The Business
Amkor Technology (NASDAQ: AMKR) is an OSAT — Outsourced Semiconductor Assembly and Test. It's the back end of chipmaking: after a foundry fabricates the silicon, Amkor packages the chips (flip chip, wafer-level, 2.5D/3D, chiplets, system-in-package) and tests them. It's the world's largest US-based OSAT and the clear #2 globally behind Taiwan's ASE.
The reason it matters for the AI trade is advanced packaging. As chipmakers hit the limits of shrinking transistors, more of the performance gain now comes from how you assemble chips — bonding HBM stacks to a GPU, integrating multiple chiplets on one substrate. That's the same bottleneck Applied Materials and the memory makers keep pointing at, viewed from the assembly side. In Q2, advanced products were $1.557 billion — roughly 82% of revenue and up 27% year over year — so this is increasingly an advanced-packaging company, not a commodity assembler.
End markets span computing (AI/HPC — the growth driver), smartphones (a big, seasonal second-half business), automotive & industrial, and communications. Packaging is ~88% of sales, test ~12%.
Where the Price Is
As of the September 29 session (quote pulled live from Yahoo Finance):
- Price: ~$54 (+1.8% on the day)
- 52-week range: $28.03–$96.68 — it nearly hit $97 in the AI-packaging enthusiasm, then round-tripped to ~$54; that's ~44% below the high, but still nearly 2x off the low
- 50-day MA: ~$52.47 — price is a few percent above it
- 200-day MA: ~$57.14 — price is ~5% below it; the 200-day is the overhead line the chart is working against
- RSI(14): ~56 — neutral. Not washed out, not extended
- Market cap: ~$15 billion
The chart reads as a name consolidating in the low-to-mid $50s, pinned between its 50-day (support) and 200-day (resistance) after giving back a big euphoric move. It's a coiled range, not a trend — the direction gets decided on a reclaim of the 200-day or a loss of the 50-day.
The Numbers
Q2 2026 (reported July 27) was a record, and the operating leverage is the story:
- Revenue: $1.90B, up 26% YoY and up 13% sequentially (from $1.685B in Q1) — a record
- EPS: $0.70 vs $0.22 a year ago, a ~45% beat. One honest note: that $0.70 included a ~$0.07 one-time tax benefit, so underlying EPS was closer to $0.63 — still a large beat, just not quite the headline
- Gross margin: 16.8%, up from 14.2% in Q1 and 12.0% a year ago — expanding on higher utilization
- Operating income: $200M, operating margin 10.5% (up from 6.1% a year ago); EBITDA $400M (~21% margin)
- Balance sheet: $2.5B cash against $2.5B debt — roughly net-neutral — plus a small dividend (~$0.08/quarter)
- Guidance (Q3 2026): revenue $1.95–2.05B (~$2.0B midpoint) with gross margin guided up to 18.5–19.5%
Notice the margin trajectory: 12% → 14% → 17% gross margin over four quarters, with Q3 guided toward ~19%. That's the operating leverage of a high-fixed-cost manufacturing business as utilization climbs — the same dynamic that drives the semicaps, just off a much lower margin base.
The Strategic Catalysts
This is what separates Amkor from a run-of-the-mill packaging name right now — it has spent 2026 turning itself into contracted AI infrastructure:
- A 10-year advanced-packaging agreement with TSMC. This is the big one. It establishes a framework to expand advanced packaging and test capacity and pairs TSMC's leading-edge fabrication with Amkor's back-end — explicitly to build a more resilient US semiconductor supply chain. A decade-long tie-up with the most important foundry on earth is a profound validation of Amkor's position.
- A multi-year strategic partnership with NVIDIA. Advanced packaging for the exact AI/HPC accelerators driving the whole cycle — putting Amkor directly in the Nvidia supply chain on a multi-year basis.
- The Arizona plant. Amkor is building a large advanced-packaging and test facility in Arizona, next to TSMC's new fabs, supported by CHIPS Act funding. It localizes the AI back-end in the US — a policy tailwind and a structural onshoring story, letting a chip be fabricated and packaged domestically for the first time at scale.
Put together, Amkor isn't just riding AI demand — it's being designed into the Western AI supply chain by the two companies (TSMC and Nvidia) that matter most, with government money underwriting the capacity.
Why It Sold Off
So why is a strategically-blessed, record-setting company 44% off its high? A few reasons, and they're the crux of the bear case:
The guide was light. Q2 beat handily, but the Q3 revenue midpoint (~$2.0B) came in a few percent below the ~$2.09B analysts wanted. In this tape, a strong quarter with a soft guide gets sold — the market read it as a question mark over how sustainable the momentum is, the same reaction Marvell and Ciena got.
