Verdict: KEEP at 1.04% benchmark weight.

Healthcare name in a tech index. Factor mismatch screams remove but company is executing well, trailing alpha is positive, and removal would be a 7th correlated factor short with zero IC. The two things consensus is potentially lazy on — Prolia biosimilar cliff velocity and IRS total exposure scope — don't create a trade at this weight.

Factor Profile

iev regress AMGN (n=250 trading days)

SPY          beta=+0.13      3.1% of variance
XLV          beta=+1.20     52.2% of variance
MTUM         beta=-0.21     -3.7% of variance
Idio                        48.5% of variance

alpha=+19.6%  sigma_idio=20.3%  R-sq=51.5%
Orthogonal Sharpe: 0.97

52.2% of AMGN's variance comes from XLV. QQQ beta 0.36. Anti-momentum tilt (MTUM beta=-0.21). Idio variance 48.5% — well below the 75% target. In a QQQ rally driven by tech/momentum, AMGN systematically lags. In risk-off/defensive rotation, AMGN outperforms. This is factor exposure, not stock-specific weakness.

Fundamentals

Earnings history (4 consecutive beats, declining magnitude):

Next: April 30, 2026. Consensus $4.76.

Bull/Bear Tension

Growth engines (accelerating):

• Tezspire: $1.5B (+52% YoY), Q4 exit rate $1.9B annualized. CRSwNP approved Oct 2025. COPD and EoE Phase 3 ongoing.
• Repatha: $3B (+36% YoY). Volume-driven. MK-0616 (Merck, oral PCSK9) threat is real but 12-18 months from launch.
• EVENITY: $2.1B (+34% YoY). >60% bone builder market share. Dual mechanism moat.
• BLINCYTO: $1.6B (+28% YoY).
• Biosimilar portfolio: ≈$3B annualizing (+50% YoY). PAVBLU (Eylea), WEZLANA (Stelara), BEKEMV (Soliris).

Legacy erosion (accelerating):

• ENBREL: $2.2B (-33% YoY). IRA/340B/Medicare Part D price destruction. Volume +4% but price -36%. Terminal decline.
• Prolia: $4.4B (+1% YoY). US $2,978M (+3%), ROW $1,436M (-4%). Patents expired Feb 2025 (US), Nov 2025 (EU). 10-K warns "accelerated sales erosion" in 2026.
• XGEVA: $2.1B (-6% YoY). Same patent cliff as Prolia.
• Otezla: $1.2B impairment charge. Medicare price setting 2027.
• Combined at-risk: ≈$10B in declining/eroding revenue (≈28% of total product sales).

Revenue trajectory:

Pipeline (outside 15-week window):

• MariTide (obesity): 6 Phase 3 studies fully committed. CWM studies fully enrolled per Q4 call. Data readouts late 2026 at earliest (52-68 week treatment period). $4.3B later-stage R&D spend.
• Two pipeline kills in Q4 2025: bemarituzumab (gastric cancer) and rocatinlimab (returned to Kyowa Kirin). Reduces optionality.
• Biosimilar pipeline: ABP-206 (Opdivo) and ABP-234 (Keytruda) enrollment complete. Future contributors.

FY2026 guidance (Q4 2025 transcript):

• Revenue: $37.0-38.4B
• Non-GAAP EPS: $21.60-$23.00
• Q1 2026: "lower mid-single-digit YoY growth," lowest margin quarter (≈42% non-GAAP OM)
• ≈$250M inventory build from 2025 "potentially impact first-quarter sales"
• Full-year non-GAAP OM: ≈46% (flat vs FY2025)

Balance sheet (Dec 31, 2025 10-K):

• Debt: $54.6B ($50.0B noncurrent + $4.6B current), down from $60.1B after $6.0B retirement
• Feb 2026: $4.0B issuance (4.20-5.65%, 2031-2056) refinances $4.6B current maturities
• Cash: $9.1B. OCF: $10.0B. Capex: $1.9B (+69% YoY, manufacturing expansion)
• Ratings: BBB+/Baa1/BBB+ (investment grade, stable)

What Consensus Thinks

Thirty-four analysts independently arrived at the same non-opinion. Mean target $350.55 (current price). Median $350.00. The Street says: "Fair value. No catalyst. Wait for MariTide."

