AMD

Current Thesis

The clean $360 retrace we waited six straight DEFERs for (2026-05-23) HELD — and then ripped +45% to $521 (6/4 premarket $521.82) while we sat out. That is exactly the Momentum-Realignment alpha leak this book exists to avoid: we deferred an ACCELERATING, cluster-confirmed name for "RSI-high / no-pullback" and it ran. Lesson logged. BUT the trade has now matured into late-mania. AMD prints new ATHs at RSI 75.34 (6/3), Cathie Wood dumped 56M shares to chase NVDA (6/3) "capitulating after missing the rally," and Broadcom's "repeat-not-raise" AI guide (6/4) just triggered complex-wide profit-taking. The fundamental narrative is still ACCELERATING (Intel confirmed surging agentic-AI CPU demand 6/3, independently validating AMD's EPYC TAM call; Helios rack-scale + Computex showcase 6/2). But price is +45% in 8 sessions, stretched far above the 20-EMA, into distribution tells. A FRESH long at $521 is chasing a parabola into a blow-off — not a risk-defined entry. Posture: LOW-conviction watch; re-arm only on a pullback to the ~$420 breakout shelf that holds.

Bull Case

• Agentic-CPU TAM validated by a COMPETITOR: Intel CEO Lip-Bu Tan reports "surging CPU demand fueled by rapid growth of agentic AI" (2026-06-03) — independent confirmation of AMD's $120B AI-CPU / 1:1 CPU-to-GPU thesis (2026-05-07). EPYC is the under-modeled lever distinct from the crowded GPU trade.
• Helios rack-scale platform + $10B+ Taiwan commitment for 2H-2026 deployment (2026-05-21) — moves AMD from chip vendor toward full-system NVL-rack competitor.
• Computex showcase with Supermicro of next-gen AI infrastructure (2026-06-02); SMCI +5.5% premarket same day on persistent AI demand — keeps AMD central to the buildout narrative.
• AMD bets on the next AI bottleneck (networking) as DriveNets raises $410M (2026-06-02) — positioning beyond compute into the interconnect layer.
• Q1 2026 print (2026-05-05): rev $10.3B (+38% YoY), Data Center $5.8B (+57% YoY), EPS $1.37, record FCF $2.566B, +170bps GM YoY; Q2 guide $11.2B vs $10.5B consensus — fundamentals validated the re-rating.
• TD Cowen Buy PT $500 (2026-05-06) was the most bullish mainstream target — price has now EXCEEDED it ($521), forcing the next round of sell-side upgrades (narrative still catching up).
• Sector tape confirms demand intensity: MU +940% rally / +260% YTD, Dell +260% YTD (2026-06-02), SMCI breakout — the AI-infra complex is being repriced wholesale.

Bear Case

• Cathie Wood / ARK persistent distribution: sold $56.2M (6/2), dumped 56M shares to load NVDA (6/3) "capitulating after missing the rally," sold another $40.6M (6/4) rotating into GOOGL/META/BABA. A structural AI bull repeatedly selling AMD into strength = name-brand distribution.
• Broadcom "repeat-not-raise" AI guide (2026-06-04) triggered AMD -3.82% premarket to $521.82 and complex-wide profit-taking — the first crack in the serial guide-up cascade that powered the move.
• Overbought / stretched: RSI 75.34 (6/3), new ATH, +45% in 8 sessions, far above 20-EMA — textbook late-mania extension. Repeated "investors lock in gains after massive rally" headlines (6/2, 6/4).
• Retail rotation tell: May's most-searched tickers were Micron, SanDisk, Qualcomm (6/2) — retail crowding the memory leg; Webull CEO flagged retail "moving beyond AMD/MU/NVDA" (5/22). Peak-retail-sentiment marker.
• ASIC / custom-silicon encroachment (Broadcom) on the GPU TAM remains the structural overhang on the crowded MI-series leg.
• Group-wide mean-reversion risk: SOXX was 60% above its 200-DMA / Burry chip short (5/6); the parabola has only stretched that further. SMCI +145% then falling (6/3) shows how fast these unwind.

Setup & Price Structure

• Last dossier setup (retrace to ~$360 shelf, 5/23) RESOLVED BULLISH — held and launched. We missed it. The $360→$421→$360→$521 sequence is now a completed momentum leg, not a pending entry.
• Current: $521.82 (6/4 premarket, -3.82%), new ATH / 52-wk high, RSI 75.34 (6/3), extended far above the 20-EMA.
• No risk-defined long exists here: ATH = no overhead supply but also no support shelf beneath for a stop. A stop at any sane level (e.g. ~$470) sits >9% away and in air.
• Breakout shelf = ~$420 (mid-May high cleared on this leg). The only clean re-entry is a pullback to $420–440 that HOLDS (higher low, daily close back above 20-EMA) — risk-defined against a sub-$410 stop.
• Trim discipline if held (archetype 2): weekly close below 20-EMA, OR a daily close back below the $420 shelf = momentum leg broken → exit.
• Do NOT average down and do NOT chase. A fresh entry at [entry redacted] violates the beginner-trap rule: peak sentiment + stretched + no NEW catalyst beyond the one that already ran.

Catalyst Calendar (next 30 days)

• ~2026-06-25 (est.): Micron fiscal Q3 print — sector read-through; "can the 940% rally continue ahead of Q3 earnings?" (6/3). A MU miss/guide-down would hit the whole memory/AI-infra complex including AMD.
• Computex 2026 aftermath (early June, ongoing): MI355X benchmark / rack-scale detail follow-through from the 6/2 Supermicro showcase.
• NVDA narrative encroachment ongoing: Jensen Huang pitching PC reinvention + CPU market (6/3) — headline/competition risk for AMD's CPU leg.
• No AMD-specific binary in the 30-day window: next AMD print (Q2 2026) is ~2026-08-05 (est.), outside the window. Catalyst risk is therefore sector-driven (MU, BRCM read-through), not company-specific — which means the next 30 days are about whether the complex holds, not about an AMD-specific re-rating event.

What Would Change Our Mind

• Re-arm to a risk-defined LONG: pullback to the $420–440 breakout shelf that HOLDS as a higher low (daily close back above 20-EMA) with narrative intact → MEDIUM/HIGH probe against a sub-[entry redacted] stop.
• Upgrade conviction if: the agentic-CPU leg gets a HARD order/booking datapoint (named hyperscaler EPYC commitment) rather than TAM rhetoric, AND the complex stops the distribution (ARK/Cathie Wood stop selling).
• Confirm the bear / stand fully aside if: daily close below the $420 shelf (parabola failed), OR MU Q3 guides down (sector demand crack), OR a second BRCM/hyperscaler guide-repeat confirms the cascade is rolling over.
• Hard invalidation of the fundamental leg: Q2 (Aug) revenue guide below $11B, or Data Center YoY growth decelerating below ~+40%.

Correlation Notes

• Tightly correlated AI-infra cluster: NVDA, MU, AVGO/BRCM, DELL, SMCI, INTC. AMD is now largely a high-beta expression of this complex — the BRCM guide-repeat (6/4) hit AMD directly.
• Held-book overlap: prior DEFERs flagged AMD as a "correlated duplicate of held NVDA cluster." A fresh AMD long would stack AI-semi beta on existing exposure, not diversify it — size accordingly.
• Distribution leader to watch: ARK/Cathie Wood rotating AMD→NVDA→GOOGL/META is a cross-name tell for the whole leg's late stage.
• Sector mean-reversion beta: SOXX 60% above 200-DMA (5/6); any complex-wide unwind takes AMD disproportionately given its +45%/8-session extension.