Stifel: US non-residential construction spending up 0.1% in March

U.S. non-residential construction spending rose 0.1% year-over-year in March, with non-building spending up 5.1% and building spending down 3.0%. Data center spending surged 33.7% on hyperscaler AI capex, while manufacturing spending fell 17.4% and office spending dropped 7.9%. The article also notes oil prices ticked up after new attacks on ships in and around the Strait of Hormuz.

Analysis

The more important signal is not the headline energy noise, but the divergence in capex intensity by end market. Data-center construction is still behaving like a quasi-utility buildout, which should keep a high floor under electrical gear, switchgear, thermal management, and grid-adjacent materials even if broader nonresidential activity stays flat. That supports a narrower but more durable set of winners than the usual “industrial recovery” basket, with hyperscaler spend likely to remain the cleanest growth vector through 2026.

The weak spots matter because they point to a lagged overhang for several channels: manufacturing and office softness usually feed into lower backlog conversion for general contractors, steel fabricators, and commodity-linked materials with 2-4 quarter latency. The warehouse stabilization theme is more interesting as a second-order call on 2026 than as a near-term catalyst; if logistics demand improves, it can leverage incremental pricing power in insulation, roofing, concrete, and automation suppliers far more than headline construction spending would suggest.

On the energy side, Strait of Hormuz risk is a volatility event first, supply shock second. The market is likely to price in a risk premium quickly, but unless disruptions persist beyond days to weeks, the cleaner trade is through options rather than directional equity beta. The contrarian risk is that any meaningful oil spike can pressure power-intensive AI capex economics, but hyperscalers can absorb modest cost inflation better than industrial end users, so the real transmission channel is into margins for transport, chemicals, and construction rather than into data-center demand itself.

Key Decisions for Investors

• Long AMAT / ON / ETN on a 3-6 month horizon: data-center capex remains the most resilient construction subsegment; use pullbacks to add, targeting 12-18% upside with 8-10% downside if AI spend decelerates.
• Pair long ETN vs short XLI basket for 1-2 quarters: favor electrical infrastructure exposure over broad industrial cyclicals; expected relative outperformance if backlog mix continues to shift toward data centers and grid spending.
• Buy XLE call spreads or USO calls for 1-4 weeks into geopolitics headlines: prefer convexity over spot exposure because the base case is a fast fade unless physical flows are impaired; risk/reward is favorable with defined premium at risk.
• Avoid or underweight highly levered construction/material names tied to manufacturing and office exposure for the next 2-3 quarters; backlog revisions could hit harder than current consensus implies if comparisons stay weak.
• Watch CCI/AMT and AI-infrastructure beneficiaries on any oil-driven equity selloff: if higher crude tightens financial conditions, REITs with data-center adjacency may offer a better risk-adjusted entry than broad industrials.