AMKR · Amkor Technology, Inc. · Stock research

Current Thesis

The re-rating leg that carried this name from June into late July has broken. Amkor reported Q2 '26 on 2026-07-27 and beat both lines it had guided on 2026-04-27: net sales ~$1.90B (+25.6% YoY) against a $1.75–1.85B guide, gross margin 16.8% against a 14.5–15.5% guide, net income attributable of $173.8M and diluted EPS $0.70. Shares fell 6.54% in the following session to $60.71 (Investing.com report on the Q2 slides), and the 2026-08-14 close was $58.99 — through the ~$64 June low that the 2026-07-26 note named as the structural break level, and below $65, the lowest published sell-side target after B. Riley cut from $75 on 2026-07-28. Four consecutive positive events have now been distributed into: the ~50% Q1 EPS beat (-8.6% on 2026-04-27), the TSMC 10-year Arizona deal (2026-06-16, faded by 2026-06-25), the $1.5B Nvidia packaging/test collaboration (2026-07-23), and a record Q2. What an investor would be buying today is no longer the "US AI back-end anchor re-rating" — it is a margin and mix step-up (Q3 gross-margin guide 18.5–19.5%, Computing guided ~+30% sequentially) plus an undated balance-sheet event: Bloomberg reported on 2026-08-11 that Amkor is weighing a stake sale in its China unit at a $1.0–1.5B valuation. Nothing on the company calendar resolves before the Q3 print in late October.

Bullish and bearish views on Amkor Technology, Inc.

The model's bull view on Amkor Technology, Inc. (AMKR), in brief: Margin is inflecting off the OSAT ceiling: gross margin 14.2% (Q1 '26) → 16.8% (Q2 '26, reported 2026-07-27) → 18.5–19.5% guided for Q3 '26. The bear view: Management guided the Arizona facility as a gross-margin headwind in 2027 and 2028 on underutilisation as it ramps (2026-07-27 call). Both cases follow in full.

Bull Case

• Margin is inflecting off the OSAT ceiling: gross margin 14.2% (Q1 '26) → 16.8% (Q2 '26, reported 2026-07-27) → 18.5–19.5% guided for Q3 '26. That is the first sequence in this cycle that argues the mid-teens structural ceiling is moving.
• Q3 '26 guide (2026-07-27): net sales $1.95–2.05B, net income $180–205M, EPS $0.72–0.82 — a second consecutive record if delivered.
• Mix shift is doing the work: Computing guided ~+30% sequentially for Q3, Consumer mid-teens, Automotive/Industrial mid-single-digit. Growth is coming from the AI-adjacent segment rather than the smartphone segment.
• Two front-end anchors remain in place: the TSMC 10-year Arizona advanced-packaging partnership (2026-06-16, terms undisclosed) and the $1.5B Nvidia advanced-packaging and test collaboration (2026-07-23).
• Optional de-risking of the Arizona spend: Bloomberg (2026-08-11) reported an adviser engaged to prepare a carve-out of the China business, valued at $1.0–1.5B, with Amkor potentially retaining a minority stake. Proceeds at that scale would offset a meaningful share of FY26 capex of $2.5–3.0B.
• Price sits below every published target in the post-print cluster: B. Riley $65, Goldman $65, Morgan Stanley $70, Needham $90 (all dated 2026-07-28), against a 2026-08-14 close of $58.99.

Bear Case

• Management guided the Arizona facility as a gross-margin headwind in 2027 and 2028 on underutilisation as it ramps (2026-07-27 call). The margin inflection above has an explicit expiry attached by the company itself.
• Communications is guided DOWN in Q3 on two counts: memory supply and pricing dynamics compressing smartphone volumes, and disruption from migrating system-in-package production from Korea to Vietnam. This is the segment that grew ~32% in Q2 and carries the Apple concentration.
• The market's verdict on the print was -6.54% to $60.71 the next session, then a further drift to $58.99 by 2026-08-14. Record revenue plus a two-point gross-margin beat did not hold a bid.
• Sell-side marked down into the beat: B. Riley $75 → $65 on 2026-07-28, Goldman at $65, Morgan Stanley $69 → $70.
• FY26 capex $2.5–3.0B (65–70% facilities, 30–35% advanced-packaging capacity) against Q2 net income of $173.8M — the funding gap is structural through the ramp.
• The China carve-out is preliminary: no buyer, no agreed valuation, and Bloomberg's own framing says details are subject to change. A stalled process removes the only near-term source of new information.

