US spot bitcoin ETFs absorbed roughly $1.07 billion across the first three sessions of last week before flows reversed, leaving price pressing against an aggregate ETF buyer cost basis near $81,300. The relevant question for equity investors is not what happens to the coin but what happens to marginal risk capital when the newest cohort of buyers sits at breakeven. The answer in US megacap was rationing rather than withdrawal: a single share of each of the ten names below opened $25.16 cheaper than the prior close, then finished $30.52 above that opening print for a net gain of just $5.36, or 0.14%.
The Print
Breadth was negative while the basket was positive, which is the first clue that this was a dispersion session. Four of the ten names closed green and six closed red. AMD carried the entire aggregate gain and then some, adding $11.43 to close at $519.36 in the upper third of a $508.40–$523.88 range; strip it out and the remaining nine names lost $6.07 per share set. That is a narrow engine, and narrow engines are how index-level calm masks underlying rotation.
| Symbol | Close | Change | Day range | 52-week range |
|---|---|---|---|---|
| AMD | $519.36 | +11.43 (+2.25%) | $508.40–$523.88 | $188.22–$658.52 |
| V | $408.97 | -5.37 (-1.30%) | $406.83–$412.51 | $293.89–$414.60 |
| JPM | $333.13 | -3.41 (-1.01%) | $330.45–$333.58 | $279.10–$366.50 |
| TSLA | $377.14 | +2.56 (+0.68%) | $365.30–$377.31 | $297.38–$498.83 |
| META | $647.06 | -1.79 (-0.28%) | $633.39–$648.01 | $520.26–$779.82 |
| AMZN | $249.20 | -1.18 (-0.47%) | $246.49–$249.50 | $196.00–$287.20 |
| GOOGL | $323.59 | -0.63 (-0.19%) | $320.30–$324.49 | $251.85–$408.61 |
| NVDA | $212.60 | +0.39 (+0.18%) | $210.20–$213.85 | $164.27–$243.37 |
| MSFT | $524.84 | +3.76 (+0.72%) | $515.79–$525.80 | $349.20–$553.72 |
| AAPL | $347.82 | -0.40 (-0.11%) | $346.45–$351.68 | $243.42–$351.68 |
The two most informative closes sat at opposite ends of their own histories. AAPL traded up to $351.68 intraday — precisely its 52-week high — then faded to close at $347.82, down $0.40 but still marking a 90-day closing high and a third consecutive green close. GOOGL, in contrast, closed at $323.59, which is the bottom of its 90-day closing range, on a fourth straight red close and a 6.0% five-session decline. Same session, same sector weighting bucket, opposite structural positions.
The Quality Check
The recovery off the lows was broad, and that matters more than the headline change column. Eight of the ten names gapped lower at the open, which is consistent with the cautious Asian session Reuters flagged and with Seoul trading weaker ahead of chipmaker earnings and the Federal Reserve decision. What distinguishes a real bid from mechanical mean reversion is whether the names that gapped down finished above where they opened.
Six of the eight names that opened below their prior close finished above their opening print. The two exceptions were both financials.
The qualifier is Apple. The one name that touched a 52-week high was also the only name in the group to close in the lower third of its session range, surrendering $3.86 from the high on the session's second-heaviest listed volume. A tag of a 52-week extreme that cannot hold into the close is a weaker signal than the closing price alone implies. The implication is that the intraday recovery was real in the aggregate but did not extend to the name with the least valuation cushion.
CNBC's reporting that Wall Street is hunting for a hedge on the AI trade — including an ex-Citadel quant building futures on physical GPU capacity — is a sentiment datapoint, not an earnings datapoint. Translated into per-share terms, a functioning hedge market for compute would compress the option value embedded in accelerator pricing power. NVDA closed at $212.60, up $0.39 but down 3.5% over thirty sessions and 12.2% below its 90-day high, which is roughly what that de-rating looks like in slow motion.
Peer Context
The financials are where my own work needs correcting. I argued JPM would re-rate above $350.00 within twenty sessions; that call expired unfilled, and the reason is visible in the structure. JPMorgan has now posted five consecutive red closes, is down 3.1% over five sessions and 6.6% over thirty, and at $333.13 sits 8.8% below its 90-day high with only 2.1% of cushion above the 90-day low. A re-rating thesis requires the revision cycle to turn; the price series says it has not.
V offers the counterweight. Visa fell 1.30% to $408.97, a larger single-day decline than JPMorgan's, yet it remains 0.8% below its 90-day high and up 10.6% over thirty sessions. One payments network near the top of its range and one money-center bank near the bottom is not a sector call — it is two different earnings trajectories being priced correctly. On AMD, my pending range call from last week's piece on interconnect pricing remains live, with both the $523.88 and $508.40 boundaries untouched on a closing basis.
Verdict
Fair. The aggregate basket moved 0.14% and the gap-down recovery was genuine, so there is no evidence here of risk capital leaving equities because crypto ETF buyers are underwater. But the internals are splitting hard enough that index-level exposure is now the least efficient way to own this group. The thesis lives or dies at Alphabet: a close below $320.30 within five sessions confirms the 90-day breakdown, while a reclaim of $325.21 would mark the four-session slide as exhaustion. On the other side, Apple closing above $351.68 within ten sessions would establish a 52-week closing high and argue the dispersion resolves upward rather than down. What would change this read is a JPMorgan close below $330.45, which would extend the de-rating from AI-adjacent names into financials and make the rationing look like something broader.