Situation

Harmony Biosciences ($27.50, $1.6B market cap) is down 25% in a week. Two things happened simultaneously: Q4 EPS missed by 46.5% ($0.57 vs $1.07), and the bench trial for the last remaining ANDA patent challenger (AET Pharma) concluded February 17-19. Market is pricing this as a death sentence. It isn't — but it's not a gift either.

The Q4 Miss Is Not What It Looks Like

Full-year numbers tell the real story: revenue $868.5M (+21.5%), operating cash flow $348M, net income $158.7M. The business is growing and generating enormous cash.

Q4 got hit by two non-recurring items:

$39M in ANDA settlement payments. Five of six generic challengers settled their patent disputes in 2025, and the legal fees and settlement amounts were booked to G&A. G&A jumped $42.2M year-over-year — almost entirely from patent litigation costs. These settlements are done. They don't recur.

State deferred tax asset remeasurement. Effective tax rate spiked from 24.1% to 26.2%, of which 3.5% was a one-time state DTA adjustment. This too is done.

Strip out those items and normalized EPS is roughly $3.25-3.50 for the year, putting the trailing P/E at ≈8x. The street will revise 2026 estimates down because they're anchoring on the headline miss, but the underlying earning power hasn't changed.

The Patent Binary

WAKIX's core patent protection rests on two patents: the '947 (expires September 2029) and the '197 (expires March 2030). Seven generic companies filed ANDAs challenging these patents. Here's what happened:

Five of six settled to July 2030. The sixth (MSN) settled with confidential terms — a genuine wildcard. Only AET went to trial.

The settlement timing is the most important signal in this entire analysis. Four companies settled in January 2026 — weeks before the February 17 trial date. Critically, all of these settlements occurred after the March 2025 claim construction hearing, meaning every challenger had full visibility into how the court interpreted the patent claims before deciding whether to fight or fold.

They folded.

When generic companies see a favorable claim construction and believe the patents are vulnerable, they go to trial. When they see an unfavorable construction and believe they'd lose, they settle for a license date near patent expiry. Four companies seeing the same court ruling and all choosing settlement over trial is a strong signal about patent validity.

The question is whether AET has something different — a stronger non-infringement argument, a better invalidity theory, or just more appetite for risk. They're the outlier. The market is pricing AET's outlier status as if it's dispositive.

Insider Behavior: The Counter-Signal

This is where the thesis gets complicated — and where my initial analysis was wrong.

I initially reported "coordinated insider buying" on January 23. The yfinance data showed CEO Dayno and CFO Kapadia acquiring shares. Looked bullish.

The actual Form 4s tell a different story. Transaction code analysis:

CEO Dayno (December 12, 2025): Exercised 20,000 stock options at $8.22 (Code M), immediately sold all 25,933 shares at $40.11 (Code S). Ended with zero direct shares. The options expired November 2027 — he had two years left and chose to exercise and sell everything rather than hold through the patent verdict.

CFO Kapadia: Sold to zero direct shares three times in 90 days:

• November 21: Sold 3,427 shares at $35.00 → zero
• December 5: Exercised 20,000 options at $29.03, sold all 20,000 at $39.54 → zero
• January 15: Sold 20,961 shares at $35.92 → zero

The January 23 "acquisitions" were annual equity compensation grants (Code A) and RSU vestings (Code M) — standard compensation, not open market purchases. The CFO then sold 3,746 of his newly vested shares three days later on January 26.

There are zero Code P (open market purchase) transactions in any HRMY insider filing. Not one.

The people with the deepest visibility into the patent litigation — CEO and CFO — maintain zero voluntary stock ownership. They exercise and liquidate immediately, every time. The CEO didn't even wait for his options to get closer to expiry.

Fidelity (FMR LLC), the largest institutional holder, cut their position 19% between November 2025 and February 2026 (6.7% → 5.4%, selling ≈734K shares). The smart money is trimming.

