Analog Devices, Inc. ADI

Key data

The business

Analog Devices makes the chips that sit between the physical world and the digital one: data converters, amplifiers, power management, signal conditioning. A sensor reads a temperature or a current; an ADI part turns it into a clean digital number a processor can use, and back again. The moat is in the count and the life of the parts, tens of thousands of them, most costing a few dollars, each designed onto a customer's board and left there for a decade because swapping it means redesigning the whole thing. It sells through distributors and direct to equipment makers. The end markets, in size order, are industrial, automotive, communications including data center, and consumer. That single converter in a factory controller or a cell-tower radio is cheap to buy and expensive to replace, which is the entire business in one part.

The numbers

Earnings have inflected off a cyclical trough, hard.

Revenue rose 40% year over year in the August quarter, but net income rose 159% and EPS 164%, because operating income nearly doubled to $1.61B: the classic operating leverage of a fabless-heavy analog maker coming off the bottom. Non-GAAP EPS of $3.45 beat consensus for the fourth straight quarter. The sequence is still accelerating, not flattening.

The five-year record is the whole argument. FY24 earnings halved from the FY23 peak, $6.55 to $3.28, as industrial and auto customers worked off inventory. Nine months into FY26, ADI has already earned $6.83, past the old peak, and the August quarter annualizes near $16B of revenue. The multiple, 43x trailing, sits at the 92nd percentile of its twelve-year range, which is the market saying this step-up is durable. What that price does not settle is the one fact this run explains least well: the same customers who drove the doubling drove the halving two years ago, and gross margin, back to 67.3%, held flat quarter to quarter, so the earnings from here have to come from volume, not more margin.

Management

Vincent Roche has run this company since 2013 and been inside it for four decades, so the record is his. Over the last year insiders bought nothing and sold 86 times for $20.6M, Roche himself selling $3.6M to $4.0M on the first of most months; the cadence looks like a program though plan status is not disclosed in the filings. Capital return is the real tell: ADI bought back $2.16B of stock in FY25 and $2.96B in FY23 near higher prices, which is aggressive buying across a cycle rather than only at the bottom. R&D runs at 14% of sales. Net debt is a manageable 1.1x EBITDA, interest covered 14x.

How it fails or surprises you

The cycle turns again. Industrial and automotive are the swing markets, and they just proved they can take earnings from $6.55 to $3.28 in two years. An inventory correction returns, orders soften, and a business bought at 43x re-rates on both lower earnings and a lower multiple. The first tell is sequential revenue rolling over and book-to-bill slipping below one.

AI data center scales (right tail). The fastest-growing sub-market is the data center, driven by AI power and connectivity, and today it is small enough that the market pays ADI a cyclical multiple, not a secular one. If that content grows from a rounding error to a reported driver, the story changes from "analog cycle" to "AI beneficiary." The print: communications and data-center revenue disclosed as a sustained accelerator.

You are paying peak for peak. At 43x, near the top of its historical range, with margins already recovered to 67% and little obvious room above, the earnings have to keep compounding on volume alone. Customer concentration and China exposure are not disclosed in this quarter's 10-Q, so the buyer is trusting a mix he cannot see. The tell is margin stalling while orders cool.

Closing thoughts

The order trend and FY27 guide tell you whether industrial demand holds, which settles the nearest question: is this recovery durable or another cyclical peak. But the underlying cycle, the one that halved earnings in 2024, will return at some point, you just don't know when, so entry price and balance-sheet strength matter as much as the near-term print. The left tail is fatter and recent: if industrial rolls over, you lose on both earnings and multiple compression at once, from an already-elevated starting price. The right tail, AI data center scaling into a re-rate, is real but too small in today's numbers to lean on. What is at risk is buying peak earnings at a peak multiple; what you are paid for is a deep catalog, 67% margins, and steady buybacks through the cycle.

The bet is still that factories, cars, and AI data centers keep buying more of these chips, and the slump that cut sales a quarter two years ago is over for good rather than paused. What breaks it is industrial orders rolling over while margins slip, and the one pair of numbers that tells you first is sequential revenue and gross margin, watched together. If revenue stalls and the 67% margin starts to give, the durable step-up was a cyclical peak, and 43x was the wrong price to pay for it.