$AMD · Advanced Micro Devices

Thesis

Up +22% in 10 trading days into a print where options imply an 8% move and RSI sits at 72 — a clean beat-and-raise is the floor expectation, not the upside surprise. PLTR last night ran the same setup (clean beat + 10% FY guide raise) and still fell -7%. AMD needs a discrete MI350 hyperscaler-customer commitment AND gross-margin-expansion guide to break the priced-in ceiling.

What would flip it

Lisa Su names a MI350 hyperscaler customer (Microsoft/Meta GPU commitment) on the call — that's the bull-case asymmetry.

The market's narrative

AI-accelerator pure-play that has rallied +22% in 10 trading days into the print on MI350 production-ramp anticipation; Cathie Wood liquidating a $70M position into the move and options pricing an 8.1% implied — the bar for ANY positive surprise to break the ceiling is now extreme.

Where the Street may be wrong

• Data center segment is highly likely to print $5.7-5.8B vs $5.56B Street consensus — but at +52% YoY growth, that's still right on the GOOGL-Cloud-+50% growth-justifies-spend threshold; not above it.
• MI325 → MI350 transition timing risk is asymmetric — every quarter of MI350 production-ramp delay vs Q3 schedule is a -2-3% multiple compression event, regardless of headline beat.

Peer read: PLTR last night — clean beat + 10% FY guide raise + Q2 above Street and STILL fell -7.04% on options-implied 10% pre-priced. AMD setup is structurally identical: hype_high + options-implied >8% + RSI 72 + just rallied +22%.

Reasoning

• RSI 72.78 + price +22% in 10 trading days + ATR 16.44 = textbook over-extended runup into print; entire beat-and-raise scenario is mathematically priced in.
• Options-implied 8.1% move + Polymarket beat-odds elevated = PLTR-template sell-the-news regime where clean beat-and-raise is the FLOOR not the ceiling.
• Data center revenue likely beats ($5.7B vs $5.56B Street), but +52% YoY is right at the GOOGL Cloud growth-justifies-spend threshold — needs to be >55% to unlock the multiple-expansion bid; high $50s % is the FLAT zone.
• Gross margin guide stays at 55% (no expansion) — AMD needs to walk margins UP to 56-57% range alongside MI350 ramp commentary to break the priced-in pattern; if margins stay 55%, the Mag-7-capex-penalty narrative resurfaces.
• Lisa Su has historically been measured on hyperscaler customer-name disclosures; this print needs a discrete MI350 hyperscaler announcement (Microsoft/Meta/OpenAI committing GPU counts) to deliver the upside-amplifier the priced-in setup demands.