Oil, rates and AI financing are separating the demand story

The market is not rejecting the idea that earnings growth and resilient demand can support this scope. It is applying a higher hurdle: growth must arrive with funding capacity, pricing power and enough balance-sheet resilience to absorb a geopolitical shock.

The market tell: software is being rewarded, infrastructure is being financed

At 12:28 p.m. ET on October 21, DDOG was up 7.4% to $275.46 and SNOW was up 1.9% to $325.53, while CRWV was down 1.7% to $61.08. RH was up 5.3% to $126.10; AN was down 3.5% to $156.25, WSM was little changed at $252.87, and LZB was down 1.7% to $28.68. ETH was up 3.0% to $23.92. These are delayed FMP regular-session snapshots, not end-of-day closes.^[1]

That dispersion is more informative than the direction of the group as a whole. The market is willing to pay for evidence that AI and digital workloads translate into software consumption. It is less willing to extend the same benefit to a capital-intensive infrastructure model whose returns depend on debt markets staying open and reasonably priced.

Snowflake’s latest reported quarter offers the cleanest fundamental support in the software part of the thesis: product revenue was $1.16 billion, up 29% year over year; net revenue retention was 125%; and remaining performance obligations rose 37% to $7.88 billion.^[2] Datadog’s recent tape has also responded to a reported upgrade tied to AI demand, although the next test is whether usage converts into durable revenue and margins rather than only a stronger narrative.^[2]

Why geopolitics matters even when oil flows continue

The current geopolitical channel is running through shipping economics rather than an immediate disappearance of supply. Reports on October 21 described elevated risk around the Strait of Hormuz and crude near multi-month highs; one market report put Brent near $110 a barrel and WTI near $99.^[3]

The International Energy Agency’s current commentary is explicitly about shipping disruptions lifting the price consumers pay for oil, but its page was behind a security-verification screen in this research pass, so the article relies on the accessible market reporting rather than attributing additional details to the IEA.^[4]

This is a tax on the thesis, not yet a verdict against it. Higher freight, fuel and insurance costs can pressure household budgets and furniture logistics, which matters for AN, RH, WSM and LZB. They can also keep inflation expectations and long-duration discount rates elevated, raising the bar for richly valued software and AI infrastructure. The relevant question is whether the disruption remains a cost shock or becomes a sustained supply interruption.

Rates are the second transmission channel

The latest macro snapshot shows a mixed but uncomfortable backdrop: unemployment at 4.2%, CPI inflation at 3.72% year over year, the federal-funds rate at 3.75%, the 10-year Treasury at 5.15%, VIX at 18.44, and high-yield spreads at 3.31%. Consumer sentiment was 51.7, down 11.17% year over year, while real GDP growth was 2.1% year over year.^[5]

That combination is not a classic recession signal. It is closer to a selective-financing regime: economic activity is still positive, but capital is expensive and consumers are not uniformly confident. In that regime, recurring software revenue can remain resilient, while companies that must continuously refinance large build-outs face a separate test.

CoreWeave is the clearest example. CNBC reported that its shares fell 15.5% during the October 14 stress episode while yields on bonds due in 2031 and 2032 approached 15%. The company said its backlog supports its debt obligations, but outside analysts highlighted the mismatch between hyperscaler-like capital expenditure and a balance sheet without hyperscaler capacity.^[6]

This does not prove that AI demand is false. It shows that demand, financing and equity returns are three different claims. A customer can want more compute while the infrastructure provider’s cost of capital rises enough to change who captures the economics.

The consumer names provide a useful cross-check

The home-furnishings part of the scope is not uniformly weak. Williams-Sonoma reported second-quarter comparable-brand revenue growth of 6.2%, a 22.9% GAAP operating margin and raised its full-year 2026 outlook.^[7] RH’s reported second-quarter revenue was about $922 million, but the earnings beat included a sizable tariff refund, making the quality of the margin signal more important than the headline EPS surprise.^[7]

That distinction matters in a geopolitical story. A one-time policy or tariff benefit can cushion a quarter; it does not establish that consumers will absorb higher freight, energy or financing costs indefinitely. The market’s mixed response across AN, RH, WSM and LZB is consistent with investors separating brand strength from broad housing and household-demand exposure—not with a single sector-wide conclusion.

What to watch next

1. Funding stress versus workload growth. For CRWV and the wider AI build-out, watch bond yields, refinancing terms, backlog conversion and operating cash flow together. A strong backlog is not the same as low-cost funding.
2. Software durability. DDOG and SNOW need continued evidence that AI-related usage expands paid workloads, retention and margins. DDOG’s next scheduled report is November 4, estimated, before the open; SNOW’s is December 2, estimated, after the close.^[8]
3. The oil-shipping path. A contained risk premium would be manageable; a prolonged disruption through Hormuz would feed more directly into freight, inflation and consumer demand. Follow tanker rates, insurance costs and inventory data—not only the headline Brent price.
4. Whether the consumer can absorb the shock. AN is scheduled to report October 30, estimated, before the open; RH December 10, estimated, after the close; WSM November 18, estimated, before the open; and LZB November 17, estimated, after the close.^[8]

The base case is therefore conditional rather than binary: resilient demand can support the software and selected consumer names, but only if geopolitical costs remain contained and financing markets do not convert AI enthusiasm into a balance-sheet event. The evidence currently supports a split market—not a blanket endorsement or rejection of the thesis.