We toured an AI data center to see how our stock names make these facilities work

The article highlights strong AI-driven demand for data center infrastructure, with Big Tech on track to spend at least $608 billion this year and suppliers seeing meaningful benefit across power, chips, fiber, networking, and cooling. GE Vernova reported Q1 revenue up 16% year over year to $9.3 billion and added $13 billion to backlog, while Corning’s sales rose 18% to $4.35 billion and Eaton, Broadcom, Nvidia, and Dover were all cited as key AI beneficiaries. The piece is broadly constructive for the named suppliers and data center ecosystem, with several companies reporting robust orders and raised expectations.

Analysis

The market is still underestimating how AI capex is shifting from a compute story into a utilities-and-rights-of-way story. The binding constraint is no longer just GPU supply; it is interconnect, thermal management, and especially grid access, which creates a multi-year bottleneck that should keep pricing power elevated for power equipment, liquid cooling, and fiber suppliers even if hyperscaler capex growth moderates. That makes the “pick-and-shovel” beneficiaries more durable than the chip trade because they are exposed to installed-base density rather than model cycles.

Within the chain, Eaton and GE Vernova look best positioned on a 12-24 month horizon because the secular need is turning into backlog conversion, not just order growth. The second-order winner is Corning: fiber is becoming a harder requirement as data centers chase lower latency and lower heat, and that should support mix improvement and factory utilization even if the broader industrial cycle softens. Broadcom is more nuanced: custom silicon is a legitimate share-gain vector, but the real upside is increasingly tied to networking content, not just accelerators, which means the market may still be underpricing its role as a system-level enabler rather than a pure chip competitor to Nvidia.

The key risk is that the trade has become crowded in the obvious beneficiaries while the market may be extrapolating current build rates too linearly. If hyperscaler capex pauses after earnings or if power delivery delays stretch timelines, the most levered names can de-rate quickly even though the medium-term thesis remains intact. Over a 1-3 month horizon, watch for any sign that utility interconnection queues or permitting bottlenecks are forcing project slippage; that would hit near-term sentiment before it hits revenue, creating a better entry point.

Contrarianly, the market may be too focused on who sells the chips and not enough on who owns the bottlenecks: electrical gear, cooling, and fiber could sustain above-trend growth longer than semis because every incremental watt requires more supporting infrastructure. The other underappreciated angle is that custom silicon adoption does not necessarily weaken the ecosystem; it can expand total AI infrastructure spend by lowering inference cost and therefore increasing deployment density. That argues for staying overweight the enablers while being selective on the semiconductor pair.

Key Decisions for Investors

• Add on weakness to ETN for a 6-12 month hold; use any post-earnings dip as entry. Risk/reward favors continued backlog upside from AI power distribution, with downside limited unless data center capex rolls over materially.
• Stay long GEV on a 12-18 month horizon, but size with discipline after the run. This is the cleanest way to own grid-constrained AI buildout; trim only if order growth or backlog conversion decelerates two quarters in a row.
• Initiate/maintain a long GLW position on a 6-12 month basis. The thesis is factory utilization and mix expansion from fiber intensity; risk is valuation compression if hyperscaler builds pause, but the structural demand curve remains intact.
• Prefer AVGO over NVDA tactically for the next 3-6 months via a relative-value pair: long AVGO / short a smaller notional of NVDA. Rationale: AVGO has more diversified AI exposure through networking and custom silicon while Nvidia remains more exposed to sentiment swings around capex digestion.
• Use any sharp post-earnings rally in AMZN or MSFT to fade via call overwrites rather than outright shorts. Their data-center spend is a long-duration moat, but near-term the market may overcapitalize the rate of build, creating volatile but not thesis-breaking pullbacks.