Notes on what the market is pricing in.

Micron closed below $900. Our rule said sell. We didn’t. Single-name update

The two best prints of the cycle — ASML’s beat-and-raise, TSMC’s record quarter with capex hiked to $60–64B — and memory broke on both: the market bought the spenders and sold the supplied, because record tool orders are tomorrow’s memory supply. On July 16 Micron closed at $853.20 — the first close below the $900 stop we published in June. The rule said out in full. We didn’t sell — and after today’s 12.2% rip to $970.82, the violation sits ~7 points ahead of the discipline (~+57% marked vs. ~+50% rule-honored vs. ~+45% never selling). Ahead is not the same as right: we grade the freeze an F on process, name the reasons without accepting them as excuses, and install the line with no discretion left — one close below $848.95 and the runner is gone, with the hyperscaler capex prints (MSFT Jul 29) as the hard deadline for a re-underwritten hold case or an exit regardless. Plus the ASML either/or the market answered with a third branch, and the barbell — the basket and ASML — that did its job while the single name tested ours.

ASML — initiation. Paid by both sides of the memory war. Coverage initiation

We refused to chase it at the $1,989 record, named a $1,550–1,700 buy zone, and the July 1–2 memory crash delivered — we initiated at $1,752.87. On print morning the Q2 numbers graded the entry: net sales €9.3B above the top of guidance, gross margin 54.0%, and the full-year guide raised a third time in six months, to €43–45B — with a committed 30% EUV capacity expansion for 2027 behind it. The only monopoly in the AI supply chain, paid by Micron, SK hynix, Samsung, and TSMC alike — whichever side wins the memory war we’ve been writing about. Initiating at ACCUMULATE, probability-weighted target ~$2,000. The honesty clause: the market front-ran it (+2.9% Tuesday, ~+1% reaction), so the thesis got confirmed and the entry didn’t get cheaper. Add-zone $1,550–1,650; we don’t chase.

The wires fired. The rules answered. Week ahead

Two of last week’s four trip-wires fired, one confirmed, one never triggered — and every response was already in writing. Micron closed $853.20 Thursday, the first close through the $900 line we published June 23, and the rule fired — and we did not sell: the runner is still held, graded F on process in the July 21 stop note, now under a zero-discretion $848.95 line. Friday morning a bullish BofA note landed, the stock ripped ~12% intraday to ~$957 — and closed at $848.95, below the firing close. CPI came in cool (the branch we pre-answered), ASML beat and raised while the market paid the toolmaker and punished the spenders, and oil re-armed the inflation trade — WTI up ~15% on the week after Iran struck Kuwaiti utilities. Now the argument gets its verdict: Alphabet opens hyperscaler capex week Wednesday — the same evening as Tesla, IBM, and ServiceNow — with Microsoft, Meta, Amazon, and Apple behind it. Four new wires are set.

Micron round-tripped again. KeyBanc raised its target to $1,750. We didn’t. Single-name update

Six days after the reckoning, Micron swung from $948.80 up to $991.64, then down to a $937.00 retest, and back to $983.12 — a ~$55 range, about +3.6% net, on no company news. The bookend told the story: the July 1 crash was a Korean leveraged-ETF unwind; today’s rip was US 2× single-stock ETFs launching on SK hynix (+20%), dragging the complex up — flows, not fundamentals, exactly as we said. Into the noise KeyBanc raised its target to $1,750 and the Street held near $1,500 (its high is $2,000), while our ~$1,100 sits at the low end of the range — level with Goldman’s Neutral, below every Buy — on a tape with zero Sells. The argument isn’t demand (Micron just committed $250B to it) — it’s what Micron keeps. And it gets settled tomorrow, when ASML reports and the options price a ±8% move.

The market tested every line we drew. The rules held. Week ahead

All four trip-wires got tested in five sessions. The Warsh Fed's first minutes put a June hike argument on the record; the 10-year closed through 4.50% while the $119B auction slate stopped through everywhere (we cut duration and concede the long-end argument to Schwab); Micron pierced $900 intraday and never on a close — then SK hynix raised $26.5B in the largest-ever US foreign listing, seven times oversubscribed, and memory rallied through it. The Iran ceasefire broke and oil turned back up. The graded ledger, plus the loaded week: CPI, five megabanks, and Warsh's debut testimony in one Tuesday morning — then ASML (our newest position) Wednesday and TSMC Thursday.

