AI demand is holding up better than the market’s duration trade

The September 23 tape offered a useful stress test for the idea that earnings growth and resilient demand can carry a mixed basket of software and home-related companies over the next year. The answer is not uniform: AI-linked usage remains a tangible operating signal, while higher long-term yields and softer consumer confidence are making the valuation and demand bar more demanding.

The opening snapshot is a split signal

At the 16:00 ET close, SPY fell 0.72%, QQQ fell 0.84%, and DIA fell 0.71%. The AP attributed the session’s pressure to a stronger economic report, renewed inflation worries, higher bond-market pressure, and a halt in oil’s decline.^[1][2]

That is a meaningful backdrop for growth stocks. The latest macro snapshot shows unemployment at 4.1%, real GDP growth at 2.1% year over year, CPI inflation at 3.35%, and the 10-year Treasury at 4.96%. Credit spreads remained contained at 2.66% and the VIX was 14.81, so this looks more like a valuation and rates test than a broad financial-stress event.^[3]

Within the watch set, the close was notably uneven: DDOG rose 1.62%, while SNOW fell 0.53%, RH fell 4.63%, WSM fell 2.09%, ETH fell 2.66%, LZB fell 2.12%, and LESL fell 5.27%. The TPX quote returned an old February 2025 timestamp rather than a current September 2026 print, so it is excluded from today’s comparison rather than treated as a live signal.^[1]

The strongest evidence for the hypothesis is operational, not thematic

Datadog’s Q2 2026 results are a clear positive data point. Revenue grew 36% year over year to $1.12 billion, the company ended the quarter with about 33,400 customers, and its $100,000-plus ARR customer count rose to roughly 4,720 from about 3,850 a year earlier.^[4]

Management’s earnings-call explanation also matters. Datadog said revenue growth accelerated across both AI-native and non-AI customers, with non-AI growth reaching the high 20s percent year over year. The company tied that performance to continued adoption of cloud and modern technologies, alongside rising use of its platform to observe, secure, and act on AI workloads.^[5]

Snowflake provides a similar, though not identical, signal. In its latest call, management said AI is adding workloads, extending the platform to new users, and increasing consumption. It also raised FY2027 operating-margin guidance from 13.5% to 14.5%, while acknowledging that AI workloads carry lower product gross margins today.^[6]

That combination is important: demand can be real without being frictionless. Snowflake’s own comments describe both a growth engine and a margin-mix issue. The bullish case requires AI workloads to scale faster than their current cost burden; the cautious case says consumption growth may normalize while infrastructure costs remain visible.

Consumer and home exposure is a harder test

The home-related names do not provide a single clean read. Williams-Sonoma reported Q2 comparable brand revenue growth of 6.2%, a 22.9% GAAP operating margin, and raised its full-year 2026 outlook.^[7]

RH’s Q2 revenue increased 2.6% to $922.2 million, but its shareholder letter also highlighted a 600-basis-point tariff benefit embedded in adjusted EBITDA. That makes the headline margin picture less straightforward: reported resilience is present, but investors still have to separate underlying demand from pricing, mix, and trade-related effects.^[7]

For WSM, RH, ETH, LZB, and LESL, the key question is not whether consumers are spending at all. It is whether discretionary and replacement demand can remain durable while sentiment is weak and financing costs are elevated. Consumer sentiment was 55.2 in the latest macro snapshot, down 10.53% year over year even after a month-over-month improvement.^[3]

What the basket says so far

The hypothesis therefore passes an initial evidence check in software, especially where demand is visible in usage, customer additions, and production workloads. It is less established in home-related names, where company-specific execution and the consumer’s willingness to fund larger purchases matter more than a broad “resilient demand” label.

What to watch next

1. AI workload conversion: Look for evidence that pilots are becoming production workloads at DDOG and SNOW, rather than merely expanding product announcements.
2. Consumption quality: Track retention, usage growth, optimization, and gross-margin commentary together. Strong revenue with worsening unit economics would weaken the thesis.
3. Home demand breadth: WSM’s follow-through, RH traffic and order trends, and demand commentary from ETH, LZB, LESL, and TPX will help distinguish a company-specific story from a broad consumer cycle.
4. Rates versus earnings: A 10-year yield near 5% raises the hurdle for long-duration growth equities even when operating results are solid.^[3]
5. Data quality: TPX needs a fresh, current quote before it can be compared with the rest of the basket; the available snapshot was stale.^[1]

The balanced conclusion is that resilient demand can still matter more than the macro tape for selected companies, but the market is demanding proof. Software currently has the clearer proof points; home-related names need broader evidence that demand is durable rather than concentrated in a few strong operators. This is research, not investment advice.