Buy 3 AI-Led Stocks Amid Solid Estimate Revisions and Upside Potential

Zacks highlights Quanta Services, Keysight Technologies and Amphenol as AI-infrastructure beneficiaries with positive 30-day earnings-estimate revisions and implied brokerage-target upside of 25.3%, 17.7% and 21.4%, respectively. Quanta's backlog reached a record $53.4 billion as of June 30, up 49% year over year, while Keysight's fiscal Q3 orders rose 56% to $2.091 billion and Amphenol's IT datacom sales grew 63% organically in Q2. The bullish outlook is supported by AI data-center power, testing and interconnect demand despite sticky inflation, elevated rates and oil-price risks tied to Middle East conflicts.

Analysis

The investable signal is not the estimate revisions themselves but where AI capex is bottlenecked. PWR monetizes the physical power-delivery constraint before servers can be energized; APH captures rising interconnect content per rack; KEYS is more exposed to the design/validation cycle and therefore offers the highest operating leverage if advanced-node, optical and custom-silicon programs continue accelerating. This creates a useful sequencing trade: PWR backlog conversion should support 6-18 month visibility, while KEYS and APH remain more sensitive to quarterly hyperscaler procurement timing.

The principal second-order risk is that elevated power and financing costs delay data-center energization rather than cancel silicon demand. That would defer PWR revenue recognition and potentially shift spending toward liquid cooling, power-management and optical architectures, benefiting Vertiv (VRT), Eaton (ETN), Credo (CRDO) and Coherent (COHR). APH is relatively insulated by defense/aerospace and industrial diversification, whereas KEYS has greater downside if semiconductor customers pause R&D tools after a capacity buildout.

Consensus likely overweights published price targets and underweights valuation risk after AI-infrastructure rerating. For the next 1-3 months, the key catalyst is evidence that orders remain firm through budget season; over 6-18 months, grid interconnection queues, transformer availability and utility capital plans determine whether announced data-center demand becomes deployed load. A widening gap between PWR bookings and revenue conversion, or decelerating APH IT-datacom organic growth, would be the first sign that the physical build cycle is slipping.

Inflation is not uniformly negative here: it can increase nominal utility and infrastructure project values, but persistent higher rates pressure long-duration multiples and customer project IRRs. Favor companies with pricing power and contracted demand rather than treating the group as a single AI beta basket; QBTS is not a comparable beneficiary of this capex chain and should not be used as a proxy hedge.

Key Decisions for Investors

• Overweight APH versus KEYS over the next 3-6 months: long APH / short KEYS in equal dollar amounts. APH has diversified end-market support and direct content-per-rack exposure; the spread should work if semiconductor test orders normalize. Exit if APH IT-datacom organic growth drops below 30% or KEYS orders sustain greater than 25% sequential growth for two quarters.
• Accumulate PWR only on market-driven pullbacks, not on backlog headlines; target a 6-18 month holding period. The thesis requires continued conversion of electric-infrastructure backlog into revenue and stable project margins. Reduce if backlog growth materially outpaces revenue for two reporting periods or if major utilities reduce capex plans.
• Use a basket long of PWR, APH and KEYS against a modest short in an AI-capex beta proxy such as SMH for 1-3 months only if earnings revisions continue rising. This isolates power, interconnect and test bottlenecks from broad semiconductor multiple compression; stop out if SMH outperforms the basket by 10% after the next earnings cycle.
• Monitor ETN, VRT, CRDO and COHR as confirmation names rather than chase targets. Broad strength in power distribution, cooling and optical interconnect suppliers would validate deployment; isolated strength in PWR/APH/KEYS without these peers raises the probability that the move is analyst-driven rather than order-driven.