Amkor Technology · NASDAQ: AMKR · Post-earnings decision memo

The print beat. The mix outlook reset expectations.

Q2 proved that advanced-compute mix and factory loading can lift earnings quickly. Q3 says the same mix shift will conceal a communications air pocket—and the Arizona investment makes the cash-flow path much less clean than the P&L.

Reported July 27, 2026 Quarter ended June 30, 2026 After-hours snapshot: $55.95 at 7:59 p.m. ET Evidence through July 27, 2026 Q2 10-Q not yet filed

Verdict: business thesis strengthened; stock setup improved after the reset, but still needs proof.

Revenue, gross margin, EBITDA and EPS all beat the pre-print bar. After removing the real-estate gain and discrete tax benefit, an analyst-derived ~$0.57 partially normalized EPS proxy remains roughly 22% above the common $0.47 consensus. It is not fully normalized because $20.3 million of other income remains unparsed. S1 S6

The problem is cadence, not current demand collapse: Q3 revenue midpoint is about 5.4% below the pre-print Fiscal.ai mean, communications is guided down high single digits, and one SiP migration headwind can last through the first half of 2027. Compute is expected to grow nearly 30% and push gross margin to ~19%, so the earnings guide is better than the revenue guide. S3 S5

At $55.95, AMKR is roughly 24.0–24.4× a partially normalized FY2026 EPS bridge—not obviously cheap for a company spending $2.5–3.0 billion this year and running negative free cash flow. The selloff creates a better watchlist entry, not yet a low-risk one.

Q2 revenue

$1.898B

+25.6% YoY; 4.3%–4.6% above consensus. S1

Q2 gross margin

16.8%

+260 bp QoQ; 160 bp above reported consensus. S1

Part-normalized EPS

~$0.57

After property gain and discrete tax benefit; other income remains.

Q3 revenue gap

–3% to –5%

Versus estimates; midpoint is still +5.4% QoQ and +0.7% YoY. S5 S6

Q3 margin midpoint

19.0%

Mix-led and +220 bp sequentially. S3

AH earnings reaction

–7.8%

$55.95 vs $60.71 regular close; total –13.9% vs Friday. S8

The scorecard: broad Q2 beat, split Q3 message

Metric	Q2 actual	Pre-print bar	Delta	Read-through
Revenue	$1.898B	$1.815–1.820B	+4.3% to +4.6%	Compute and auto/industrial drove the upside.
Gross margin	16.8%	15.5% guide high	+130 bp	About two-thirds of sequential expansion came from volume/utilization.
EBITDA	$400M	$331M Fiscal mean	+21%	Partially normalized for the $21M property gain, ~$379M still beat by ~15%.
Diluted EPS	$0.70	$0.47	+$0.23 / +49%	Headline beat overstates recurring quality.
Part-normalized EPS proxy	~$0.57	$0.47	~+$0.10 / +22%	Still a meaningful beat after two disclosed benefits; other income remains.

Comparison bases combine company guidance, a Fiscal.ai mean and common public EPS estimates. Accounting bases and estimate timestamps are not fully disclosed; see the evidence ledger. S5 S6

Q2 actual versus Amkor's own prior guide

Metric	Prior guide	Actual	Vs high end
Revenue	$1.75–1.85B	$1.898B	+$48M
Gross margin	14.5%–15.5%	16.8%	+130 bp
Net income	$105–130M	~$174M	+$44M
Diluted EPS	$0.42–0.52	$0.70	+$0.18

Net-income guide wording and headline attributable net income differ by about $1.1M of non-controlling interest; economically immaterial, but not identical. S1 S2

Likely revision map

Q3 revenue: down about 3%–5% toward the $2.00B midpoint, depending on provider.
FY2026 revenue: flat to modestly down; the Q2 beat offsets most of the Q3 shortfall.
Q3/FY2026 EPS and EBITDA: materially up on the margin bridge.
FY2026 free cash flow: unchanged to worse; capex remains $2.5–3.0B.
FY2027 revenue: Asia compute offers upside, but the H1 communications headwind offsets it; U.S. strategic agreements primarily de-risk 2028+ capacity.
FY2027 EPS: less certain because Arizona start-up burden begins.

What the tape rejected: a straight-line revenue acceleration. What the tape may underweight: earnings mix and customer-funded capacity.

