This report tracks the 9-name hyperscaler/chip/foundry basket (ai-infra.json — AMZN, NVDA, ASML, TAC, TSM, PLTR, MRVL, ARM, P), unchanged from the 2026-07-25 edition, not the separate 52-name physical-buildout universe (ai-infrastructure.json). Because the prior edition used this same 9-name basket, the comparisons below are a valid direct read against 2026-07-25. The clearest move: eight of the nine names weakened this cycle, led by TAC and ASML both flipping from a modest positive week to a sharp decline — PLTR is the lone name that improved, if only marginally.

FROTH READ: vs VWAP measures distance from the average buyer's cost basis across the instrument's full history.

TAC is the sharpest reversal in this basket. Two cycles ago its slide had reversed into a positive week (RSI 42.7→49.7); this cycle it flipped straight back down — RSI fell to 36.7, its 7-day change flipped from +3.47% to -8.93%, and its 30-day change turned negative again (+0.22%→-6.46%).

ASML posted a nearly identical reversal. Its RSI fell from 48.3 to 37.1, its 7-day change flipped from +0.54% to -12.13% — the sharpest weekly move in the whole basket — and its 30-day change deepened from -0.32% to -15.94%.

MRVL and ARM, which had stabilized off their craters two cycles ago, resumed deteriorating. MRVL's 30-day drawdown widened from -29.79% to -37.17% (RSI 39.0→34.3) and ARM's from -27.59% to -28.77% (RSI 38.0→35.9) — both back into acceleration rather than basing.

NVDA and TSM both cooled as well: NVDA's RSI fell from 51.0 to 43.0 (7D +1.99%→-4.96%, though its 30-day change stayed positive at +1.05%), and TSM's RSI slipped from 42.3 to 39.0 as its 30-day drawdown deepened to -13.8%. AMZN's decline continued at a similar pace to last cycle (RSI 37.2→36.3, 30D -0.92%→-3.86%) — still the name that reversed from steadiest to weakest two cycles ago, now just grinding lower rather than accelerating. P (Everpure, formerly Pure Storage) also weakened modestly (RSI 50.3→48.0, 7D +7.47%→-1.79%). PLTR is the one exception in the basket: its RSI ticked up from 41.9 to 44.5 and its 7-day change, while still negative, improved slightly (-7.15%→-6.88%) — it remains death-crossed and well below VWAP either way.

This is a valid direct ticker-for-ticker comparison — the 07-25 edition of this scan already used the same 9-name basket.

1. TAC's slide reversed straight back down: RSI 49.7→36.7, 7D +3.47%→-8.93%, 30D +0.22%→-6.46%, 52wkHi gap -21.09%→-28.75%.
2. ASML cracked hard: RSI 48.3→37.1, 7D +0.54%→-12.13%, 30D -0.32%→-15.94%, 52wkHi gap -9.85%→-20.85%.
3. MRVL resumed deteriorating: RSI 39.0→34.3, 30D -29.79%→-37.17%, 52wkHi gap -36.55%→-47.11%.
4. ARM resumed deteriorating in tandem: RSI 38.0→35.9, 30D -27.59%→-28.77%, 52wkHi gap -37.48%→-45.94%.
5. NVDA cooled: RSI 51.0→43.0, 7D +1.99%→-4.96%, though 30D stayed positive (+3.94%→+1.05%).
6. TSM's drawdown deepened: RSI 42.3→39.0, 30D -8.49%→-13.8%, 52wkHi gap -13.24%→-18.1%.
7. PLTR was the lone improver: RSI 41.9→44.5, 7D -7.15%→-6.88% (still negative but less so), 30D +8.3%→+6.77% (roughly flat).
8. P weakened modestly: RSI 50.3→48.0, 7D +7.47%→-1.79%, 30D +2.42%→+0.93%.
9. AMZN's decline continued at a similar pace: RSI 37.2→36.3, 7D -6.12%→-6.74%, 30D -0.92%→-3.86%.

RSI < 30 = capitulation zone (check trend/death-cross before treating as a buy). RSI 30-40 = oversold, often still structurally weak in this group. RSI 40-70 = neutral-to-healthy. RSI > 70 = extended. "vs VWAP" measures distance from the average buyer's cost basis — the further above 0%, the more "paid up" the current price relative to its own trading history.

None.

MRVL (34.3), ARM (35.9), AMZN (36.3), TAC (36.7), ASML (37.1), TSM (39.0) — six of the nine names, up from three last cycle. Only PLTR, NVDA, and P sit above RSI 40 now.

PLTR (44.5), NVDA (43.0), P (48.0).

• Whether TAC and ASML's sharp reversals this cycle mark the start of a real leg down or a one-week dislocation inside a longer basing pattern.
• Whether MRVL/ARM's resumed deterioration (30-day drawdowns now -37.17%/-28.77%) continues toward fresh lows or stabilizes again.
• Whether PLTR's marginal improvement extends into a real recovery from its death cross, or reverses back down alongside the rest of the basket.
• Whether the ai-infrastructure scan slot stays on this 9-name basket going forward (unchanged for three cycles now) or reverts to the 52-name physical-buildout universe.

• Tape data (price/RSI/SMA/returns): full-scan pre-computed summaries as of the 2026-07-28 close — ai-infra.