Amphenol Corporation APH

Key data

The business

Amphenol makes the unglamorous physical connections inside almost every electronic system: connectors, cables, antennas and sensors that carry power and data between components. It sells through three segments. Harsh Environment Solutions serves military, aerospace, industrial and automotive customers with rugged connectors designed into equipment meant to last decades. Interconnect and Sensor Systems covers industrial, auto and sensing. Communications Solutions is the one that matters most right now: the high-speed copper and fiber interconnect that links chips, servers and switches inside data centers, plus mobile-network and broadband gear. In January the company closed its purchase of CommScope's connectivity and cable business, folding a large fiber and data-center supplier straight into that segment. Amphenol runs as roughly a hundred separate units under a lean center and buys companies constantly. The durable edge is being designed into a customer's product, where the part is a rounding error on cost but ripping it out means requalifying the whole system.

The numbers

The last five quarters show a business that shifted gears.

Revenue rose 55% year over year in the June quarter to $8.76B and net income 62% to $1.77B, the largest quarter Amphenol has ever reported. The one wrinkle is Q1 2026, where profit fell to $0.93B even as revenue jumped: the CommScope deal closed early in January and carried acquisition and integration costs that landed in that quarter, and Q2 snapped back once they cleared. Sequential growth of 15% from Q1 to Q2 says the demand is real, not a one-quarter spike.

Across the last four full years revenue compounded about 21% a year and net income about 28%, then both accelerated in 2025 as AI data-center orders and the acquisition stacked on top of each other. The two-for-one split distributed on September 2 exactly as declared, so the earnings figures above are as filed on the old share count, while the $82.78 price and the multiples are on the new one; EPS across the full year window also spans an earlier 2021 split and is not comparable end to end, though the dollar lines are. What I believe the price only half-credits: much of this growth is content, not just spending, because each new generation of AI rack needs more high-speed interconnect than the last, so Amphenol's dollars per server climb faster than the number of servers. If that holds, 26 times next year's earnings is not expensive; the print that settles it is Communications Solutions organic growth over the next two quarters.

Management

CEO Richard Norwitt has run Amphenol since 2009 and compounded it well, but insiders have been heavy sellers into this strength: about $411M sold over the last twelve months across nineteen sales against a single small purchase, including Norwitt's own $80.8M sale on July 31, the day the quarter was filed, and $75.9M in February, with a senior executive selling $36.9M in November. Plan status is not disclosed, so I can't separate scheduled sales from discretionary ones, but the scale and the timing next to record results are worth holding in view. Capital allocation is otherwise steady and shareholder-friendly: about $665M of stock bought back last year, a quarterly dividend of $0.25 per pre-split share, returns on equity near 37% and on invested capital around 15%. The company's whole history is buying businesses and earning good returns on them; CommScope, its largest deal in years, is the real test of whether that record extends to a purchase made near the top of a cycle.

How it fails or surprises you

AI capex digestion (downside). Communications Solutions is now the growth engine, and much of its demand traces to a handful of hyperscalers building AI capacity. If they pause to digest, the fastest-growing, highest-incremental-margin revenue reverses first. Watch Communications Solutions sequential revenue and any book-to-bill commentary in the Q3 print; a flat-to-down sequential quarter there is the first crack.

The Q1 wrinkle was demand, not deal costs. My read blames Q1's profit dip on CommScope closing charges. If the next quarter shows soft margins again with no deal to explain them, that read is wrong and the integration is dragging. Watch Communications Solutions margin over the next two quarters.

Content per rack keeps climbing (right tail). If each GPU rack generation keeps needing more copper and fiber than the last, Amphenol grows well faster than data-center unit counts and datacom stays above 40% organic into 2027. The market is paying for a capex cycle, not a content ramp; the tell is organic Communications Solutions growth holding above 40% in Q3 and Q4.

Closing thoughts

The evidence points to a genuinely great business at a price that already assumes the good case continues. This is mostly a distribution the market prices well: everyone can see the AI data-center build and Amphenol's place in it, so the edge is not in direction but in duration. The honest split is between two worlds one number tends to resolve, organic growth in the datacom segment: if it is content-led it stays high for years and the multiple is fair, if it is a capex spike it fades and 40 times trailing earnings compresses toward the low-20s the stock has usually paid. The fatter tail, on the current run of orders and the content argument, is the upside one, but the left tail is a fast one, because the same hyperscaler concentration that drives the surge drives the reversal, and the CEO selling $80M into it is not nothing.

The bet is still that the companies building AI data centers keep spending more and keep buying Amphenol's cables and connectors, the physical parts that link chips, servers and switches together. What breaks it is that spending stalling, or the content-per-rack story turning out to be an ordinary capex cycle; the one pair of numbers that tells you first is Communications Solutions organic growth and its sequential revenue in the Q3 report. If that segment prints another quarter above 40% organic, the price is earned; if it rolls over toward the mid-teens, you were paying top-of-range for a cyclical peak.