Setup. Aehr Test Systems (AEHR) makes semiconductor burn-in test equipment and has ridden the AI infrastructure narrative from ≈$8 to $90 over twelve months (+1032%) on a Q3 FY26 bookings inflection. On April 20, 2026, the company filed an Item 8.01 8-K disclosing it fully utilized a $60M at-the-market equity offering in nine calendar days — the second consecutive $60M ATM to run dry.

What the filing says. Two paragraphs. The ATM, announced via prospectus supplement on April 8, completed $60.0M of sales by April 17. No press release, no business commentary — signed by CFO Chris Siu. At the observed pace AEHR sold $7.5-8.5M of stock per trading day, roughly the maximum rate the market could absorb without moving price. The prior ATM raised $19.6M across nine months before being rushed to completion in March 2026. Combined raise across consecutive ATMs: $120M+. The underlying S-3 shelf (File No. 333-282661) registered $100M of securities — the arithmetic implies the shelf is near-exhausted or already over-drawn.

The valuation is through the ceiling.

Stock is $90. That is above the full-bull scenario — a 10x P/S multiple on $250M FY27 revenue, itself a 5x ramp from the "high side of $45-50M" FY26 guide. Peer semi-equipment names trade 1.7x (COHU), 5x (FORM), 5.7x (TER). Four analysts cover AEHR, all rated Hold, mean price target $63.67 — 29% below spot.

The edge. Market-implied versus our probability:

The widest gap is fundamental, not pathwise. Options already price some reversion (IV 130%, OTM put skew +24.9% over ATM vol, Jun max pain $60). This is the entire trade.

Mechanical risks to trade expression — constraints on how to play it, not on the thesis.

• Squeeze persistence. 14% short interest, 2.4 days-to-cover, retail call chasing ($125 strike volume ran 83x open interest). Stock can stay bid above fundamentals for a quarter. → defined-risk structure only.
• IV premium. 130% implied vol means puts pay peak volatility. → expiration must be long enough for delta to overcome premium decay.
• Borrow cost. Punitive at +1032% 1Y. → equity short is wrong expression.

Catalysts.

• Q4 FY26 earnings, June 2026 (date TBD). Revenue print against ≈$20M internal expectation, versus a price that requires >$30M. Binary.
• Shelf refill: new S-3 registration or third ATM, imminent. The $100M shelf is near-exhausted after $120M of consecutive ATM draws. A new S-3 filing within 30 days of this one would document management's dilution intent across three distinct filings — signal amplification.
• FY27 guidance, September 2026. First official revenue guide — the $200M+ implied number becomes defendable or not.

What would change our mind.

• Insider open-market purchase (Form 4 code P) at any price — informed-party signal reversal.
• Q4 FY26 revenue > $25M combined with FY27 guide > $150M — ramp real, valuation defensible.
• Hyperscaler follow-on order announced at $50M+ within 30 days. (Management disclosed expectation; magnitude would determine whether the ramp narrative is real.)
• IV collapse to <80% without price movement — consensus has shifted bearish and the setup is crowded.

Trade structure. Sep 2026 $70 puts offer a defined-risk expression that covers Q4 FY26 earnings (June), FY27 guidance (September), and the dilution-drag window in one expiration. The 22% OTM strike aligns with the 60-90 day probability-weighted target of $74. Defined risk bounds the squeeze and thematic tail; equity short is structurally wrong given borrow cost at this return.

Evidence

Independent units only. Multiple filings on the same event count as one signal.

LR signal: 0.65. Bearish. The four bear signals above (ATM, insider, operating loss, peer valuation) treated as independent compound to ≈0.05 — which would be an irresponsible memo LR because they are not independent: all four are downstream expressions of "a company priced as growth while insiders and the capital structure say distress." They are correlated evidence of one underlying pattern. The combined LR is further bounded by the hyperscaler follow-on order tail — a $50M+ contract announcement would invalidate three of the four signals simultaneously, a real ≈15% probability. 0.65 reflects the net of both adjustments and matches the calibration band for "market hasn't fully priced" without performing conviction.