indicated that contractors continued to exercise restraint in the face of weak demand, actively absorbing the supply of unsold new homes. Many contractors have discounted prices and subsidized mortgage rates for clients in order to find buyers, while builders have significantly slowed down the production of "speculative model homes," which are homes built before a sale contract is signed. The latest US real estate market statistics may be somewhat lukewarm, but they do not completely overturn Piper Sandler's chief global economist Nancy Lazar's latest prediction of a "mild recovery" in the US economy. Instead, they indicate that the current housing cycle has not entered a speculative expansion phase yet as builders continue to lower prices, subsidize mortgage rates, and reduce speculative construction. The economy seems to be more in a phase of rebalancing after the long period of high interest rates maintained by the Federal Reserve, rather than in the slumping period of the US economy and housing market on the eve of 1999. The total number of building permits, which point to future construction activity, unexpectedly saw a slight decline of 0.7% in May, translating to an annual rate of 1.41 million units. Single-family housing permits saw a slight increase. President Donald Trump has been trying to encourage builders to increase housing production in order to improve housing affordability before the midterm elections in Congress. The president has criticized housing builders on social media for "sitting on over 2 million vacancies," and has proposed a ban on institutional investors buying single-family rental homes. Legislation aimed at restricting some institutional investors from buying homes, although weakened from its initial version, is still pending in Congress. Construction activity in almost all regions of the US, excluding the Midwest, saw a decline. In the largest residential construction area in the South, housing starts plummeted to the lowest level since May 2020, dragged down by a sharp drop in the construction of apartment buildings. The US national Association of Realtors will release the May report on existing home sales on Wednesday, providing a glimpse of the US resale housing market. As the US and Iran seem to be getting closer to a long-term ceasefire agreement, if energy trade volumes gradually return to pre-war levels through the Strait of Hormuz, with international oil price benchmarks like Brent crude futures peaking and inflation pressures easing, real income in the US will improve, providing new support to the consumption side of the economy. The housing market may see gradual recovery with the three factors of rising nominal personal income, falling new home prices, and potential further decline in mortgage rates. Piper Sandler compares the current cycle to 1996 rather than 1999, with the key difference being that the policy environment is not clearly out of control: after a 75 basis point rate cut by the Fed in 1995, the economy accelerated again, while after a cumulative 175 basis point rate cut from 2024 to 2025, there was a growth rebound led by industrial activity in the second half of 2025. The implicit judgment is that as long as the Fed maintains relative restraint and inflation expectations are met, the economy is more likely to follow the typical soft landing path of the US economy in 1996, characterized by "improvement in household/consumer income - consumption recovery - gradual warming of housing," rather than a super bubble created by overly loose monetary policy as seen in 1999. However, the latest housing start data adds an important constraint to this optimistic narrative: the recovery may be slower and more differentiated, rather than linearly upward. Especially with building permits decreasing by 0.7% to an annual rate of about 1.41 million units, with a slight increase in single-family permits, it indicates that future supply has not completely stalled, but the sharp decline in multifamily housing and the lowest housing starts in the South since May 2020 reflect that housing affordability remains a hard constraint. Nancy Lazar's latest prediction of a so-called "mild recovery in the housing market" emphasizes "bottom repair after marginal affordability improvement," rather than re-entering a strong expansion cycle in real estate; it needs oil prices to fall, inflation to cool down, long-term interest rates to decline, and continued growth in household income to work together. If the development follows the path predicted by Piper Sandler's Lazar, namely, the US economy is closer to that of 1996, the market should continue to lean towards a soft landing trade, with resilient consumption, bottoming out in the housing chain, acceleration in industrial activity, coexistence of growth stocks and AI capital expenditures; if oil prices rebound, inflation remains high, and mortgage rates start climbing again, then the weakness in housing data will shift from being evidence of rate cuts to being important evidence of demand decline. Piper Sandler emphasizes that for a super bull market dominated by AI computing power, the "1996-style path" is most favorable because it implies that nominal growth remains stable, interest rate pressures ease, corporate capital expenditure remains sustainable, and valuations in the AI semiconductor chain, data center optical interconnection/DCI interconnection, storage chips, and data center power infrastructure chains can continue to receive strong discount rate support; while the "1999-style path" means that the market enters a stage of high valuation, late cycle, and gradual bubble breakdown, with tech stocks becoming more sensitive to any rebound in inflation expectations or US bond yield curve. The latest housing start data imposes an important constraint on this optimistic narrative: the recovery may be slow and differentiated rather than linearly upward. The fact that building permits decreased by 0.7% to an annual rate of about 1.41 million units, with a slight increase in single-family permits, indicates that future supply has not completely stalled. However, the sudden decline in multifamily housing construction and the lowest housing starts in the South since May 2020 suggest that housing affordability remains a key constraint.