Quick hits - July 17, 2026

SNDK bounces off the 100d, memory maybe leads the recovery too: $SNDK undercut the 100d ema in premarket and is already trading back above it. Memory led this whole selloff, so if it's first to bounce that's worth watching, not betting on yet. Interesting premarket action, nothing confirmed.

GOOGL capex cut talk doesn't survive contact with reality: BofA now sees GOOGL raising CY26 capex 5% to $190-200B on higher memory costs, and Bloomberg's reporting GOOGL's own engineers can't get enough internal compute. Hard to square that with a capex-cut thesis. More corroboration the AI capex narrative isn't dying, it's still growing.

NVDA rebuts the Feynman delay chatter, power semis thesis still intact: Morgan Stanley reiterated that power semis get a real expansion off NVDA's Feynman platform, even after NVDA pushed back on the delay rumors making the rounds. Multi-year thesis still intact despite the recent momentum unwind in semis.

DRAM shortage seen running to 2028, but stocks may already be peaking in 2027: Morgan Stanley's now calling for a DRAM supply shortage through Q4 2028, the longest timeline seen yet on this. Memory stocks historically peak about 6 months ahead of spot prices though, so the read here is a 2027 stock peak, not 2026. Worth keeping in mind if you're riding the memory names into strength.

2yr yield sinks on soft inflation, gold along for the ride lower: The 2-year is grinding lower on soft inflation prints and gold's pulling back right with it. Counterintuitive on the surface, but gold should still benefit if the Fed leans less hawkish from here. Still calling it a 12-24 month accumulation story, not a trade for this week.

NFLX prints a mid quarter, sell-off looks fair: $NFLX revenue landed at $12.6B, right on estimates, and EPS beat by a penny at $0.80. Free cash flow disappointed hard though, and that's the number driving the stock lower today. Mid print, the red honestly looks earned.

AAPL shopping for AI chip startups, in house silicon isn't cutting it: $AAPL is reportedly looking to acquire semiconductor startups to build its own AI server chips, because its current silicon can't keep up with what it needs internally. Doesn't read like a company backing off the AI trade.

ARRY steps into SHLS's lane, not a thesis breaker: $ARRY just acquired Affordable Wire Management for up to $203M, a step right into $SHLS's turf. Scale's way smaller though, so this doesn't move the needle on the SHLS thesis we've been watching since the breakout call.

AAOI down 40% in a month, but the revenue math still works: $AAOI is the poster child right now for 'thesis doesn't matter when a stock's down 40% in a month.' Except here it kind of does: the read is AAOI still does $5.62B annualized revenue at a minimum. If that number holds, the drawdown looks more like noise than a broken thesis.

ABI's near miss: 15.4 vs the sub-14 trigger: Yesterday's ABI reading came in at 15.4, so the signal didn't fire, but it missed the sub-14 threshold by less than it looks. Not the all-clear it might read as. Worth watching if the next print slips back under 14.

SPX levels for a bounce, but next week looks heavy: $SPX has resistance stacked at 7580 then 7553, with support underneath at 7500-7510, 7485-7488, and 7442. Good chance of an intraday bounce off this zone, but the read past today is bearish into next week. If that support stack cracks, don't be a hero trying to catch it.

$SPY's 741 bounce holds, but 745 is the lid on a brutal gamma setup: SPY held the 741 level on today's bounce and QQQ clawed back 690, but the book underneath is ugly, negative gamma reportedly 7x worse than anything else this week. 745 is the lid to watch on any follow-through, and even if we clear it the structure isn't repaired so don't get too excited. 740 is the soft floor below that, lose it and this turns into something worse than chop.

$PENG defends the 50d, flexing relative strength in AI: $PENG bounced off its 50-day and is defending the 9-week, one of the better relative-strength showings in the AI trade right now. Nothing's broken yet, worth watching whether it can keep holding this while other AI names are still chopping around.

The bear case on memory capex, filed for later: One of the more coherent bear arguments floating around right now: hyperscalers can't keep plowing 60%+ of capex into memory forever, that mix has to normalize eventually. No action here, just filing it away as the counter-argument if you're leaned into the memory bull thesis.

$AVGO bouncing off the 200, double bottom chatter: $AVGO put in a strong candle off the 200-day today, and the forum's asking if that's a double bottom setting up a run back to the 400 area. We're also about halfway to earnings, so IV should be sitting near a low point here before it starts creeping up into the print. Nothing confirmed, just a level worth watching if you're looking for a swing entry.

NDX fighting for the trendline, 700 is the line: $NDX/$QQQ is trying to claw back this trendline against the odds, so we watch and see. Lose 700 and the heatseeker setup points to a more likely fade into the close. Getting some encouraging red-to-green flips on individual names, but nothing here worth getting excited about yet.