AI Demand Is Holding; Power, Credit and Shipping Are the Market Test

The market’s current question is not whether companies want artificial intelligence. It is whether the power, equipment, financing and logistics needed to serve that demand can be delivered at returns that survive a more difficult macro backdrop.

That distinction matters for the growth basket tracked here: DDOG, SNOW, CRWV, AN, RH, WSM, ETH and LZB. The latest evidence does not support a clean “AI boom is over” conclusion. It does support a more selective one: demand can remain resilient while the cost of converting demand into cash flow rises.

The market tell: demand stocks are firmer than financing-sensitive infrastructure

At the October 27, 2026 regular close, CRWV rose 7.94% to $59.57, while DDOG gained 0.21% to $280.82, SNOW gained 0.74% to $337.42, and WSM gained 0.20% to $250.45. The after-hours tape was comparatively quiet: CRWV was $59.16 at 16:33 ET, down 0.69% from the regular close, while DDOG was $281.00 at 16:32 ET, up 0.06%. These are snapshots, not a causal explanation, but they show that the market is still willing to reward evidence of demand even as it remains sensitive to infrastructure risk.^[1]

The more important tell is in the dispersion. CoreWeave is the clearest expression of the “demand versus delivery” problem: its Q2 FY2026 transcript described near-term capacity as effectively sold out, revenue of $2.6 billion, and backlog of $104.2 billion, but the same business is scaling a power- and debt-heavy asset base.^[2] Datadog offers the asset-light counterpoint: in Q2 FY2026, management reported 36% year-over-year revenue growth and said non-AI customer revenue growth had accelerated to the high 20s.^[3]

Why geopolitics enters the earnings model

The geopolitical channel is running through energy, trade and transport rather than through a single headline. The San Francisco Fed’s October 15 assessment said AI infrastructure investment was boosting growth, while energy and core-goods inflation linked to the Middle East conflict and higher tariff rates remained important sources of price pressure. It also described a longer-conflict scenario in which inflation would follow a higher path than in the baseline.^[4]

That matters because a data center is not just a software story. It needs electricity, cooling, chips, construction, network equipment and reliable access to capital. A disruption in an energy corridor can raise operating costs indirectly through fuel, shipping and equipment prices. A tariff change can alter the cost of imported hardware. Higher rates can reduce the present value of long-duration growth and increase the burden of debt-funded capacity.

The shipping channel is similarly concrete. The International Energy Agency’s October commentary, as summarized in current search results, linked disruption in shipping to higher oil costs for consumers. S&P Global’s October commodity outlook also warned that geopolitical conflict and weaker Gulf shipping traffic could keep manufacturing costs elevated into 2027.^[5]

Three pieces of evidence for the resilient-demand case

1. Software usage is broadening beyond AI-native companies

Datadog’s recent management commentary is important because it does not rely only on a narrow group of AI startups. Management said both AI-native and non-AI customers were contributing to growth, with customers adopting AI, cloud and modern technologies and using the platform to observe, secure and act on those workloads. In Q2 FY2026, the company said non-AI revenue growth had reached the high 20s year over year, while total revenue reached $1.12 billion.^[3]

That is evidence in favor of the supplied thesis for DDOG, and potentially for SNOW as a data-platform read-through: enterprises may trim waste, but they are not necessarily abandoning the systems needed to operate increasingly complex digital workloads. The counterpoint is that Datadog has also discussed optimization and customers seeking efficiency, so usage growth does not automatically translate into unchanged pricing power.

2. AI infrastructure demand is real, but its economics are more exposed

CoreWeave’s Q2 FY2026 commentary described strong demand, favorable pricing for newer GPU systems and a backlog of $104.2 billion. Its Q1 FY2026 commentary, however, showed the other side of the model: interest expense was $536 million, up from $264 million a year earlier, as debt supported infrastructure expansion and customer commitments.^[2]

This is not a contradiction. It is the central risk. Contracted demand can make capacity easier to finance, but debt service, power availability, deployment timing and customer concentration still determine how much of that demand becomes durable equity value. Recent reporting has focused on AI financing worries, credit risk and the gap between announced data-center capacity and what can realistically be powered and built.

3. Consumer operators can still offset shocks, but not indefinitely

Williams-Sonoma provides a useful non-AI stress test. In its Q2 FY2026 call, management said the home-furnishings industry was essentially flat while the company’s ecommerce and retail comparable sales rose 6.5% and 5.5%, respectively. It also said tariffs had reduced merchandise margins by about 230 basis points, partly offset by supply-chain efficiencies and occupancy leverage.^[6]

The implication for AN, RH, WSM and LZB is two-sided. Resilient demand, market-share gains and operational discipline can absorb part of a policy shock. But fuel, ocean freight, tariffs and housing turnover remain external variables. Williams-Sonoma’s own Q1 FY2026 commentary explicitly assumed volatility in geopolitics, war, fuel prices, trade policy, tariffs and interest rates rather than assuming a clean macro recovery.^[6]

What would falsify the thesis?

The bullish version of the thesis requires three things to remain true: enterprise AI usage continues to expand; infrastructure contracts remain financeable on acceptable terms; and energy, shipping and tariff costs do not accelerate faster than companies can pass them through or offset them.

The bearish version would gain credibility if several signals arrived together: cloud optimization becomes a sustained contraction rather than a renewal-cycle adjustment; CoreWeave-like operators report delayed deployments, weaker utilization or deteriorating financing terms; software growth remains strong only because of price increases while usage slows; or consumer companies lose their ability to protect gross margins as fuel and tariff costs rise.

The current macro backdrop is not benign. The latest FRED snapshot shows 3.72% CPI inflation, a 5.00% 10-year Treasury yield, a 19.45 VIX and a 3.38% high-yield spread, alongside 2.1% real GDP growth and 4.2% unemployment.^[7] That combination is consistent with growth that has not broken, but with a higher discount rate and less tolerance for execution mistakes.

What to watch next

1. AI infrastructure financing: Watch interest expense, capital expenditure, backlog conversion and active-power deployment at CRWV. Backlog growth is not enough if delivery timing or financing costs deteriorate.
2. Usage versus optimization: For DDOG and SNOW, separate customer counts and AI adoption from usage-based expansion, renewal pricing and cloud-cost optimization.
3. Energy and shipping: Track oil, freight disruption and equipment lead times. These are the channels through which a geopolitical conflict can reach margins without directly affecting end demand.
4. Consumer margin defense: For AN, RH, WSM and LZB, watch full-price selling, freight, tariff mitigation, housing turnover and the ability to preserve gross margin without weakening demand.
5. Rates and credit: A sustained rise in long-term yields or high-yield spreads would matter most for the longest-duration and most capital-intensive names, even if operating demand remains firm.

The base case is therefore neither “AI demand is fake” nor “backlogs make risk irrelevant.” It is a sorting process. Demand evidence favors the software and infrastructure theme; geopolitical energy, shipping and policy shocks determine how much of that demand survives the cost of delivery. The next leg of the trade will be decided by conversion, financing and margins—not by demand headlines alone.

This article is for research and education, not investment advice. It does not recommend buying or selling any security.