Weekly Digest

Week Ending 2026-08-02

Regime: Near-term oversold — only 16% of the universe is above its 50-day moving average after Tuesday's broad selldown, though 74% remain above the 200-day MA. The longer-term uptrend is intact but fraying at the edges.

Movers

Up: MSFT led the week at +21.8% after a clean Q4 FY2026 beat — EPS $4.74 vs $4.33 estimated, Azure grew 43% vs ~39% consensus, and critically, capex guidance didn't expand further, calming FCF fears. AMZN gained +17.0% after AWS grew 37% — a genuine 6-point beat vs 31% consensus — with the first-ever $200B revenue quarter; the EPS headline was inflated by a $53B non-cash Anthropic mark-up. ORCL rounded out the top three at +13.0%, bouncing from near 52-week lows after record Q4 results, though the stock had initially sold off on a capex hike from $35B to $50B.

Down: VRT fell -16.8% on a Q2 revenue miss ($3.27B vs $3.38B estimated), hitting with elevated RVOL of 1.72; EPS beat and raised full-year guidance soften but don't erase the damage. SNDK dropped -15.4% on two severe alert days (7/27 -11%, 7/28 -14%), now sitting mid-range at 0.51 of its 52w band. KLAC fell -13.2% in the broad 7/28 semi-equipment selloff.

The week's defining event was Tuesday, July 28: Nearly the entire universe triggered high-severity price-drop alerts in a single session — 26 names across semis, equipment, power, and AI infra all down 4–14%. This wasn't name-specific; it was a macro derisking wave ahead of Big Tech earnings. The universe generated 78 alerts this week (35 high, 43 medium), with the single-day carnage on 7/28 accounting for the bulk.

Risk

Three portfolio positions are currently below their stop prices as of 7/31: MRVL ($187.56 vs $200.13 stop, -14.5% from entry), CRDO ($206.99 vs $217.93 stop, -12.6% from entry), and MU ($822.60 vs $862.04 stop, -12.2% from entry). All three were caught in the 7/28 broad selldown rather than name-specific deterioration. The portfolio drawdown from the June 22 peak (116.99) was -9.6% as of Sunday 7/27 rebalance; estimated current drawdown after this week's moves is approximately -12% — well clear of the -20% circuit breaker.

Portfolio

The weekly rebalance ran July 27 — the day before the selloff. ANET was exited (rank dropped to 15, mechanical score decayed as 3-month momentum turned negative) to free capital for two new entries: AMD (rank 6, score 0.855 — EPYC Venice on TSMC 2nm + AMD Advancing AI catalyst) and ASML (rank 7, score 0.854 — Q2 beat, second FY26 guide raise to €43–45B, cheap vs own-history EV/EBITDA). The timing was unfortunate — both new positions absorbed the 7/28 selldown immediately. All other positions (DELL, MRVL, NVDA, EQIX, CRDO, MU) were held.

Earnings just landed (news-sourced; no surprise table entries yet — [UNPROVEN-LIVE-ONLY])

MSFT Q4 FY2026: EPS beat +$0.41, Azure +43% — clean sweep, +8% AH → +21.8% for the week. AMZN Q2: AWS +37% (6pp beat), record $27.5B operating income → +17% for the week. META Q2: revenue beat ($60.8B vs $55.5B est) but EPS missed by $1.04 and FCF fell 91% on $130–145B raised capex → stock -6.5% (classic capex trap; retail bought revenue, buy-side sold FCF). ARM Q1 FY27: record $1.29B revenue, double beat on EPS and revenue, but flat Q2 sequential guide disappointed — stock -8% post-print → -7.8% for the week. PWR Q2: EPS $4.24 vs $3.33 est (+27%), FY guide raised to $39.3–39.7B vs $35B consensus → +6.6% for the week; JPMorgan downgrade into earnings wrong-footed. VRT Q2: revenue miss but EPS beat and raised FY guide — market focused on the miss, -16.8% for the week.

Next week (Aug 4–7)

Eight names report. AMD (Aug 4, held, -3.8% from entry) is the key portfolio event — consensus EPS $1.61 (+235% YoY); stock enters deeply oversold. ALAB (Aug 4, not held) is top-quintile valuation — elevated downside on any miss. ANET, SMCI, ENTG report Aug 4 without portfolio exposure. SNDK and WDC report Aug 5. NBIS and CEG (Aug 6) both carry top-quintile valuation flags — risk of sell-the-news. VST (Aug 7) dropped -9.3% this week and reports Thursday.

---

> Caveats: Momentum signals are validated on historical data along a single price path — results may not generalize to different regimes. Commodity and power prices can surprise independently of AI infrastructure demand. News items are [UNPROVEN-LIVE-ONLY] and not verified against primary sources. This digest is a system output, not investment advice.