What The Options Market Says

The stock price implies roughly 32% probability that the patents hold (using $48 bull / $18 bear scenario values). July 2026 options deltas give a more granular view:

• ≈20% probability of full recovery to $50+ (patents definitively hold)
• ≈37% probability of recovery above $35 (positive signals, maybe no verdict yet)
• ≈31% probability of further decline below $25 (negative signals or early generic path)

ATM implied volatility is 69-87% across expirations versus 47% historical — the market expects a large move but is positioned heavily bearish (P/C ratio 6.92 on March options). One contrarian signal: 752 July $50 calls traded on 296 OI today (2.5x unusual activity). Someone is making a directional bet on a positive verdict.

Valuation Floor

Even in the bear case, this isn't a zero. Net cash is $717M ($882.5M cash and investments minus $165M debt). That's $12.40/share — a hard floor at 55% below current price. The company generated $348M in operating cash flow in 2025 and has a growing pipeline (EPX-100 for Dravet/LGS in Phase 3, pitolisant GR with PDUFA Q1 2027, pitolisant HD with PDUFA 2028).

The next-gen formulations (GR and HD) have utility patents extending to mid-2040s. If the original WAKIX patents fall, generic immediate-release pitolisant enters the market, but the new formulations remain protected. The GR formulation eliminates the need for dose titration — 100% of patients in the dosing optimization study tolerated the therapeutic dose from day one. That's a real clinical differentiation. Whether it's enough to defend revenue against a cheap generic is the classic lifecycle management question — mixed track record (Nexium worked, Nuvigil didn't).

At EV/OCF of 2.5x and EV/Revenue of 1.0x, the market is pricing roughly 4 years of cash flows and then nothing. If patents hold, that's absurdly cheap. If generics enter in 2027, it's about right.

Where My Evidence Nets Out

The settlement timing pattern (LR 3.0) is the strongest signal in the analysis — it's specific, primary-source, and under-analyzed by the market. But the insider liquidation pattern (LR 0.5) partially neutralizes it. The people with the most information aren't acting like the patents are safe.

These signals are in genuine tension. The settlement pattern says the patents are probably valid. The insider behavior says the people closest to the situation don't want the equity exposure. Both can be true simultaneously — the patents might hold but the insiders might be hedging career risk, or the settlements might reflect economic convenience rather than patent strength.

Scenario Framework

EV: $33.45 (+21.6% from current)

The positive EV looks attractive until you run it through the edge filter. My edge is concentrated in two places: recognizing the Q4 miss as non-recurring (≈80% edge on 15% of forward variance) and the settlement timing signal (≈30% edge on 70% of forward variance). After edge-weighting, the alpha compresses to roughly 1-2% annualized — a starter position at best.

The Honest View

This is a doorway state. Two coherent interpretations of the evidence:

Interpretation A (bull): The mass pre-trial settlement pattern, combined with the new pediatric cataplexy approval enabling exclusivity, means patents hold. The Q4 miss is a one-time settlement expense that masks a $870M revenue / $350M OCF business trading at 2.5x cash flow. Insiders are selling because they're taking compensation off the table, not because they know the patents will fail. The stock doubles within 12 months of a favorable verdict.

Interpretation B (bear): AET went to trial for a reason — they have a differentiated argument the settlers didn't. The CEO exercised options with two years remaining rather than hold through the verdict. The CFO sold to zero three times. Fidelity is cutting. The people who know the most are voting with their feet. When the verdict comes, the patents fall, and generic entry accelerates.

I can't collapse this superposition with the evidence available. The settlement pattern gives me a modest tilt toward the bull case (combined independent LR of 0.6 adjusts my posterior to roughly 40-45% P(bull) from the market-implied 32%). But that's not enough alpha to size meaningfully after edge adjustment.

Not a position. A monitoring situation. The catalysts that would create actionable signal:

1. AET verdict (filed as 8-K) — collapses the binary
2. Pitolisant GR PDUFA Q1 2027 — validates franchise extension
3. Buyback execution ($150M authorized, nothing used) — would signal management confidence
4. Pediatric exclusivity grant — locks in the 2030 settlement dates

The first one matters most. Everything else is noise until the patent question is resolved.