AI Capex in Mid-Cycle Boom, Says Morgan Stanley’s Shalett

Morgan Stanley Wealth Management CIO Lisa Shalett expects the growth narrative for technology earnings to persist. She cautioned that shifting bottlenecks in AI capital spending could raise questions for investors; the report provided no figures or specific company impacts.

Analysis

The investable signal is not simply that AI earnings growth persists; it is that the location of the constraint may determine who captures the next dollar of capex. If spending shifts from compute toward power, grid equipment, cooling, or networking, earnings leadership could broaden beyond chip suppliers—but only where orders convert into revenue and pricing power, rather than merely longer backlogs. A bottleneck can also delay deployment and reduce near-term returns on already-installed accelerators, challenging the assumption that strong capex automatically translates into durable earnings.

Over days, this is more a narrative and positioning risk than a standalone fundamental catalyst. Over 1–3 months, track hyperscaler capex commentary, delivery lead times, and evidence that infrastructure orders are accelerating without project delays. Over 6–18 months, power availability, utilization, and returns on AI investment should determine whether the broader supply chain sustains premium valuations. The contrarian risk is that investors may extrapolate spending growth while underweighting capacity constraints and the possibility that capex shifts are a symptom of delayed deployments, not incremental demand. No company-specific financial impact is established by this commentary.

Key Decisions for Investors

• Avoid treating the broad AI earnings narrative as a fresh buy signal by itself; there is no company-specific estimate change or quantified catalyst here.
• Set a 1–3 month watch on hyperscaler capex mix, grid-connection timelines, and power/cooling equipment order conversion. Favor infrastructure exposure only if reported demand converts into revenue and guidance, not backlog alone.
• If infrastructure orders strengthen while accelerator delivery or deployment timelines worsen, consider a relative-value tilt from concentrated AI compute exposure toward diversified electrical-equipment and power-infrastructure exposure; size it as a conditional trade, not a confirmed earnings divergence.
• Falsify the rotation thesis if hyperscalers indicate capex remains concentrated in compute, infrastructure suppliers fail to convert orders into sales, or deployment delays and weaker AI utilization undermine spending plans.