Earnings quality: $0.70 reported, roughly $0.57 partially normalized

Management disclosed two benefits on the call that are not identified in the release's headline tables: a $21 million real-estate sale gain in operating expense and a $14 million net discrete tax benefit. The property gain had been preannounced and embedded in Q2 guidance, so it did not drive the upside versus Amkor's own bar; the unanticipated tax benefit did. The bridge below uses exact attributable net income and a 20.8% underlying tax rate inferred before the discrete benefit. S2 S3

$173.8MReported attributable net income
~$0.70 EPS

–$16.6MReal-estate gain after inferred normal tax

–$14.0MNet discrete tax benefit

~$143.1MPart-normalized proxy
~$0.57 EPS

What looks operationally durable

Revenue grew 13% sequentially and 26% year over year.
Network utilization moved from the 50s to the 70s in the first half; Q2 was in the high 70s.
Roughly two-thirds of Q2 margin expansion came from volume/utilization and one-third from mix.
Operating income excluding the property gain was approximately $179M, a 9.4% margin.
EBITDA excluding the property gain was approximately $379M, still about 15% above the Fiscal mean.
Advanced-product revenue rose 13.5% sequentially; mainstream rose 8.9%.
Top-ten customer concentration eased to 66% from 68% sequentially and 72% a year ago.

What remains noisy or unconfirmed

The release does not separately identify the property gain or discrete tax item; both are call-only disclosures.
Other income, net was $20.3M, versus $13.7M in Q1. It is not removed from the partially normalized proxy because its components are not disclosed.
Removing all other income would reduce after-tax EPS by another roughly $0.06–$0.07, but that is a sensitivity—not a normalized result.
The Q2 10-Q was not filed as of publication; tax, interest-income and capex detail still need tie-out.

Quality conclusion. The headline EPS surprise overstates fully normalized earnings, but the partially normalized proxy still clears consensus. The core debate is therefore durability of mix and utilization—not whether the quarter was fundamentally weak.

The operating inflection is visible in the quarterly series

Fallback: revenue rose from $1.51B in Q2 2025 to $1.90B in Q2 2026; gross margin rose from 12.0% to 16.8% and operating margin from 6.1% to 10.5%.

Fallback: actual/consensus EPS by quarter was $0.22/$0.16, $0.51/$0.42, $0.69/$0.43, $0.33/$0.23 and $0.70/$0.47.

Quarter	Revenue	Gross margin	Operating margin	Net income	Diluted EPS	Consensus EPS
Q2 2025	$1.511B	12.0%	6.1%	$54.4M	$0.22	$0.16
Q3 2025	$1.987B	14.3%	8.0%	$126.6M	$0.51	$0.42
Q4 2025	$1.888B	16.7%	9.8%	$171.8M	$0.69	$0.43
Q1 2026	$1.685B	14.2%	6.0%	$83.4M	$0.33	$0.23
Q2 2026	$1.898B	16.8%	10.5%	$173.8M	$0.70	$0.47

Historical P&L is company/Fiscal.ai; consensus EPS is a common public-provider series. Q2 2025 included a $32.4M Nanium insolvency benefit in SG&A, making the year-over-year operating comparison harder, not easier. S1 S5

Q3 guide: lower revenue bar, higher-quality mix

Metric	Q3 guide	Midpoint	Pre-print bar	Midpoint delta	Interpretation
Revenue	$1.95–2.05B	$2.00B	$2.07–2.115B	–3.4% to –5.4%	Expectation gap, not absolute decline: midpoint is +5.4% QoQ and +0.7% YoY.
Gross margin	18.5%–19.5%	19.0%	16.8% Q2 actual	+220 bp QoQ	Predominantly product mix, not another utilization step.
Net income	$180–205M	$192.5M	—	+10.1% QoQ	Consolidated basis; despite about $140M of operating expense.
Diluted EPS	$0.72–0.82	$0.77	$0.62–0.66	+17% to +24%	Earnings guide is materially better than revenue guide.

Fallback: using rounded Q2 mix and midpoint growth descriptors, computing adds roughly $121M, auto/industrial $21M and consumer $40M, while communications subtracts about $64M.

Directional end-market bridge

End market	Q2 mix	Q3 sequential guide	What matters
Communications	42%	High-single-digit decline	Memory constraints, build patterns and SiP migration.
Computing	22%	Nearly +30%	Largest HDFO CPU ramp and broad data-center demand.
Auto / industrial	22%	Mid-single-digit growth	ADAS and advanced packaging.
Consumer	14%	Mid-teens growth	Broad IoT demand.

Q2 mix percentages are rounded and growth descriptors are qualitative. Dollar contributions are analyst approximations and sum to about $2.02B, not exact guidance. S1 S3

Decision implication. Q3 can miss the revenue consensus and still beat the earnings consensus. The key test is whether a compute-led mix genuinely produces 18.5%–19.5% gross margin without pulling demand forward.

