Expected idiosyncratic alpha: ≈0%. Four kill conditions met simultaneously. Founding family sold $487M of stock two weeks ago at a 13% discount to market. NIST filing confirms 5-year dividend and buyback restrictions from CHIPS Act.


What the Market Prices

At 34x trailing, 22x forward, AMKR prices:

  • HDFO (advanced fan-out for AI) revenue nearly triples in 2026
  • Gross margin expands from 14.0% toward 15%+ on mix shift
  • Arizona facility built on schedule ($7B total, Phase 1 by 2028)
  • Computing end market grows 20%+ YoY
  • Apple/Qualcomm base (41% of revenue) stable

Forward P/E of 22x on ≈$2.30 EPS requires ≈15-19% revenue growth AND margin expansion simultaneously. Q1 guided $1.6-1.7B with 12.5-13.5% gross margin.


Factor Decomposition

Single-factor OLS, AMKR ~ SMH, 250 trading days. No multi-factor collinearity — adding SPY improves R-squared by 0.2% (worthless; corr SMH/SPY = 0.87).

Factor        Beta    Var Explained    Edge?
SMH (semi)    +1.30   57.0%            NO
Idio                  43.0%            NO (see edge audit)

Orthogonal alpha: 22.9% annual
sigma_idio: 41.8% annual
R-squared: 57.0%

43% idiosyncratic. Target is >75%. This is a 1.30x levered semiconductor ETF that does packaging. ASE (ASX) is structurally identical: 43% idio, similar SMH loading. The entire OSAT sector is indistinguishable from levered semi beta.

1Y Move Attribution

AMKR total return:          +133%
SMH return:                 +75%
SMH contribution (1.30x):  +98%    → 73.5% of AMKR move
Idiosyncratic residual:    +35%    → 26.5% of AMKR move

The +35% idio component IS the HDFO narrative repricing. It already happened over 12 months. The question is whether there's another +35% of idio upside — and nothing in the primary sources suggests a second repricing catalyst the market hasn't already priced.


Edge Audit

SignalAMKRThresholdVerdict
Market cap$12.6B<$5BFAIL
Analyst coverage9<8FAIL
1Y momentum+133%<100%FAIL
Insider behaviorKim family sold $487MNet buyingFAIL
Source qualityAll public filings/transcriptsDifferentiated sourcesFAIL

No informational advantage across five passes. Every finding — HDFO demand, margin tailwinds, Arizona risk, competitive dynamics — available to 9 sell-side teams. The ASE cross-reference (equipment spend gap) is one transcript read from obvious.


Variable Templates

1. HDFO Revenue Ramp

FieldValue
VariableHDFO revenue growth 2026
Consensus"Nearly triple" (≈$450-500M), H2-weighted
Our estimateAchieves but timing may slip 1Q
DeltaNone — consensus already prices the guide
EvidenceAMKR transcript: "expecting nearly triple over course of year." Two programs in final qualification for AI datacenters. Equipment vendors (CAMT, KLIC) confirm OSAT demand broadly but can't distinguish AMKR-specific installs from ASE. CAMT: "OSATs important business...significant growth year, already POs on hand."
Base rateCoWoS ramp at TSMC slipped 6+ months in 2023. OSAT qual programs routinely slip 1-2 quarters.

2. Gross Margin

FieldValue
VariableFY2026 gross margin
Consensus≈14-15%, Q1 guided 12.5-13.5%
Our estimate13.5-14.5% FY, H2 possibly 15%+
Delta+50-100bps — not enough to move a $12.6B stock
EvidenceVietnam breakeven Q4 2025 (removes 90bps headwind). ASE transcript: "favorable pricing environment." SQNS transcript: "AI eating most capacity OSAT...increasing price." 10-K materials 55.2% of revenue (stable).
Base rateAMKR gross margin range 2019-2025: 14.0%-16.3%. Never sustained above 16.3%. ASE runs 24-25% on similar services — structural gap is scale, not technology.

