Two demand regimes, one tape: AI names chase highs while a 5.1% 10-year prices the housing trade

The tape almost always looks flat when it isn’t

On Friday, September 25, 2026, the S&P 500 closed at 7,747, up about half a percent, with the Nasdaq 100 up a similar amount and the Dow leading at +0.9% ^[1]. A headline like “broadly higher” would be technically true and analytically useless. The interesting part is underneath: the index is masking a market that is splitting into two distinct demand regimes, and the eight names in this research sweep example the divide cleanly.

One group is accelerating on customer demand that management is describing in terms of durability, and the market is paying up for it. The other is defending margins against a 10-year Treasury yield near 5.1% — a number that does most of the heavy lifting on any housing-adjacent discretionary purchase. And one name, Leslie’s, is no longer a thesis to debate; it is a thesis that failed.

Regime one: AI demand is still the cleanest tell

Datadog closed the day at $268.13, up 4.4%, and kept extending in the aftermarket, approaching its 52-week high amid AI-observability momentum ^[2][1]. There was no single same-day catalyst — the move sits on top of a seven-day run near 9% and a thirty-day advance — which matters. This is a stock being re-rated on a sustained narrative, not a headline: a broad cloud-sector optimism tied to AI-linked product launches and continued institutional engagement ^[1].

The demand evidence is on the tape. Over the course of its 2025 calls, Datadog management described customers moving AI experiments into production and an observability platform built to catch that traffic — roughly 5,500 customers using one or more Datadog AI integrations as of early 2026, with the number of LLM-observability users doubling and the volume of spans sent up tenfold in six months ^[3]. CEO Olivier Pomel framed it as a “generational growth opportunity,” churn described as low and revenue retention stable in the mid-to-high 90s — the language of a mission-critical platform, not a discretionary product ^[3].

Snowflake is the same story with harder numbers. In its Q2 fiscal-2027 report on September 2, product revenue came in at $1.49 billion, up 37% year over year — the third consecutive quarter of acceleration — and the company raised its full-year product revenue forecast on cloud and AI demand, sending shares sharply higher ^[4]. Net revenue retention sits at 125%, and management attributed roughly half of recent outperformance to AI product adoption ^[4][5]. Snowflake reports next on the calendar-2026 deck; Datadog’s is scheduled for November 5 before the open ^[5].

Regime two: the housing-adjacent group is priced off the 10-year

The second regime has a different lead indicator: the 10-year Treasury at 5.11% in the latest read, after peaking near 5.19% in the global bond selloff that dominated the week ^[6][7]. That yield is the swing factor for furniture, mattresses and big-ticket home discretionary spend, and it explains why this group rises and falls more on rates than on any single earnings print.

The context is uncomfortable on paper. Real GDP is growing at about 2.1% year over year, unemployment is 4.1% and the Fed funds rate sits at 3.63% — a soft-landing picture at first glance ^[6]. Strip out the near-term and the macro fingerprint is closer to the mid-2000s analog the data flags: a 10-year above 5%, a barely-positive yield curve, and consumer sentiment under pressure at 55.2 ^[6]. Median monthly housing payments are back near multi-year highs, which flows straight into whether a consumer buys a new couch or a bed ahead of a move ^[8].

Within that, the operators are splitting on execution, not on the macro. Williams-Sonoma has roughly doubled over two years and keeps producing solid quarters and expanding operating margins even in a sluggish housing market — an outcome that nudged the stock up again on September 20 ^[8][7]. RH reported Q2 net revenue up 2.6% to $922.2 million with an 18.3% GAAP EBITDA margin, including 600 basis points of tariff benefit in the quarter ^[8]. These are margin-defense stories: they are profitable, but they are not demand-acceleration stories in the way regime one is.

The smaller furniture names — Ethan Allen and La-Z-Boy, plus Tempur Sealy on mattress demand — round out the group and were roughly flat on the day ^[2]. The current pulse across these housing-adjacent names is summarized below (LESL aside; its pricing data did not come through fresh to all tools this session, and it is treated separately).

The falsifier: Leslie’s is the counter-example the thesis needed

A thesis is only as good as the case that breaks it, and Leslie’s plays that role here. The pool and spa retailer fell roughly 14% on Friday to around 29 cents after reports that it is preparing to file for Chapter 11 bankruptcy protection as soon as next week ^[9]. The distress is not new — it has been building for a quarter. Fiscal Q3 revenue fell 8.4% year over year to $458.5 million with comps down 6.2%, the company withdrew guidance, and it flagged substantial doubt about its ability to continue as a going concern, all in August ^[9]. S&P had already downgraded the company to CCC- in late August on a deemed-likely distressed exchange and a projected liquidity shortfall ^[9].

Leslie’s belongs to the same broad discretionary-home family as the other consumer names but fails for a company-specific reason: its credit situation makes its fate a balance-sheet event, not a demand question. Its next report is calendared for December 1 after the close ^[5]. The lesson it carries for the rest of the group is the one the group’s better-run members are already proving — in this rate environment, execution and balance sheet separate the survivors from the stories.

What to watch next

The single most important number for the housing-adjacent half of this table in the next quarter is not any one company’s comp — it is the 10-year Treasury. Each of these operators can defend margin, but a sustained 5.1%+ yield keeps the demand ceiling low because it pushes the monthly cost of new-home discretionary purchases up. Watch whether RH (December 10 after the close), Williams-Sonoma (November 18 before the open), La-Z-Boy (November 17 after the close) and Datadog (November 5 before the open) each confirm or contradict their current read ^[5].

For the software half, the tell is retention and product acceleration: Datadog’s AI-integration customer count and LLM-observability adoption, and Snowflake’s net revenue retention and whether its third straight quarter of product-growth acceleration extends. What would have to be true for each side of this split to hold is different — AI software holds if enterprise budgets replicate the last two quarters; housing-adjacent names hold if rates stabilize and the marginal buyer returns. As of the September 25 close, the evidence supports the first thesis far more strongly than it supports the second, and Leslie’s sits squarely in the column against.

Research commentary for educational purposes; not personalized investment advice.