Transcript debate map: what analysts pressed and what can falsify it

UBS · communications / SiP

Management attributed about half the missing normal Q3 seasonal lift to market factors and half to the SiP move. One application headwind extends through Q4 and the first half of 2027.

Bull: Vietnam improves cost and frees Korea for compute. Bear: weak phones overlap with a self-inflicted transfer. Gate: Q4 communications and Vietnam qualifications.

UBS · advanced-packaging breadth

Amkor cited 11 customers in 2.5D, five customers and 10 active HDFO engagements, and still targets four launches in each platform during 2026. The current CPU ramp is the largest.

Bull: funnel breadth. Bear: near-term economics remain concentrated. Gate: 4+4 launches and disclosed production ramps.

B. Riley · pull-in and Q4

No Q2 benefit was attributed to the transfer; there may have been some general communications pull-in. Management does not expect the usual significant Q4 lift.

Read: the easy “one-quarter air pocket” thesis is not supported. Gate: Q4 guide and transfer completion.

B. Riley · NVIDIA economics

The $1.5B agreement is a prepayment expected in 2027 and returned through U.S. services over an estimated five to 10 years. R&D remains within normal engagement levels.

Bull: customer-funded committed capacity. Bear: not immediate revenue or free cash. Gate: receipt and service ramp.

Melius · Q2 margin bridge

Q2 utilization was in the high 70s; advanced areas were very high while mainstream retained spare capacity. Two-thirds of sequential gross-margin expansion came from volume/utilization and one-third from mix.

Bull: proven operating leverage. Bear: less incremental advanced-capacity slack. Gate: throughput while assets move.

Melius / Needham · Q3 margin

Q3 expansion should be predominantly mix, driven by the steep compute ramp versus communications decline. Management saw no obvious pull-in outside possible communications.

Bull: structural value per package. Bear: concentrated execution. Gate: compute near +30% and GM ≥18.5%.

KeyBanc · Q4 margin

If utilization and mix remain similar, management expects no significant gross-margin delta from Q3.

Bull: ~19% could plateau. Bear: no extra seasonal step-up. Gate: Q4 margin guide.

KeyBanc · Arizona burden

U.S. manufacturing becomes a 2027 headwind; underutilized Arizona capacity will dilute gross and operating margin in 2028 and was contemplated in the long-term targets.

Bull: management says projected phase I capacity is fully committed. Bear: public contract definitions are incomplete and earnings are non-linear. Gate: 2027 start-up cost and 2028 utilization.

Prepared remarks and Q&A are from the timestamped Fiscal.ai/Quartr transcript, cross-checked against the official release/deck and an independent Investing.com transcript. Amkor had not posted an official text transcript at publication. S3

Cash conversion is the counterweight to the P&L inflection

H1 operating cash flow

$382M

Only 11% of H1 revenue.

H1 cash capex

$688M

Before the heavier second-half build.

H1 free cash flow

–$307M

Analyst-derived CFO less PP&E payments.

Net cash

$26M

$2.512B cash/investments less $2.486B debt.

Capex payables

$621M

Up from $244M at year-end.

FY2026 capex guide

$2.5–3.0B

Unchanged despite stronger earnings.

Funding bridge

Amkor issued $1.15B of 0% convertible notes in May, paid $56.4M for capped calls and $20.5M of issuance costs.
The convertible's initial conversion price is $106.37; the capped-call cap is $139.50.
Q2 standalone cash flow, derived from six-month less Q1 totals: CFO ~$236M, capex ~$464M and FCF ~–$227M.
A straight cash subtraction suggests $1.81–2.31B of H2 spend, but capex payables rose about $378M. A payable-aware incurred-capex proxy is ~$1.07B for H1, leaving roughly $1.43–1.93B of H2 additions; cash spend depends on ending payables.

Why the NVIDIA prepayment matters

Customer funding can lower financing risk and validates future U.S. capacity. But the cash is expected in 2027, is matched to services over five to 10 years, and should not be recognized as near-term revenue.

For valuation, P/E and EBITDA multiples therefore look cleaner than economic free cash flow during the Arizona build.

Cash gate. Require a clearer bridge among capex, customer prepayments, CHIPS/credits, debt and eventual Arizona returns—not just higher EPS.

Cash-flow and balance-sheet figures are from the official release; quarterly FCF and H2 spend are derived. S1 S10

The tape repriced cadence; valuation is reasonable, not distressed

Fallback: AMKR closed at $64.96 Friday, $60.71 in Monday's regular session and $55.95 after hours; the 52-week range was $20.86–$96.68.