3. Arizona CapEx and Incentive Funding

FieldValue
VariableArizona funding gap and FCF impact
Consensus$2.5-3.0B CapEx, offset by government incentives
Our estimate$2.85B incentive figure is plausible but multi-year and conditional
DeltaManageable — previously overstated this gap
EvidenceNIST confirms $407M direct CHIPS + $7M workforce. ExecutiveGov: $200M proposed loans. Treasury ITC: up to 25% of qualified capex. On $7B total, math works to ≈$2.4-2.8B if all tranches close. BUT: NIST confirms "dividend and buyback restrictions for 5 years" and milestone-based disbursement. Funding is real but backloaded and conditional.
Base rateU.S. semi fabs: TSMC AZ +63% overrun, Intel OH +40%, Samsung TX +47%. AMKR is building packaging facility (less complex than a fab) but still first U.S. build for this company.

Correction from prior draft: The $2.85B is not fabricated. Federal direct ($414M) + proposed loans ($200M) + 25% ITC (up to $1.75B) + state/local = plausible total. The risk isn't that the number is fake — it's that disbursement is milestone-conditional and the ITC requires qualified expenditure certification. This shrinks the funding gap from "alarming" to "manageable but execution-dependent."

The NIST dividend/buyback restriction for 5 years is a new constraint not discussed on the earnings call. AMKR is locked out of returning capital to shareholders while burning cash on Arizona.

4. Competitive Position vs ASE

FieldValue
VariableAMKR share of advanced packaging vs ASE
ConsensusBoth grow, AMKR maintains #2
Our estimateGap widens. ASE outspending 3-4x on equipment.
DeltaNegative for relative positioning
EvidenceASX transcript (direct quote, Tien Wu): "Machinery CapEx totaled $3.4 billion. Building facilities CapEx $2.1 billion 2025, mainly LEAP services testing investment." AMKR guided $2.5-3B total CapEx, 65-70% = Arizona facility. AMKR equipment spend: ≈$750M-$1.05B. ASE $3.4B equipment vs AMKR ≈$900M = 3.8x gap on equipment specifically. ASE also acquiring Infineon and ADI factories for geographic expansion.
Base rateOSAT is oligopoly. Scale advantages compound via yield learning and equipment utilization.

Correction from prior draft: The equipment-vs-facility distinction matters. AMKR's CapEx is mostly concrete (Arizona construction). ASE's CapEx is mostly tools (machinery for LEAP). Equipment generates revenue within quarters. Facilities generate revenue in years. The competitive gap on productive investment is ≈4x, not 2x.

5. No Backlog Risk

FieldValue
VariableRevenue visibility / order backlog
ConsensusStrong demand assumed via AI tailwind
Our estimate"Absence of backlog" = genuine downside risk if cycle pauses
DeltaUnderpriced tail risk, but Q1 IS a potential trigger
Evidence10-K: "our absence of backlog and the short-term nature of our customers' commitments" and "double bookings by customers." Revenue fell 8% in 2023 when smartphone cycle weakened despite strong narrative.
Base rateOSAT revenue standard deviation ≈8% annually. A 1-quarter demand pause reprices a 34x stock violently.

Correction from prior draft: Q1 earnings (Apr 27) was dismissed as "LOW" catalyst. That's wrong. On a 34x trailing multiple with no backlog, a miss IS the downside catalyst. The seasonally weak quarter is exactly when a demand pause would surface. If Q1 misses on computing or communications, the $50 put wall breaks and put gamma cascades the stock toward $42. Right about the mechanism, wrong to dismiss the timing.


The 424B7: Kim Family Sold $487M

Filed Feb 12, 2026, three days after Q4 earnings beat.

Facts from the prospectus supplement (SEC filing, primary source):

  • Seller: Kim family (49.4% owner, Board Chairman Susan Y. Kim)
  • Shares: 10,000,000 (+ 1,500,000 greenshoe option)
  • Price: $48.75/share (13% discount to prior close)
  • Proceeds: $487.5M to Kim family. $0 to company.
  • Underwriter: Goldman Sachs & Co. LLC
  • Lock-up: Kim family 180 days (~Aug 11, 2026). Company 75 days clear market (~Apr 28).
  • Post-sale ownership: Kim family ≈112M shares (≈45%)

Goldman underwrote the Kim family sale at $48.75 and simultaneously carries a $43 Neutral rating. No insider has made a Form 4 code P (open market purchase) at any price in the reviewed period. Every real-money transaction is a sale.