Direction of travel is toward neutral. Post-Q4 earnings actions (Feb-Mar 2026): 6 of 10 are holds/neutrals. One downgrade (Freedom Broker, Buy to Hold). Zero upgrades. Only Piper Sandler (Overweight $432, the MariTide optimist) and Argus (Buy $400) have conviction. Morgan Stanley is the outlier bear at $309.

Forward EPS consensus of ≈$23.28 exceeds management's guide high of $23.00 — Street is pricing beat continuation. With declining beat magnitude (15% to 12%) and structural Q1 headwinds, this is mildly aggressive.

The consensus narrative: growth brands replacing legacy erosion, race tightening, MariTide is the swing factor that won't generate revenue for years. 14.98x forward is the standard pharma-compounder multiple. Nobody expects upside, nobody expects collapse.

Where Consensus May Be Wrong

1. Prolia Biosimilar Cliff: Step Function, Not Slope

This is the most interesting finding. The Street is extrapolating from 2025's gentle trajectory — Prolia ran flat at ≈$1.1B/quarter all year despite 3 biosimilars in market. Q1 2026 is materially different.

The competitive landscape shifted in Q1 2026. The 10-K's patent litigation section (In Re: Denosumab Patent Litigation) reveals:

• 3 biosimilars already in market through 2025
• Hikma/Gedeon Richter: Settlement consent judgment Nov 24, 2025 — enjoined from selling "before January 1, 2026." 4th entrant launched Jan 1.
• Amneal/Mabxience: Defendants responded to complaint Jan 9 and Jan 23, 2026 — preparing for launch
• Dr. Reddy's/Alvotech: Filed counterclaims Jan 9, 2026 — preparing for launch
• Alkem: Filed amended answer Feb 12, 2026 — preparing for launch

Four active, three more pressing to clear the runway. And EU patents expired Nov 2025, so ROW erosion (already -4% in FY2025) accelerates simultaneously.

Management's own language shifted: "competitive dynamics in line with expectations" (Q3 2025 call) became "expect accelerated sales erosion" (10-K, Feb 13, 2026). They upgraded the warning between the call and the filing.

Transcript trajectory confirms the setup:

• Q1 2025: Prolia +10% YoY (+13% volume) — pre-biosimilar
• Q2 2025: Prolia -4% YoY — first price concessions
• Q3 2025: Prolia +9% YoY — "3 biosimilars launched, dynamics in line"
• Q4 2025: ~flat (FY $4,414M implies ≈$1.1B/quarter throughout)
• Q1 2026: 4+ biosimilars, "accelerated erosion," EU patents expired

If Street models ≈$1.0B for Q1 Prolia (gradual -5% sequential) but actual is ≈$0.90B (-18% YoY), the delta is ≈$100M revenue, or ≈$0.12 EPS headwind not in the $4.76 consensus (2.5% miss from Prolia alone).

Reference analog: Neulasta US went $710M to $318M (-55%) in one year once biosimilar traction hit. Prolia has higher switching costs (6-month injectable, elderly patients, physician inertia) so the cliff is slower. But the direction is the same.

2. IRS Total Exposure Is 2-3x the Headline

Everyone cites $10.7B ($8.7B tax + $2B penalties). That's just 2010-2015. The 10-K buries the rest:

From the 10-K: "We expect that the IRS will begin its audit of 2019-2022 in the first half of 2026, and we believe that it may seek to continue to audit similar issues related to the allocation of income between the United States and our foreign jurisdictions."

Amgen's pre-tax income grew from ≈$5B/yr (2010-2012) to ≈$10B/yr (2019-2022). If IRS applies the same profit allocation methodology to 13 years of growing profits, total exposure could be $20-25B+ (10-13% of market cap). Net after $3.1B repatriation credits, still substantial.

The behavioral tell: Management has never discussed the IRS dispute on an earnings call. Searched all 7 available transcripts — zero mentions. Analysts don't ask. The trial lasted 2.5 months (Nov 4, 2024 to Jan 17, 2025) with extensive post-trial briefing and oral argument. That's not a nuisance case.

Consensus assigns ≈0% probability based on "firmly believe without merit." If P(adverse) = 20% on just the 2010-2015 period, expected loss = 20% x $7.6B net = $1.5B = 0.8% of market cap not in the price. Include 2016-2022 and it's 2-4%.