Fallback: the stock initially traded up after the release, returned near the regular close around the call, and fell to $55.95 by 7:59 p.m. ET.

Yahoo provides chart granularity; the final $55.95 mark is Nasdaq's 7:59 p.m. ET snapshot. The timing is consistent with call details weighing on the after-hours reaction, but does not prove causation. Monday's –6.5% regular-session move occurred before the release, is unbenchmarked here and is not attributed to earnings. S8

Valuation lens at $55.95	Revenue / earnings base	Multiple	Use and limitation
Equity / enterprise value	$13.87B / $13.84B	—	247.828M shares and official $26M net cash.
Pre-print FY2026	$7.60–7.67B / $2.08–2.14 EPS	1.80–1.82× EV/sales 26.1–26.9× P/E	Stale after the print; useful only as the starting bar.
Headline FY2026 EPS bridge	$2.42–2.46 EPS	22.7–23.1× P/E	Adds full reported Q2 and Q3 midpoint upside to $2.08; includes Q2 property/tax benefits. Analyst arithmetic.
Part-normalized FY2026 bridge	$2.29–2.33 EPS	24.0–24.4× P/E	Replaces the Q2 $0.47 bar with ~$0.57 and adds Q3 midpoint upside; holds Q4 unchanged.
Fiscal NTM re-mark	Same estimate denominators	22.8× P/E 9.0× EV/EBITDA 1.67× EV/sales	Mechanically scales a prior-close snapshot to the AH price.
Pre-print FY2027	$8.44–8.59B / $2.30–2.62 EPS	1.61–1.64× EV/sales 21.4–24.3× P/E	Arizona start-up costs make the range unusually fragile.

Entry discipline

Roughly $46 equals about 20× the partially normalized FY2026 bridge. That is not a price target; it is an illustrative hurdle where execution and capex risk receive a larger discount.

At $55.95, the alternative is time rather than price: wait for Q3 proof that compute, mix and margin offset communications without a cash or capex surprise.

One-provider peer screen

Company	NTM P/E	NTM EV/EBITDA	NTM EV/sales
AMKR · AH re-mark	22.8×	9.0×	1.67×
ASE	31.8×	13.3×	3.45×
JCET	53.9×	19.3×	3.33×

Fiscal snapshots around 8:00 p.m. ET; AMKR is mechanically re-marked to $55.95. Listing venue, FX, business mix, estimate horizon and Arizona capex materially reduce comparability. S5

Long-term upside is real—and deliberately non-linear

Investor Day framework	2025 actual	2028 target	2030 ambition	Current read
Revenue	$6.7B	$9.0B ± $0.5B	~$11B+	Pre-print 2027 estimates already approached $8.5B.
Gross margin	14.0%	17.5% ± 100 bp	~22%+	Q3 2026 midpoint is 19%, but Arizona pulls against it later.
Diluted EPS	$1.50	$2.50 ± $0.25	~$5.00+	2026 may approach $2.5 before 2027–28 start-up burden.

TSMC

Ten-year Arizona agreement

TSMC will procure advanced packaging and test services from Amkor. It validates the site, but public economics remain undisclosed. S9

NVIDIA

$1.5B multi-year agreement

A 2027 prepayment supports U.S. capacity and is earned through services over an expected five to 10 years. S9

Arizona

Projected committed, then underutilized

Management says projected phase I capacity is fully committed; the public definition is not contractual backlog. Construction completes in 2027, production begins in 2028, break-even is estimated in 2029 and full utilization in 2030. S3 S4

The tension: 2026 profitability can look like the 2028 target before Arizona dilution arrives. Investors should not extrapolate Q3 margin linearly; the value case requires accepting a deliberate earnings valley to reach a larger U.S. platform.

Proof points, break points and the next decision window

What proves the thesis

Q3 compute: nearly 30% sequential growth and a visible HDFO CPU ramp.
Q3 margin: gross margin at or above 18.5% with no broad pull-forward.
Portfolio: four 2.5D and four HDFO launches during 2026.
Communications: high-single decline contains the damage; qualifications progress.
Cash: capex stays within $2.5–3.0B and customer/government funding arrives as framed.
Arizona: explicit unit economics and a credible 2029 break-even path.

What breaks the thesis

Mix: compute misses while communications remains weak, pulling GM below guide.
Transfer: the SiP headwind extends beyond the first half of 2027 or costs more than expected.
Concentration: the largest CPU ramp slips and the broader 2.5D/HDFO funnel stays pre-production.
Capex: spend exceeds $3B without matching committed economics or non-dilutive funding.
Arizona: 2027 operating-margin dilution exceeds the previously discussed 1%–2% full-year effect.
Cash quality: payables and debt rise faster than durable operating cash flow.

