The Data Center Build-Out Is Accelerating. Here Are 3 Stocks to Buy Beyond Chips.

The article highlights accelerating AI infrastructure spending as a favorable tailwind for Astera Labs, CoreWeave, and Vertiv. CoreWeave said revenue more than doubled to almost $2.1 billion in Q1 and it has a $100 billion backlog, while Vertiv reported 30% Q1 revenue growth and raised full-year guidance to $13.8 billion. Astera Labs also nearly doubled revenue to $308 million in Q1, underscoring strong demand for AI networking hardware and data center power/cooling systems.

Analysis

The cleanest read-through is that AI capex is broadening from compute into the picks-and-shovels bottlenecks that actually determine deployment speed: interconnect, power, and thermal management. That matters because these segments have higher switching costs and less customer concentration than the headline GPU vendors, so they can keep compounding even if hyperscaler procurement gets choppy. In practice, this is a second-derivative trade on AI intensity: the more chips per rack and the more heterogeneous the rack architecture, the more incremental value accrues to networking and infrastructure control points.

The market is likely underestimating how binding electricity and cooling constraints become once utilization rates rise. That creates a favorable setup for Vertiv-type exposure because the spend is not discretionary software budget; it is physical capacity that must be installed before revenue can be recognized. The implication is that margins can expand faster than consensus if service attach and installed-base revenue grow, while vendors with deep integration may gain pricing power as operators prioritize uptime over unit cost.

CoreWeave is the highest-beta expression, but also the most fragile: backlog visibility is helpful, yet customer concentration means sentiment can turn abruptly if one or two frontier-model spenders pause. The more interesting angle is that its backlog and rapid facility conversion point to a supply-chain arbitrage: whoever can source power, land, and equipment fastest wins share. That should keep pressure on smaller GPU-cloud competitors with weaker balance sheets and slower site activation, even if the category remains crowded.

The contrarian view is that the easy money in AI infrastructure may be behind us for the most crowded names, and the trade is shifting from multiple expansion to execution quality. If AI capex decelerates even modestly, high-multiple hardware enablers can de-rate faster than earnings grow. The setup argues for favoring companies with recurring service revenue, embedded software, and operating leverage over pure narrative beneficiaries.

Key Decisions for Investors

• Go long VRT on pullbacks over the next 2-6 weeks; use a 12-18 month horizon and look for a 2-3x upside/downside skew if data-center power constraints keep tightening.
• Buy ALAB only on weakness or via a call spread rather than outright stock; 6-12 month view, because the multiple already discounts strong execution and any capex pause could compress valuation sharply.
• Initiate a smaller, high-conviction long CRWV position paired against a basket of weaker GPU-cloud or unprofitable AI infra names; 3-9 month horizon, betting the best-capitalized operator wins the activation-speed race.
• For a cleaner risk-defined expression, buy VRT Jan-2027 call spreads and finance them with a partial short in a high-multiple semi proxy; this captures continued AI power spend while limiting multiple-risk.
• Avoid chasing NVDA on this headline; the better risk/reward is in second-order beneficiaries where revenue is more directly tied to deployment bottlenecks and less dependent on model-cycle timing.