From NIST CHIPS filing: AMKR agreed to dividend and buyback restrictions for 5 years. The Kim family may have concluded that if the company can't return capital to shareholders for 5 years, and the stock is at +133% on narrative, this is the liquidity window.


Positioning

Effective float:    ≈125M shares (Kim 49.4% locked, minus 10M sold)
Short interest:     5.9% of float / 2.0 DTC
Options:            P/C OI 0.59 (bullish). Max pain $40 (-21% below).
                    $50 put wall (6,669 OI near-term)
                    $55 call wall (7,048 OI June)
                    $42 put floor (6,524 OI June)

Forced buyers if right: Short cover (2 DTC, trivial), $55 call gamma (moderate). No structural bid.

Forced sellers if wrong: Put gamma below $50 (real), 1.30 SMH beta amplification (real), thin float.


Repricing Catalysts

DateEventPotential
Apr 27Q1 earnings (est $0.24)MODERATE — miss on 34x/no-backlog reprices violently
H2 2026HDFO programs enter HVMHIGH — binary validation
~Aug 11Kim family lock-up expiresMODERATE — will they sell more?
2028 H1Arizona manufacturingLOW — too distant

Kill Conditions

ConditionStatus
No delta on any variableMET — every variable within consensus range
No factor edgeMET — 43% idio, 57% semiconductor sector, no insight on SMH
No forced actorMET — no structural buyer within 6 months
No informational advantageMET — $12.6B cap, 9 analysts, +133% 1Y, all public sources

Unfilled Gaps

GapImpact
AMKR-specific HDFO tool installsCan't verify ramp vs ASE from equipment vendor data
CHIPS Act milestone progressNo milestone filings yet, first expected mid-2026
Arizona construction cost trackingNo independent data; base rate says overrun likely
Customer prepayment binding termsVerbal "loading agreements" on call, not in filings

Watchlist Triggers

Re-evaluate if ANY of these occur:

  1. Equipment vendor transcripts (May-June 2026) show AMKR-specific HDFO tool divergence from ASE
  2. Insider Form 4 code P purchase >$500K by Engel or Faust
  3. CHIPS Act milestone 8-K showing ahead/behind schedule
  4. Q1 earnings miss (would confirm no-backlog vulnerability)
  5. Kim family lock-up expiry Aug 11 — monitor for additional 424B7

Alpha Calculation

P_target:     $56.25 (consensus mean)
P_current:    $50.94
T:            1.0 year
r_f:          4.5%

Raw excess return = 5.9%
Confidence = 40% (consensus target, no differentiated view)
Edge% = 43.0% (idio component from clean regression)

alpha = 5.9% x 0.40 x 0.43 = 1.0% annualized

Position size = negligible

LR Rationale: 1.0

This is consensus confirmed, not bearish divergence.

Every bearish finding — ASE competitive gap, no backlog, Arizona cost risk — is already priced by the 6 of 9 analysts at Hold/Neutral. Goldman at $43, Morgan Stanley at $45. The Kim family secondary was a public event that moved the stock from $56 to $51. The market absorbed it.

The temptation was to call this LR 0.85 (mild bearish) based on the 424B7 and the competitive analysis. But a sell-side analyst wouldn't disagree with a single thing in this memo. That's the definition of LR = 1.0. The debate was confirmed, not resolved.

For semiconductor packaging exposure, SMH at 1.30x the beta offers the same factor return without the 14% gross margin, $3B CapEx bill, 5-year shareholder return lockout, and founding family selling into the bid.


Primary sources: AMKR 10-K (2026-02-20), AMKR 424B7 (2026-02-12), AMKR Form 4 (2026-02-18, 2026-01-05), NIST CHIPS Award page (amkor-technology-inc-arizona-peoria), ExecutiveGov (amkor-7b-expanded-arizona-chips-project), AMKR Q4 2025 transcript (2026-02-09), ASX Q4 2025 transcript (2026-02-05), CAMT Q4 2025 transcript (2026-02-18), KLIC Q4 2025 transcript (2025-11-20), MSFT FQ2 2026 transcript (2026-01-28), GOOG Q4 2025 transcript (2026-02-04), yfinance market data.