AI demand meets an energy shock

The market is not treating this geopolitical episode as a uniform risk-off event. It is sorting companies by how directly they depend on long-duration financing, imported goods, discretionary demand and the physical buildout of AI.

At 12:27 ET on October 28, the requested basket was split: Snowflake (SNOW) was up 1.1%, CoreWeave (CRWV) 1.2% and Restoration Hardware (RH) 0.7%, while Datadog (DDOG) was down 1.5%, AutoNation (AN) 1.1%, Williams-Sonoma (WSM) 0.3%, Ethan Allen Interiors (ETH) 0.6% and La-Z-Boy (LZB) 1.1%. These are delayed FMP snapshots by 15 minutes, not a real-time trading feed.^[1]

That dispersion is the useful tell. Investors appear to be distinguishing between demand tied to AI infrastructure and businesses exposed to household budgets, imported goods or financing costs—while still questioning whether AI spending can earn an adequate return.

The geopolitical shock is arriving through energy and inflation

The World Bank’s October 28 Commodity Markets Outlook says global commodity prices are projected to rise 23% in 2026, with energy prices up 32%. It attributes the pressure to the Middle East conflict, depleted oil inventories and disrupted grain shipments; its baseline assumes no major further escalation. Brent is forecast to average $91 a barrel in 2026, versus $69 in 2025, before easing to $79 in 2027.^[2]

That baseline matters because it defines the difference between a shock that fades and one that becomes a rates regime. The same World Bank release says global oil supply is projected to fall 5.2% in 2026, while consumption falls 2.5%, leaving production below consumption and inventories more exposed to renewed disruption.^[2]

The San Francisco Fed’s October assessment reaches a similar conclusion from a monetary-policy angle: headline PCE inflation was 3.4% year over year in August, with energy and core goods linked to the Middle East conflict and higher tariffs making the larger contribution. Its baseline expects inflation to decline toward 2% over the next two years, but the alternative scenario is a longer conflict and a higher inflation path.^[3]

For markets, the key risk is not simply higher oil. It is the possibility that higher energy costs delay disinflation, keep policy restrictive for longer and compress the value investors assign to distant cash flows.

AI infrastructure still has real macro support—but financing is the test

The bullish case has tangible evidence. The San Francisco Fed says real investment in information-processing equipment has grown at an average annualized rate above 20% since the start of 2025, exceeding the pace seen after the Great Recession and during the late-1990s internet and computer boom.^[3]

That backdrop helps explain why CRWV and SNOW were holding up in the midday snapshot. CoreWeave is more directly exposed to the buildout of compute capacity, while Snowflake is a software beneficiary of data-intensive workloads. Datadog’s weaker print is a reminder that “AI exposure” is not a single factor: the market can reward infrastructure scarcity while still repricing software duration, valuation or customer-budget risk.

But the counterargument is equally current. NBC reported that the October 22 selloff combined Iran-war concerns, higher oil and fears that capital financing the AI boom was becoming scarce; Brent rose nearly 4% that day to $113 a barrel, while the Nasdaq fell 2.6%. The report also cited Nvidia CEO Jensen Huang saying financing for AI investment was becoming scarce amid higher rates and existing spending commitments.^[4]

This is the central question for the next year: can revenue growth and resilient demand compound fast enough to justify the capital intensity before energy costs and rates raise the hurdle rate? The answer is likely to differ across DDOG, SNOW and CRWV even if all three are placed under the broad “AI” label.

Export controls add a second geopolitical channel

China-related technology restrictions add policy uncertainty on top of physical and financial risk. Current reporting describes proposed Chinese limits on some technology exports, including AI-related technologies, framed around national-security concerns.^[5]

A separate report on October 28 said the United States was considering additional restrictions tied to Chinese access to remote AI compute and advanced technology. That report could not be independently read here because the hosting page blocked automated access, so it should be treated as a reported policy signal rather than a verified rule or implementation date.^[6]

For CRWV and the broader infrastructure chain, the issue is not only whether demand exists. It is whether hardware, customers and compute can move across borders under changing licensing rules. For SNOW and DDOG, the direct exposure is less physical, but customer budgets and international growth assumptions can still be affected by a slower or more fragmented technology cycle.

Consumer and housing-sensitive names show the other side of the split

The consumer portion of the basket is a useful counterweight. AN, WSM, ETH and LZB were lower in the midday snapshot, while RH was modestly higher. That is not enough to establish a causal pattern from one session, but it is consistent with a market that is asking whether households can absorb higher goods, energy and financing costs.

Trade policy is relevant here. The New York Fed has reported that tariffs accounted for a substantial increase in the prices of everyday goods, while housing and furniture-related industries remain exposed to imported materials and components.^[7] The distinction among retailers will therefore be execution-specific: pricing power, sourcing flexibility, inventory discipline and the income profile of the customer matter more than a simple “consumer” label.

The practical read-through for RH, WSM, ETH and LZB is not that demand must collapse. It is that a resilient consumer may need to coexist with higher input costs and a higher-for-longer rate path. AN adds a separate sensitivity to vehicle affordability and financing conditions.

What would confirm or challenge the thesis?

Evidence that would confirm the pressure channel:

• Further escalation that keeps oil inventories under strain rather than allowing prices to normalize.
• A sustained rise in goods and energy inflation that pushes expected policy rates higher.
• New export-control rules that limit access to advanced compute or complicate cross-border AI deployment.
• Management commentary showing that AI infrastructure commitments are rising faster than cash generation or customer revenue.
• More visible pressure on discretionary traffic, housing-linked demand or promotional intensity at consumer-facing companies.

Evidence that would challenge it:

• A durable de-escalation that lowers energy risk and restores the disinflation path.
• AI revenue growth and utilization that validate infrastructure spending despite higher financing costs.
• Stable enterprise software budgets, strong renewal activity and improving free-cash-flow conversion.
• Retailers passing through costs without a meaningful demand response, supported by healthy inventories and traffic.

The current data do not justify a single-direction conclusion. The market is separating the basket because geopolitics is changing the cost of capital and physical inputs at the same time that AI is expanding investment demand. The next signal is whether that divergence survives the next round of company guidance and policy decisions—not whether one midday quote points up or down.