Iran War Worries Hit Just About Everything but Data Centers

The important signal here is not the headline infrastructure spend, but the re-pricing of power as the binding constraint for AI scale. If large-model buildout increasingly requires behind-the-meter generation, the value chain shifts away from “pure compute” toward firms that can finance, permit, interconnect, and operate distributed energy assets. That structurally favors equipment vendors and integrated industrials with grid, turbine, and controls exposure, while cloud/platform players inherit more execution risk and longer payback periods.

For GEV, the second-order upside is that this is not a one-off order story; it validates a multi-year demand cycle for firm power solutions tied to data-center localization. The market may still be underestimating margin mix improvement if spare manufacturing capacity stays tight and service attach rates rise after installation. The key risk is that these projects are capex-heavy and politically exposed, so any change in permitting, gas pricing, or AI monetization could push procurement timing out by 2-4 quarters.

For ORCL, the issue is less direct revenue and more balance-sheet and duration risk: hyperscale/AI infrastructure commitments can inflate near-term capital intensity before usage ramps. Consensus may be too focused on top-line AI participation and too lax on the financing cadence needed to support it. If power costs or interconnection delays rise, the market could start discounting slower deployment rather than faster growth, which would matter most over the next 6-12 months.

Contrarian view: the bullish read on AI infrastructure may be overstating speed of conversion from announced projects to revenue-generating capacity. A meaningful share of value could leak to suppliers of gas turbines, switchgear, EPC services, and grid equipment rather than the nominal sponsors, especially if power becomes the scarce input. The cleaner trade is to own the bottleneck, not the narrative.