Not actionable in our window. Decision is H2 2026+, likely years of appeals. But it's a real tail risk the market is pricing at zero.

Insider Activity

Mar 3-4, 2026: Annual compensation grants and conversions only. Bradway (CEO): 42,657 shares awarded (≈$14.9M), 119,782 converted (≈$18.7M). No open market transactions. Annual comp cycle noise, LR 1.0.

Q1 2026 Earnings Preview (April 30)

Management pre-guided Q1 as weakest quarter:

• "Lower mid-single-digit YoY revenue growth"
• Lowest margin quarter (≈42% non-GAAP OM vs ≈46% full-year)
• ≈$250M inventory destocking headwind
• Enbrel/Otezla "additional impact biosimilar competition Q1"
• Prolia "accelerated sales erosion" beginning

Consensus $4.76. Our prediction: 50% beat (coin flip). Beat magnitude has been declining (15.0% to 14.1% to 12.5% to 11.9%). Management has a sandbagging tendency — beaten every quarter since Q1 2025 by double digits — but Q1 has real structural headwinds (Prolia cliff acceleration, $250M destocking, ENBREL continued decline). The 4th denosumab biosimilar entering Jan 1 is a specific, quantifiable headwind that may not be fully in the number.

A Q1 miss wouldn't change the KEEP verdict unless accompanied by a full-year guide-down on growth brands. Pre-guided seasonal softness is not idiosyncratic weakness.

The Filtration Decision

The case for removal (factor mismatch):

• QQQ beta 0.36, XLV beta 1.20 — healthcare stock in tech index
• 48.5% idio variance — majority is factor noise vs QQQ
• Anti-momentum (MTUM beta=-0.21) — systematically lags in momentum rallies
• In a tech bull run (most likely QQQ regime), AMGN drags

Why removal fails the burden of proof:

1. Zero IC, correlated risk. The book already runs 6 factor-mismatch shorts (MAR, CMCSA, CSX, SBUX, CTAS on anti-momentum; CEG on utility). Adding AMGN is a 7th correlated factor bet at zero incremental IC. Book Sharpe declines.

2. Zero relative underperformance. AMGN vs QQQ = ≈0 bps over 1 year (both +17.2%). The factor mismatch creates tracking error but XLV has kept pace with QQQ over this period. No directional drag to exploit.

3. Positive idiosyncratic signal. Trailing alpha +19.6% vs factor model, orthogonal Sharpe ≈0.97. Growth brands adding ≈$3.5B incremental annually. 4 consecutive earnings beats. The company-specific direction is up, not down.

4. RSI 20.4 oversold into earnings. The -8.1% monthly decline is 100% XLV factor. Shorting into deeply oversold conditions ahead of a potential 5th consecutive beat is poor risk/reward.

5. 1.04% weight caps the prize. Maximum filtration alpha from removal: ≈10-15 bps (if AMGN underperforms by 10-15%). Expected alpha near zero. The juice is not worth the squeeze.

Decision

KEEP at 1.04% benchmark weight.

Monitoring triggers:

• April 30 Q1 earnings: If miss + guide-down on growth brands, reconsider. Watch Prolia sequential decline rate specifically — if >10% sequential, the cliff thesis is confirmed.
• H2 2026 IRS decision: If adverse, reconsider (likely post-window but could leak via legal filings)
• Prolia quarterly trajectory: If >20% YoY decline in Q1, the biosimilar cliff is faster than anyone models

What would change the verdict: Only idiosyncratic negative — growth brand deceleration, material pipeline failure, or IRS adverse. Factor underperformance alone is not grounds for removal. The Prolia cliff and IRS exposure are real risks but neither is actionable at 1.04% weight in a 15-week window.

LR Signal

Memo-level LR: 1.0 (Consensus). Two findings add nuance — Prolia biosimilar cliff velocity (step function from 3 to 4+ entrants in Q1) and IRS total exposure scope (2010-2022, not just 2010-2015) — but neither creates a divergence from the consensus KEEP. The filtration decision rests on portfolio construction (correlated factor short book at capacity), not AMGN-specific edge. Nothing shifts the prior.