Catalyst timeline

Near term · Q2 10-Q

Resolve other-income, tax, real-estate gain, debt and capex accounting.

Q3 2026 · expected late October

Test $2.00B revenue midpoint, 18.5%–19.5% GM, $0.72–$0.82 EPS and compute near +30%.

Q4 2026 / H1 2027

Test communications duration, Vietnam qualifications and the absence of a normal Q4 lift.

2027

NVIDIA prepayment, Arizona completion, Songdo capacity and first start-up burden.

2028–2030

Bridge technologies, Arizona production, estimated 2029 break-even and 2030 full utilization.

Positioning discipline

For a new position: wait for either a better price around the illustrative 20× hurdle or Q3 proof.

For an existing long: the business evidence improved, but the mandate must tolerate negative FCF and a planned earnings valley.

For a short: do not chase a 7.8% post-release decline on top of a weak pre-print session when partially normalized earnings and customer funding are improving; the bear case needs a failed compute/margin ramp.

Bottom line. Upgrade the business, not yet the entry. AMKR becomes more compelling when the market prices the investment burden—or management proves the mix can fund it.

Evidence ledger, source posture and open questions

Primary company and SEC sources control reported figures, guidance and target wording. Fiscal.ai and public providers supply estimates, transcript retrieval and price/valuation snapshots. Derived figures are labeled; no settled post-earnings consensus was available at publication.

SEC Exhibit 99.1 — Q2 2026 results and Q3 guidance · 8-K
Filed July 27, 2026 at 4:08 p.m. ET. Controls current-quarter reported figures, product/end-market data, cash flow and official guidance.

Primary · current

SEC Exhibit 99.1 — Q1 2026 · Q1 10-Q
Controls prior Q2 guide and most recent filed quarterly accounting detail.

Primary · historical

Official Q2 webcast page · Quartr timestamped transcript · Investing.com cross-check
Call held July 27 at 5:00 p.m. ET. No official text transcript was posted at publication; call-only items are identified as such.

Mixed · current

SEC Investor Day 2026 presentation
May 21, 2026. Controls 2028/2030 targets, Arizona timeline and long-term assumptions.

Primary · company

Fiscal.ai AMKR revenue estimates · AMKR overview · ASE overview · JCET overview · local wrapper snapshots
Pulled July 27 around 8:00 p.m. ET. Q3 revenue mean $2.115B from eight estimates; estimate timestamp/basis is not fully exposed. Used for screen-grade estimates and a same-provider peer screen.

Secondary · timestamped

Zacks Q2 beat and EPS consensus · StreetInsider Q3 guide comparison · MarketBeat/Fiscal earnings history
Direct public cross-checks for Q2/Q3 EPS, revenue expectations and historical consensus. Provider bases vary; no precise gross-margin consensus is used in the decision tables.

Secondary · caveated

FactSet consensus snapshot · S&P Global public forecast
July 24–27 snapshots for FY2026/FY2027 valuation ranges. May be partially stale or partially refreshed; not labeled as settled post-print consensus.

Secondary · timestamped

Nasdaq quote API · Nasdaq history · Yahoo chart API
Frozen July 27 at roughly 8:00 p.m. ET. Nasdaq controls the $55.95 final mark; Yahoo provides chart granularity.

Secondary · market

TSMC ten-year agreement · NVIDIA $1.5B partnership
Official company releases, June 16 and July 23, 2026; public terms are incomplete and the call clarifies timing.

Primary · company

S10

SEC exhibit — 0% convertible notes
Controls $1.15B final proceeds, $106.37 initial conversion price and $139.50 capped-call cap.

Primary · financing

S11

SEC submissions index · Amkor quarterly results
Checked July 27 after the call. Q2 10-Q and official text transcript were not yet available.

Primary · status

Open questions for the 10-Q

Composition and recurring nature of $20.3M other income, net.
Exact tax treatment of the property gain and discrete tax benefit.
Capex commitments, payables and customer/government funding schedule.
Top-customer concentration and any new contractual commitments.
Convertible-note accounting and diluted-share impact.

Method

Partially normalized EPS proxy = attributable NI minus estimated after-tax property gain minus discrete tax benefit, divided by diluted shares; other income remains.
Free cash flow = cash from operations minus PP&E payments.
Enterprise value = after-hours equity value plus debt minus cash and short-term investments.
Mechanical FY2026 EPS bridge holds Q4 unchanged; it is not a forecast or price target.
All inference is labeled and rounded; this is research, not personalized investment advice.