Five Reports Broke the Memory Trade. Here's Where AI Leadership Is Moving.
The Dow at records, memory punished premarket, and our watchlist for this week.
In today’s note:
⭐Watchlist⭐ $SNOW $PLTR $BE & $NBIS (reporting next week)
The last three sessions went by the book. Closed NBIS for a quick +7% gain this week ahead of earnings. The portfolio is now:
50% across CRWD DELL HPE with a smaller speculative position in LABU with stops, while the other half is in cash
Happy to hold cash here as we are short-term extended. Waiting for the moving averages to catch up to the broader market.
Day two FTD verdict: giveback, NOT distribution.
QQQ held the 50-day, suggesting yesterday looked more like healthy profit-taking than institutional selling.
Memory gets punished premarket ($SNDK, $WDC) and what it means for the $MU trigger
The week ahead: power is the new bottleneck. BE is the main focus, HUT watched for a low, NBIS after its report
The RTL portfolio update as of last night
Why the posture is very selective from here (extended tape, August chop)
Looks like one of the worst momentum selloffs has finally come to an end, with the market healthily digesting its gains off the lows.
The Qs and SPY were both down on less volume than Tuesday which makes me think this was just giveback, not institutional distribution.
QQQ sold off into the close and held its 50-day at 715. That is day two in one line: the indexes handed back most of Tuesday’s gain and still closed on the right side of the level.
Underneath a record-high Dow, leadership has rotated away from AI Memory toward cyclicals and rate-sensitives. Memory and storage are being punished premarket after SNDK and WDC disappointed last night. Oil fell for a second day, yields eased, and the dollar softened, a tailwind for the broad tape, while gold keeps rallying, which tells you the hedge stays on. The main risk to the calm is this morning’s data snapping yields back up. We don’t pick the branch. We hold our levels and let price decide.
TODAY’S POTENTIAL ENTRIES
Clean Energy
BE Pulling into the rising 20-day. Still want to see it tighten here but 241.95 is the confirmation, below 219 we stand down. Spec size, day-low stop.
Software
FTNT pullback we wanted finally arrived late yesterday, low 163.94, close 164.13, right in the zone. If it holds there today, we will be watching but likely to pass.
SNOW Extended so just watching for now but keeps both entries, 309-313 or back through 321.40.
PLTR sits on top of its 152-155 zone. Watching to see if this holds.
The memory and storage trade is broken, and I’ll say that plainly.
Five reports are in: STX last week, AMD and ALAB Tuesday, SNDK and WDC last night. The group failed every one of them.
Every chart in the complex has been rejected at its 20-day.
Good pipelines didn’t matter, in-line numbers didn’t matter, and that tells you the sellers are bigger than the news. I don’t have faith in these names again until the charts repair, and repair means time: higher lows, a reclaimed 20-day, weeks of base-building. Not a bounce. MU’s 902.50 stays on the chart as the line that would start changing my mind. Even then, a book holding HPE and DELL doesn’t stack a third hardware name on top. If MU earns a spot, it competes with what we own. STX comes off the active list with its group.
THE WEEK AHEAD
The plan got bigger than today, so here it is. The AI bottleneck keeps moving. First it was chips. Then it was hardware. Now it’s power and the software layer. Texas froze the ERCOT datacenter queue and goes to the PUCT on August 20, which makes on-site power the scarce asset. To be clear on the macro: cheap oil helps the broad tape but has nothing to do with this trade. The power story is grid access, not crude.
That’s why BE is the main focus into next week. Earnings out of the way, 237.55 the trigger, smaller size and a day-low stop because it moves.
HUT is the other side of the same trade: the ERCOT casualty worth watching for a low. Yesterday was an outside reversal, a run to 104.80 and a close at the dead low of 92.76. This group goes from oversold to overbought fast. A reclaim through 95.50 is the first tell, and if we ever take it, the stop is the low of the day. Into August 20, HUT is the safest of the group with confirmed tenants already signed.
And NBIS: the plan is to be back in after next Wednesday’s report if the reaction sets up. Not before. It’s volatile and infamously hard to trade, and the re-entry has to earn itself post-print.
Down the road, we may trim CRWD and DELL into strength to fund that next wave. By relative strength the list reads HPE, BE, SNOW, FTNT, NBIS, PLTR. Winners get the capital.
The rest: TER retests 382-386. GOOG got slammed at resistance on the news its chief AI scientist is leaving; a calm retest of 370-372 is the only version we want.
QQQ needs to hold 715. Below there the follow-through day goes on the clock and we stop adding. Above it, the same levels stand. With the tape still extended and August historically choppy, very selective is the posture. The market owes us nothing before the averages catch up to price.
THE BOOK
NBIS | banked +7% at the stop yesterday. We tightened it into next week’s earnings on purpose, and it did what a pre-earnings stop does: banked the gain and got us out ahead of the report. Revisit after the print.
CRWD | full, average 193.42, closed 209.86, +8.5%. Stop 203.40. It tagged the 214.30 extension zone and then closed on its day low. Trims could come soon.
HPE | full, average 51.01, closed 53.22, +4.3%. Stop 50.39. New high at 54.63, the only book name green yesterday, and it held through the whole AMD mess. Best-acting name we own.
DELL | full, average 468.12, closed 462.80, down 1.1%. Stop 457.90, under yesterday’s higher low. That stop sits about 1% under the close, so a soft open tags it.
LABU | spec size, 261.00 entry, closed 255.66. Stop 230.90, never threatened. The signal reads on XBI, which held its breakout by eighty cents. Today it needs to prove it with volume.
THREE READS
Two beaten-down internet names, opposite answers.
SHOP delivered: revenue up 34%, sixth straight quarter above 30%, guided ahead. It gapped 17%, reclaimed its 50-day and 200-day in one session, and went from 36% off its high to 21%.
APP missed, blamed timing, and reiterated its long-term growth. They may well be right. Price does not care yet. APP is 44% off its high and below both moving averages. “It was timing, not demand” is a fine explanation and a terrible reason to buy. A stock that has stopped going up is telling you something no matter how good the story sounds on the call. SHOP earned a watchlist spot this week. APP earned another week of watching. That distinction is the entire job.
The S&P closed within 1% of an all-time high. Semis are 15% off theirs.
Both true at the same time, and that is the entire story right now. RSP, the equal-weight S&P, is 0.6% off its high. The average stock is at highs. Financials, industrials and small caps are all within about 1%. SMH is 15% off and IGV 14% off. The index is not the market: money walked out of the AI trade and into everything else, and the tape barely flinched because buyers were waiting on the other side.
For a momentum book the read is simple. Leadership changed. If you are still holding the old leaders waiting for them to come back, you have been sitting in a 15% drawdown while the average stock printed a new high. Trade the names making highs, not the ones you wish would.
Texas just froze the AI power trade, and it has nothing to do with demand.
Abbott ordered an audit of every data center in the ERCOT queue on August 3. Nothing new gets approved until it’s done. The queue: 1,800 projects, 474GW, 90% data centers, five times the record peak of the entire Texas grid.
Earnings didn’t matter. HUT has an 8GW pipeline with confirmed tenants and fell 8% anyway. GLXY lost 14%, CIFR 8%. The only green name in the group is BE, which sells the on-site power you need when the grid says wait. I flagged 102 on HUT into the print. It closed 92.76. Off the list. August 20, when ERCOT goes to the PUCT, is the date.
If the triggers print on volume, we add. If they don’t, we wait. Rotation is not a threat to this book. It’s the reason the book looks the way it does.
See you in the chat.