AI demand is holding up; the cost of proving it is rising

The market is separating two claims that were often treated as one: that customers still want AI and cloud capacity, and that every dollar spent building that capacity will earn an attractive return. The first claim still has evidence behind it. The second is now being tested by energy disruption, financing needs and the speed at which infrastructure supply can be absorbed.

That distinction matters for the research scope. DDOG and SNOW are operating software businesses that can benefit as workloads expand. CRWV is closer to the capital-and-power bottleneck itself. AN, RH, WSM, ETH and LZB sit on the other side of the cycle, where tariffs, housing turnover and discretionary demand determine whether revenue growth converts into margin.

The market tell: growth is being rewarded selectively

At the October 19 regular close, DDOG gained 5.36% to $260.36 and SNOW gained 2.77% to $315.35. CRWV finished at $61.12 and was trading at $60.98 at 16:27 ET, 0.23% below its 16:00 ET close. LZB fell 2.73% to $29.25, while AN declined 0.83% to $160.39 and WSM slipped 0.29% to $249.43. RH rose 0.92% to $114.03. These are closing and, where available, extended-session observations—not a single sector-wide signal.^[1]

The dispersion is the message. Software exposure is being treated as a beneficiary of ongoing usage, while infrastructure financing and housing-linked consumption remain harder to underwrite. That is consistent with a market that wants evidence of durable demand, but is less willing to pay for growth whose costs arrive first.

What management evidence supports the AI side

Datadog’s recent transcript history provides the clearest evidence in this group. Management said in the second quarter of fiscal 2026 that more than 750 AI customers were using the platform, that all ten of the largest AI leaders were customers, and that MCP tool calls had grown more than 22 times from Q4 2025. Management also described accelerating infrastructure-product consumption across the customer base.^[2]

Earlier calls add useful context rather than proof of a straight line. Datadog said its infrastructure-monitoring business had passed $1.6 billion in ARR, log management had passed $1 billion and its APM and digital-experience suite had also crossed $1 billion. The company’s argument is that AI expands the number and complexity of monitored workloads, creating a recurring software layer above the compute build-out.^[2]

That supports the thesis for DDOG and, by extension, the broader cloud-software case around SNOW: resilient usage can translate into expansion if customers keep deploying workloads. But the evidence is not the same as a guarantee of valuation support. The next test is whether usage growth shows up alongside durable customer commitments, efficient sales execution and margins that do not require a permanently easy financing environment.

Why CRWV carries a different risk

CoreWeave is exposed to the same demand wave, but its economics sit nearer the physical constraint. Current reporting described a sharp prior-week selloff and continuing debate over AI infrastructure financing; one analyst note cited in the October 19 market coverage argued that customers still want compute and that compute remains scarce.^[3]

That is the bullish case. The counter-case is that scarcity today can encourage too much capacity, too much leverage or too much dependence on a small number of large customers. The market does not need to decide that demand is fake for CRWV to be volatile; it only needs to question the timing between capital spending, power availability, customer utilization and cash generation.

Geopolitics makes that timing less comfortable. Recent market reporting described continued attacks affecting shipping and energy infrastructure, emergency reserve releases and oil prices around $105 per barrel earlier in October. A separate report said the United States was considering easing sanctions on vessels that transported sanctioned oil, while analysts expected only a limited improvement in tanker availability.^[4] Energy and shipping shocks therefore matter to AI infrastructure not only through electricity prices, but through equipment logistics, project schedules and the discount rate applied to long-dated cash flows.

Furniture is the inverse test: demand can survive while margins do not

The furniture names show why “resilient demand” is not enough. RH’s management said revenue rose 9% in the third quarter of fiscal 2025 despite a very weak housing market and tariffs, but also said adjusted operating margin was below guidance because tariff expense and tariff-related opening costs were higher than expected.^[5]

The subsequent operating discussion was even more specific: furniture, outdoor furniture, lighting and rugs were difficult to resource quickly because factories and supplier networks take time to move. Management said the company expected margin growth as investment peaked, but that transition is a process, not an instant reset.^[5]

That is the relevant read-through for RH, WSM, ETH and LZB. The group can benefit if housing activity improves and consumers regain confidence, but tariff exposure, imported inputs, freight and inventory decisions can absorb the benefit before it reaches operating profit. Recent industry reporting also counted twelve bankruptcy, restructuring or insolvency proceedings among home-furnishings businesses in the first half of 2026, a sign that the pressure is broader than one company’s execution.^[6]

The thesis, tested

Evidence for: AI-related consumption is expanding in Datadog’s customer base; management continues to describe cloud migration and AI adoption as secular drivers; and some discretionary brands are still gaining share or holding demand despite a weak housing backdrop.^[2][5]

Evidence against: AI infrastructure requires large upfront commitments and remains sensitive to power, shipping and financing conditions. In furniture, tariff and sourcing costs can pressure margins even when revenue grows. The October 19 tape—stronger DDOG and SNOW, weaker CRWV and LZB—fits that split, although one session cannot establish a durable trend.^[1]

The balanced conclusion is that earnings growth and resilient demand can support parts of this scope over the next year, but the path is conditional. Software must show that AI usage becomes recurring, monetized consumption. Infrastructure must show that demand can cover the capital structure. Furniture must show that pricing and sourcing can restore margins without destroying volume.

What to watch next

1. DDOG and SNOW: customer consumption, expansion rates, remaining performance obligations and commentary on AI workloads versus experimentation.
2. CRWV: contracted backlog quality, customer concentration, power availability, financing costs, capex intensity and the pace at which new capacity becomes productive.
3. RH, WSM, ETH and LZB: housing turnover, ticket size, traffic, inventory, tariff pass-through and gross-margin recovery.
4. Energy and shipping: oil and refined-product prices, attacks or disruptions around major routes, tanker availability and any changes to sanctions enforcement. These can alter both inflation expectations and the cost of building physical infrastructure.^[4]
5. The next reporting window: the market-data calendar currently lists DDOG for November 4 before the open, SNOW for December 2 after the close, AN for October 30 before the open, RH for December 10 after the close, WSM for November 18 before the open and LZB for November 17 after the close; each date is marked estimated, not confirmed. CRWV is unsupported in that calendar.^[7]

This is a research framework, not a trade signal. The central question is not whether demand exists. It is whether demand remains strong enough, and economically valuable enough, to outrun energy, tariff, financing and execution costs.