Setup & Price Structure

The narrative is dead for the leg that was being bought. The dating is specific — 2026-07-28, when a record quarter and a gross-margin beat produced -6.54% and a target cut from the covering analyst who had been at $75; and the loss of the ~$64 June low, the level the prior note published as the structural break. The 2026-08-14 close of $58.99 is -36.9% from the 52-week high of $93.55, with a three-month return of -16.1%. The narrative did not fail on fundamentals; it failed on distribution — every catalyst since April has been sold, and the last one was sold on the best numbers of the cycle.

Positioning and crowding observables, stated as observed:

• RSI(14) at 48 on 2026-08-14 — mid-range. No washout low, no overbought condition. The drawdown has been a grind rather than a capitulation, which is the shape that historically leaves a supply overhang unresolved.
• Price is below the lowest dated post-print target ($65, B. Riley 2026-07-28) while aggregated consensus sits at an average of $77.11 / median $75 (stockanalysis.com). The gap between marks and price means the sell-side has not yet reset to the tape.
• Retail-facing coverage recurred through the drawdown: Benzinga ran backward-looking "if you invested $X years ago" pieces on 2026-06-10, 2026-06-17 and again on 2026-08-03 — the third one after the break.
• Issuance came ahead of the drawdown, not into it: the $1.0B 0%-coupon convertible due 2031 settled 2026-05-05, roughly seven weeks before the June low.
• No earnings date inside the next 30 days, so there is no imminent binary compressing the range in either direction.

Catalyst Calendar (next 30 days)

• ~2026-09 (est.): Apple's autumn product launch window. Amkor's Communications segment is already guided lower for Q3; build commentary around the launch is the nearest read-through on whether "memory supply and pricing dynamics" are a one-quarter or multi-quarter drag.
• ~2026-10-26 (est.): Q3 '26 print. Q1 landed 2026-04-27 and Q2 landed 2026-07-27, so late October is the pattern-consistent estimate. This is the next date that resolves anything.

Elapsed catalysts

• No company-scheduled event. The 30-day window from 2026-08-16 contains no confirmed Amkor print, guidance update or capital-markets day. This is the observation, not an omission. (passed 10d ago)
• Undated, live since 2026-08-11: the China unit stake-sale process. Bloomberg reported an adviser engaged and initial interest being gauged at a $1.0–1.5B valuation. Any headline confirming a buyer, a price, or an abandonment can land inside the window without warning. (passed 15d ago)

What Would Change Our Mind

The structure that has to hold is the post-print zone the stock has occupied since 2026-07-28 — the $60.71 reaction close and the $58.99 mark on 2026-08-14 bracket it. A weekly close below $58 gives that up and turns the July break into continuation of the decline from $93.55; the secondary condition is the ~2026-10-26 (est.) Q3 print arriving with Communications still guided lower and the China carve-out either shelved or unpriced.

On the other side, the specific evidence that would rebuild a leg: gross margin printing at or above the 18.5–19.5% Q3 guide with Computing delivering the ~+30% sequential step; a signed China transaction at or near the reported $1.0–1.5B valuation with disclosed proceeds applied to the Arizona build; or the first disclosed revenue-recognition detail on the Nvidia collaboration, which as of 2026-07-23 remains a development framework with no stated volume floor. Absent those, the constructive read is a company compounding revenue 25% YoY whose equity is being marked as a capex story, and that gap does not close on a schedule anyone can point to.

Correlation Notes

• Memory cycle: management named "memory supply and pricing dynamics" as the driver of the Q3 Communications decline (2026-07-27). Amkor is now partly a short on memory cost inflation flowing into smartphone bills of materials — an unusual linkage for a back-end assembler, and one that ties the name to DRAM/NAND pricing headlines rather than only to unit volumes.
• Front-end anchors: TSMC (2026-06-16 Arizona partnership) and Nvidia (2026-07-23 collaboration) headlines have both produced sold rallies. The pattern of the last two quarters is that partner news moves the stock for one to two sessions and then reverts.
• Apple: Communications was the largest growth segment in Q2 (~+32%) and is the guided-down segment in Q3. iPhone build revisions transmit here with roughly a quarter's lag.
• Semis beta: the iShares Semiconductor ETF was up more than 80% YTD as of 2026-05-26. Amkor's -36.9% distance from its own 52-week high through 2026-08-14 means it has been decoupling from the sector, not tracking it.
• China/geopolitics: the reported carve-out (2026-08-11) is exposure being reduced, not added. Any tightening of US or Chinese rules on semiconductor asset transfers directly affects whether that process closes and at what price.