1. 1
   TSM Strong Add A+ / A+ / A · PEG 1.23 · entry $360–395

   Foundry monopoly on leading-edge nodes — Apple, NVDA, AMD all anchored. +41% revenue growth at PEG 1.23 is the cleanest scale-vs-multiple combination in the table. Current $400 sits marginally above zone; stage in tranches now, accelerate on any pullback below $390.
2. 2
   META Strong Add A / A / A · PEG ~0.8 · entry $580–630

   Cleanest large-cap setup: 22× fwd P/E, +33% growth, A across all three dimensions. AI capex generates measurable ROI — Reels CTR uplift and ad targeting improvements are demonstrable, not aspirational. Current $617 sits mid-zone — the highest-probability "do something now" name in the universe.
3. 3
   CRM Strong Add A / A- / A · PEG 0.87 · entry $175–190

   13.8× fwd P/E for a category-leading enterprise SaaS franchise — value setup masquerading as software. Agentforce ARR disclosure (late May) is the catalyst. Asymmetric: even modest Agentforce traction unlocks multiple expansion; if it disappoints, you own a $180B FCF compounder at sub-15× earnings. In zone.
4. 4
   COHR Strong Add A / A / B+ · PEG 0.9 · entry $300–330

   Optical content per Blackwell rack (800G → 1.6T) is a structural unit story, not a multiple story. Each NVDA shipment carries more COHR dollars regardless of optical pricing. Direct beneficiary of the same Q2 NVDA print that risks priced-in chip names. In zone $300–330.
5. 5
   AVGO Add (pullback) B+ / A / A · PEG ~1.0 · entry $400–430

   Custom XPU thesis (Google TPU, OpenAI Jericho) is real and accelerating, but 37× fwd P/E demands discipline. Current $420 is at the top of the zone — initiate small now, reload aggressively only on retest of $380–400. Aug Q3 print is the validation event.

• !AI cohort hit record highs week of 8 May. Cohort drawdowns of 15–20% historically follow record-print weeks. Stage entries over 6–8 weeks; never lump-sum at the top.
• !Memory cycle is late. MU +68% YTD into 24 Jun print; SNDK +251% revenue growth. Don't initiate new memory exposure. Cyclical margin reversion is a when-not-if risk.
• !Pre-earnings asymmetry is bad. Implied moves on NVDA (20 May), MU (24 Jun), hyperscalers (late Jul) are 8–12%. Resist adding inside the 5-day pre-print window — better risk/reward post-print, off the gap.
• !Concentration is a hidden risk. Closing the active-picks gap by adding two more software names creates correlated AI-capex exposure. Spread across foundry / optics / software / power.
• !Insider clusters worth respecting. PLTR insiders sold $432.9M in 90 days incl. Thiel 1.9M shares. Treat as a hard signal, not noise — they have information you don't.
• !"In entry zone" ≠ "must buy now." The zone is permission, not obligation. If verdict + flow + macro all line up, take it. If two of three are weak, wait for confirmation.
• !Single-stock cap 20%. Hard rule. MU at +114% unrealized is the closest current breach risk — let it run, but the next add goes elsewhere.

• ✗Chase MU above $760. Stock +68% YTD at 52W highs ($818 intraday 2026-05-12); consensus PT $483–533 below spot. Wait for the 24 Jun Q3 print; consider trim trigger at $850–900.
• ✗Add PLTR above $130. 88× fwd P/E into Anthropic's $1.5B enterprise-services JV launch (4 May 2026), $432.9M insider sales in 90 days incl. Thiel 1.9M shares — asymmetry is wrong. Reload zone $100–115.
• ✗Add NVDA above $225 ahead of 20 May. Goldman says bar is high. Wait 7 trading days.
• ✗Treat NEE as an AI growth story. It's a utility with AI optionality. Add only for income and power hedge, not AI growth.
• ✗Add a second memory/storage name (SNDK, WDC, STX) on top of MU — duplicative exposure.
• ✗Deploy all capital at once. AI cohort hit record highs week of 8 May. Stage over 6–8 weeks.
• ✗Add Bloom Energy (BE) at $258+. Never GAAP-profitable, ~175× fwd P/E, AI fuel cell thesis is aspirational not demonstrated. Early-cycle holders have a cost-basis cushion that new entrants do not — wrong asymmetry at current price.
• ✗Add ARM at current price ($213+). PEG 4.87 — worst in the AI-chip basket. 11.46% short interest. Only 7.45% institutional ownership on a $226B mkt cap. NVDA at 25× fwd P/E / PEG 0.64 is a strictly better AI chip bet.
• ✗Chase AMD above $420. Up 327% in 52 weeks. Meta 6GW and OpenAI 6GW MI450 commitments are real, but the post-earnings spike has fully priced those wins. Wait for a pullback to $300–340 before initiating.
• ✗Buy PSTG (Pure Storage). Single Meta customer dependency ($30M license at 90% margin) with a credible Kerrisdale short thesis arguing QLC is not broadly price-competitive. Better memory/storage exposure through MU.
• ✗Buy WDC or STX at peak. WDC +845%, STX +711% in 52 weeks. The structural rerating has happened. Late-cycle. Better entry only on a 25–30% drawdown.

• ValuationTop picks (TSM, META, CRM, COHR) sit inside entry zones at PEGs 0.8–1.2 — argues to deploy. AI cohort breadth at record highs — argues for restraint.
• SentimentCohort euphoria week of 8 May — crypto, memory, custom silicon all near 52W highs. Tilt slightly defensive, not panic-cash.
• EventsNVDA 20 May, MU 24 Jun, hyperscalers late Jul — each carries 8–12% implied moves. Reserve dry powder for post-earnings down-gaps.

15% balances structural opportunity (4 high-conviction names in zone) against tactical caution (record-high week + clustered earnings). Below 12%: too aggressive given upcoming print risk — leaves no ammo for down-gaps. Above 18%: opportunity cost compounds at ~6% per quarter against the 26% target CAGR.

• ↓S&P drawdown >7% from 8 May high → accelerate deployment to ~10% cash.
• ↓Held name retests support post-earnings (MU $650, PLTR $115, NVDA $195) → redirect dry powder into existing position.
• ↑VIX >25 sustained or two of (NVDA, MU, META) miss → step back to 20–25% cash, wait for clarity.
• ↑Memory ASPs roll over on 24 Jun MU print → trim memory exposure, raise cash 3–5pp.