A record-margin quarter, a stock cut in half, and a capex plan increasingly funded by customers instead of the equity market
AMKR · Earnings Call · 2026-10-26
A record quarter the tape has already voted against
Amkor's third quarter was, on the operating numbers, the best it has looked in years: revenue of $1.99 billion, up 5% sequentially; gross margin of 18.3%, the highest since 2022 and up 400 basis points year on year; operating margin of 11.3%; EPS of $0.76. Computing rose 25% sequentially to a record, and automotive and industrial set a record for the second consecutive quarter. And yet the stock walks into this print down 54.5% over the last 90 trading days, 58% below its June 22 peak of $93.55. That is the tension of this report in one line: the operating business is accelerating while the equity has been repriced as though the advanced-packaging cycle were already over.
The drawdown is not idiosyncratic. On the market theme tape, AI data centers shows 47 decliners against 6 advancers over 90 days, with high bandwidth memory, co package optics and data center capacity all lower. The entire AI hardware complex de-rated; an OSAT sitting at the end of that chain took one of the harder hits. So the real question this call had to answer was never "is demand good" — demand is plainly good — but "is this capital structure defensible if the market stops paying up for growth."
From a demand story to a funding story
Look at what shifted inside the company's own keyword set. Last quarter, the top-ranked concepts were customer engagement and packaging complexity. Both fell out of the top-30 this quarter. In their place: advanced packaging capacity, payment arrangements, CapEx guidance, Phase 1 construction, and — tellingly — incremental debt. The narrative moved from "are we winning the AI socket" to "how does the build get paid for."
The answer management gave is the most consequential thing on the call. Arizona grew from a two-phase $7 billion plan into a campus of more than 90,000 square meters of cleanroom and an estimated $12 billion of investment. Customer prepayments have scaled with it: advance payment arrangements now exist with "a dozen customers," and the analyst on the call sized the aggregate at roughly $3 billion. Kevin Engel's framing:
Customers are not only planning further ahead, but are also making long-term commitments and providing advance payments tied to future capacity requirements. This gives us confidence as we accelerate our capital spending above historical norms.
That is a genuinely different contracting model for an outsourced assembly house — historically a business that ate the cyclical risk of building capacity on speculation. If it holds, it converts a capex cycle into something closer to tolling. The bears' case is that the prepayments are narrow (a handful of hyperscaler programs), that some of the equipment is effectively bespoke, and that the residual risk stays with Amkor. Engel pushed back on exactly that, in the most quotable exchange of the call:
They understand that these are very significant capital investments and there needs to be a return on these capital investments... They are more focused on making sure they secure the capacity.
And on the risk-management question of whether the tools can be redeployed: “the technology is very fungible to different customers” — Kevin Engel, President & Chief Executive Officer · 2026-10-26.
The mix engine is doing the work
What made the margin number possible is that communications is falling apart in a way Amkor has been telegraphing for two quarters. Comms revenue fell 8% sequentially in Q3 and is guided down high single digits in Q4; consumer is guided down mid-teens. The culprits are the same three named in July — the SiP migration to Vietnam, component shortages hitting premium-tier phones, and memory pricing. Engel was blunt that both ecosystems are down: “we do see Android being down in Q4 compared to Q3, and we also see iOS” — Megan Faust, Executive Vice President · 2026-10-26. In July the company warned the SiP headwind “is probably going to extend into Q4 and even into the first half of next year” — Kevin Engel, Chief Executive Officer · 2026-07-27 — and that is precisely what has happened.
The remarkable part is that this no longer hurts. Gross margin rose 150 basis points sequentially while the largest legacy segment shrank, because the mix rotated toward the CPU program HDFO ramp and new 2.5D program work in networking and data center. Computing is now the ballast, guided to over 50% year-on-year growth in Q4. Diversification was the explicit defence: “we are being cushioned by the strength in compute” — Megan Faust, Executive Vice President · 2026-10-26. Note the check on this. Last quarter Amkor guided computing to roughly +30% sequentially and delivered +25%, and Joe Moore asked why; Engel blamed work-in-progress loading across a flow with "over 300 steps," not supply. Credible, but unverifiable — and worth watching into Q4.
The capex cliff, and the numbers that matter
2026 capex narrows to about $2.5 billion; 2027 steps up hard, and the facility piece alone is large — Arizona up to 15% of the $12 billion, Korea $400-450 million, Vietnam around $50 million — which the CFO described as higher than $2 billion "just around those programs," before equipment. Depreciation rises mid-teens in 2026 and more in 2027. Capex had already doubled year on year by the June quarter, and free cash flow went with it — negative $227 million, a negative 12% margin. This is the arithmetic the equity market has been discounting.
The funding answer was reassuring but incomplete. Faust said the balance sheet is solid through 2027 and that the prepayments mean “even pushes through '28 where there may not even need to be incremental debt” — Megan Faust, Executive Vice President · 2026-10-26, with anything beyond that "too early to tell." Meanwhile gearing already moved: liabilities to assets rose 6.7 points year on year to 51.8%, debt-to-EBITDA sits at 1.7x versus 1.2x in February, and interest coverage of 12x is comfortable but well off its 21x peak. Back in July the CFO framed 2027 as the year the bill lands: “So then it just comes down to next year to be thinking about the U.S. manufacturing burden. So that is going to be a headwind.” — Kevin Engel, Chief Executive Officer · 2026-07-27 That headwind is now a stated number rather than a footnote.
What to watch
Three things. First, whether the dozen prepayment customers convert into disclosed, cumulative dollar commitments — that single figure decides whether this is a customer-funded build or just another OSAT capex cycle dressed up. Second, whether communications actually troughs in the first half of 2027 as signalled; if it does not, the mix offset loses its shock absorber. Third, whether computing's sequential cadence in Q4 finally matches the guide — one miss is not disqualifying, but a second would undercut the central claim that packaging, not silicon, is now the bottleneck. On a stock 58% off its high, with a pricing environment management calls attractive enough to pass cost through, and a margin profile that is genuinely improving, the debate is no longer about demand. It is about whether Amkor can spend $12 billion of other people's money well.