Mag7 Desk Note #48

Executive Takeaway

Breadth held at 4-of-7 above EMA20, but the quality behind it degraded — leaders lost premium while the tape stayed volume-less. All seven remain NEUTRAL with a 31% conviction ceiling and macro isn't the binding constraint (VIX 14.63, HY OAS 2.71%). This is a relative-value, event-gated book into NVDA on 8/26, not a directional one.

Positioning

• NVDA (+5.4% vs EMA20, +4.6pp over AMD/AVGO, 84th %ile) is still the only name with momentum and peer confirmation — and the sole reason breadth looks intact.
• MSFT leads GOOGL/AMZN by 7.5pp but RSI 70.9 on 0.42x RVOL makes it hold-not-add; Gemini's mean-reversion caution is the right posture.
• Fund relative longs from peer-confirmed laggards: AAPL -5.2pp vs MSFT/GOOGL, META -3.3pp vs GOOGL/SNAP at the 25th %ile with a trial starting Tuesday. Claude's point stands — SNAP +6.2% says the ad problem is Meta-specific.
• AMZN (in-line with WMT/COST, 0% conviction) and TSLA (Δ -0.0pp vs GM/F) are breadth arithmetic, not leadership.

Key Risks

• False comfort in EMA20 membership: AMZN +0.8% and TSLA +1.2% are rounding errors, and both fade on one down day.
• Equipment tape is worse, not repaired — AMAT -5.1% on the day, -9.6% over 21d, KLAC -7.1%. Don't read AI capex as confirmed.
• Metals broadened (copper 21d +7.94%, silver +17.59%, platinum +10.01%) — an input-cost headwind for TSLA at 0% conviction and only in-line with GM/F. All three briefs skipped it.

Watch

NVDA +5.4% vs EMA20 into 8/26 (EPS est 2.13, rev est $93.6B, IV30 41.1% — the only name where implied exceeds realized). Lose that premium and leadership narrows to MSFT alone.

What Changed

• Breadth unchanged, internals eroded: NVDA +6.1%→+5.4%, MSFT +8.4%→+7.2%, AMZN +1.8%→+0.8%, META -0.1%→-0.9% (%ile 27→25).
• TSLA is now the only name with real volume (RVOL 0.89x→1.16x) and still zero conviction.
• Headline tone flipped back two-sided, all [NEW]: MSFT's Azure $100B milestone against NVDA vendor-financing risk framing and META's addiction trial.