Applied Materials, Inc. AMAT

Key data

The business

Applied makes the machines chipmakers use to build a transistor: laying down films, modifying and shaping them layer by layer, then the equipment to package finished chips together. It touches more of those process steps than any rival, which is the moat, once its tool is qualified into a customer's recipe on a leading-edge node, switching it out means requalifying the whole flow. A services arm (AGS) sells parts and upgrades into that installed base at a recurring 30% operating margin, $1.78B last quarter. A small display unit rounds it out. The concentration is real and disclosed: two customers were 20% and 14% of revenue over the first nine months, so a third of the business sits with two leading-edge buyers you are trusting to keep spending.

The numbers

The story of the last year is acceleration in the top line and margin, on slower-improving cash.

Revenue growth ran from -2% year-on-year in January to +11% in April to +25% in July, a genuine inflection in the July quarter, not a drift. The August quarter came in ahead of consensus on earnings, the fourth straight beat. The open question from two days ago, whether growth would halve or free cash flow would finally catch reported earnings, resolved only halfway: growth accelerated instead of halving, but cash conversion crept to just 64% of net income while receivables jumped to $7.7B from $5.2B and inventory built to $6.6B, both outrunning the 25% sales gain.

For four years this was a 5%-a-year revenue grower whose earnings per share climbed faster only because buybacks shrank the count, and it traded for 15 to 21 times earnings the entire time. Now the market pays 39 times, half again above its highest multiple in twelve years, on the belief that July's 25% is the new run rate. Consensus underwrites exactly that: FY2027 revenue of $46.7B and EPS of $18.67, more than double FY2025. What it does not yet underwrite is the cash. Buybacks fell to $337M in the January quarter from a $4.9B annual pace, redirected as capex rose 90% to $2.3B to build capacity, so the earnings growth is real but the free cash behind it lags and the working capital is absorbing the ramp. The variant view here: this is a cyclical toolmaker priced as a secular compounder, and the print that settles it is whether receivables and inventory keep growing faster than revenue into the October quarter.

Management

The record is capable operators selling their own stock and slowing the buyback while paying up for capacity. Insiders bought nothing over the last twelve months and sold about $180M across 43 transactions; CEO Gary Dickerson sold roughly $50M in two days at the end of June and President Prabu Raja about $13M in early June, with plan status not disclosed on the Form 4s, so I cannot tell you which of those were pre-scheduled. Capital allocation shifted this year from repurchase toward capex and R&D (13% of revenue), a defensible choice if the AI ramp is durable and a poorly-timed one if it is a peak. The company also paid a $253M settlement to the Commerce Department in the second quarter to close a China export-controls inquiry, a clean resolution but a reminder of where the regulatory risk sits.

How it fails or surprises you

Two customers, one-third of sales. The 20% and 14% concentration means a single leading-edge foundry or memory maker trimming its equipment budget takes the 25% growth with it. The next 10-K's concentration note and the semiconductor-systems revenue line are where you would see it turn first.

The growth is being financed (the number the read explains least). Receivables rose about 48% and inventory 11% against 25% revenue growth, and free cash flow is only 64% of net income. If AR keeps outrunning sales into Q4, the reported earnings are lending customers the growth, and the 39x multiple is paying for it.

A real WFE step-up (right tail). If gate-all-around, backside power and advanced packaging drive a durable multi-year rise in equipment spending and China restrictions ease, Applied's process breadth compounds and FY2027's $18.67 proves conservative. You pay 39x on trailing earnings but only 24x on next year; a real secular ramp at $47B revenue means the forward multiple compresses to mid-teens by year two.

Closing thoughts

The October quarter settles it. If receivables and inventory shrink as percentages while revenue grows, the 24x forward is cheap and the ramp is real; if they keep building as growth slows, you paid 39x trailing for the cycle peak. The July quarter converted the case from hope to evidence on growth and margin but left the cash unresolved. The fatter risk is the left tail: a 39x multiple, half again above any level in twelve years, on a business where a third of revenue is two customers and free cash flow trails earnings, means an ordinary cyclical pause costs far more than the beat is worth. What is at risk if concentration or cash breaks is the entire re-rating back toward the low-20s the stock lived in for years; what the right tail is worth is a cheap-looking 24x forward compounding for a decade. On my read the downside is the more probable of the two, and that is a judgment, not a number.

The bet is still that the world's biggest chipmakers keep buying more of Applied's machines every year to build the transistors inside AI chips, and keep paying up for the hardest steps. What breaks it is those two customers slowing, and the pair of numbers that tells you first is receivables and inventory against revenue: if both keep growing faster than sales while October's growth decelerates from 25%, you paid the highest multiple in twelve years for the top of a cycle.