It de-rated with the whole complex. The AI-capex-peak worry that pressured the semicaps and optical names hit Amkor too — and as the lowest-margin, most capital-intensive link in the chain, it's treated as higher-beta and lower-quality when sentiment turns.
It ran too far, too fast. The move to ~$97 priced in a lot; the round-trip to ~$54 is the market repricing the euphoria, not necessarily the business.
The Thesis
Bull case: Amkor is the indispensable Western back-end of the AI chip, and advanced packaging is the performance bottleneck now that transistor scaling is slowing. It has a 10-year TSMC agreement, an Nvidia partnership, and CHIPS-funded Arizona capacity coming online — contracted, multi-year demand from the most important names in the industry. Revenue is growing 26% with gross margin climbing 12% → 17% → a guided ~19% as utilization rises, and at ~$54 (roughly high-teens to low-20s times this year's earnings, less on next year's as Arizona and the TSMC/Nvidia programs ramp) you're paying a reasonable multiple for a structural onshoring-plus-AI story that has already de-rated 44%.
Bear case: an OSAT is a structurally lower-quality business than the fabless designers or the equipment makers it serves. Gross margins in the mid-to-high teens (versus 50%+ for Applied Materials) tell you Amkor is a capital-intensive manufacturing taker, not a pricing-power leader. The business is cyclical and customer-concentrated (top-10 customers ~66% of revenue, with heavy smartphone seasonality and hyperscaler concentration), it competes with a larger ASE, and the Arizona buildout consumes capex and pressures near-term free cash flow. The soft Q3 guide is the market's evidence that the momentum may be closer to a cyclical high than a new baseline.
What would change my mind, either way: the swing factors are (1) the TSMC and Nvidia programs and the Arizona ramp converting into recognized revenue and sustained margin expansion; (2) whether gross margin holds its climb toward the high teens/20s or stalls; and (3) the customer-concentration and capex trajectory. Continued margin expansion with the strategic programs ramping confirms the bull case. A margin stall, a demand air-pocket from a concentrated customer, or a capex-driven cash squeeze confirms the bear.
Trade Plan
Setup: a strategically-important, reasonably-valued name coiled between its moving averages after a sentiment reset, with real fundamental catalysts (Arizona ramp, the next print) ahead. This is a level-and-execution trade, not a chase.
Levels that matter:
- ~$57 (200-day) — the overhead line the chart is deciding at. A decisive reclaim on volume flips the intermediate trend up and is the cleaner long trigger.
- ~$52 (50-day) — near-term support. Holding it keeps the low-$50s base intact; losing it opens air toward the low $40s.
- $28.03 — the 52-week low, the reference if the cycle rolls and the capex drag bites. Naming it is the floor of the range, not a forecast.
- $96.68 — the prior high and the only meaningful overhead target if the AI-packaging story re-rates.
How I'd treat it: respect that this is a lower-margin, capital-heavy cyclical — size it accordingly. The constructive entries are a confirmed reclaim of the 200-day (~$57), or evidence in the next print that gross margin keeps climbing and the TSMC/Nvidia/Arizona programs are ramping on schedule. A margin stall or another soft guide is the tell to stand aside.
Horizon: this is a multi-quarter onshoring-plus-advanced-packaging thesis gated by capacity coming online. Trade the range; own the thesis only to the extent each quarter confirms the margin expansion and the strategic programs are real revenue, not just announcements.
Bottom Line
Amkor is a genuinely strategic business the market is pricing like a commodity — the largest US-based advanced-packaging house, written into the Western AI supply chain by TSMC and Nvidia for the next decade, with government-funded Arizona capacity on the way, growing 26% with margins expanding, at a reasonable multiple after a 44% de-rate. That combination is more interesting than a priced-for-perfection name. But the reason it trades cheap is real too: an OSAT is a lower-margin, capital-intensive, customer-concentrated, cyclical business, and the soft guide gave the market a reason to doubt the durability. So the honest framing is that this is a good business at a fair price with a genuine structural tailwind — not a mispriced compounder, and not a falling knife. I'd stalk the 200-day reclaim, watch the gross-margin trajectory and the Arizona ramp as the tells, and size for the mid-teens-margin cyclical it is rather than the AI pure-play the headlines make it sound like. The strategic partnerships are the bull's evidence; the margin structure is the bear's, and the next few quarters of execution decide which one wins.
Price and technical data pulled live from Yahoo Finance at the September 29, 2026 session. Fundamentals from Amkor's Q2 2026 release (reported July 27, 2026) and earnings-call commentary, including the TSMC and NVIDIA partnership announcements and Arizona capacity plans. Note the Q2 EPS of $0.70 included a ~$0.07 one-time tax benefit. Valuation figures are approximate. Levels are point-in-time and will move.