Samsung posted the biggest profit in the world. Memory made new lows. That’s the tell. Single-name update

Samsung guided to a world-record ~$58B quarterly operating profit — the largest of any company on the planet, past Nvidia — and memory fell anyway: Micron broke to a new low of $938 while Nvidia ripped through $204 to fresh highs. The peak-out doubt got resolved with numbers, and the market sold the numbers; the AI trade split cleanly into demand (up) and memory pricing (down). We cut the Micron target a third time, to ~$1,100 (from $1,500 in June, ~$1,300 on July 2) — a multiple cut with the full re-underwrite behind it: Micron capped its own DRAM upside while SK hynix didn’t, it may be losing HBM4 share, and single-digit is the cycle-top signature, not the bargain. Why we bought the memory basket and ASML into the crash instead of more Micron, and why the $900 line doesn’t move.

The bounce came with names attached. Micron’s was the smallest. Friday is why. Single-name update

Memory bounced Monday with names attached — UBS raising Q3 DRAM forecasts to +32% (from +17%), BofA’s “healthy reset” at $1,550, Citi’s 90-day upside watch, Samsung pushing +20% contract hikes into what consensus says will be the biggest quarterly profit in its history Tuesday. Micron still bounced least in the complex — because the ~$28B SK hynix Nasdaq debut, the largest ADR offering on record, lands Friday and takes the only-US-memory scarcity premium with it. The tell, the two tests, why the pricing data never confirmed the panic — and a new position: ASML, the toolmaker that gets paid by both sides of the memory-cycle question. Initiation note upcoming.

We said suppliers reprice last. The market sold them first. Week ahead

The June jobs print came in soft and strange — +57K against ~113–115K, with unemployment “improving” to 4.2% only because half a million people left the labor force — and the Dow closed at a record on the relief. But the week's real event was Meta moving to sell its excess AI compute: the suppliers we're long got routed while every megacap spender finished green, and we grade the inverted split call in full. The long end led yields higher all week (our “calmer long end” call is leaking too). Ahead: the first minutes of the Warsh Fed Wednesday, $119B of duration auctions, and SK hynix's ~$29B Nasdaq debut Friday — the test of whether the memory unwind was flows or the cycle.

Micron round-tripped the best quarter in its history. What the unwind is — and isn’t. Single-name update

Two days took Micron from $1,154 to $975.56 — below the close the night before the best quarter in memory history, with our $900 stop 7.7% away. It wasn’t macro: the jobs print broke dovish and the S&P closed flat while SanDisk fell 23% and the whole memory complex unwound. The catalysts: $9B of Korean 2× leveraged ETFs de-grossing, the SK Hynix Nasdaq listing on July 10 killing Micron’s only-US-memory scarcity premium, and a second derivative that turned in the company’s own guidance. We hold the runner behind the pre-committed stop, cut the base target from ~$1,500 to ~$1,300 (a multiple cut, narrated), and show why even the worst case from here locks a ~+53% call.

Micron locked in $100 billion. The AI spenders got a cost problem. Week ahead

Micron’s June 24 blowout — record $41.5B, an 84.9% margin, a $50B next-quarter guide, and $100B of contracted, prepaid demand — ripped it to a new high even as megacap AI spenders sagged on cost worries and the S&P drifted −3% on the month. The AI trade is splitting into suppliers and spenders. Into a holiday-shortened week: quarter-end, the June jobs report Thursday (first labor read under a hawkish Warsh Fed), markets closed Friday. Plus a new idea — we start looking at quantum.

Micron didn’t fade. It ripped 16%. What we got wrong — and the contract that rewrote the thesis. Single-name update

We called the FQ3 print priced and likely to fade. It ripped ~16% to a record $1,213.56 instead — we were wrong on the direction, full stop. But the scale-out still won (a ~40% runner held through the rip), and the headlines missed the real story: $100B of strategic customer agreements whose contracted floor margins sit above Micron’s all-time peaks. That structural de-risk re-rates the base target from $1,100 toward ~$1,500 — while the same ceilings cap the $2,000 the Street now prints. The honest reckoning, the collar, and what we hold from here.

Everyone agrees on Micron now. Tomorrow it has to print. Single-name pitch

MU set a record $1,211 high on Monday, then faded into tomorrow’s FQ3 print (Wednesday, after close). The sell-side modal target is now $1,500 — only Goldman ($900) and Morgan Stanley ($1,050) sit below the stock. But the in-quarter beat is already priced; what re-rates the tape is the FQ4 guide against a $38–42B bar, the HBM4 ASP read, and any FY27 tone. PT stays $1,100. We’ve booked ~50% of the position in a $950–$1,150 scale-out — the book is now 25% MU / 35% AI basket / 40% cash — and hold the rest behind a hard $900 stop. Reaction piece June 25.

We said the surprise skewed up. Warsh skewed it down. Week recap

We called Warsh’s first FOMC to underwhelm the hawks. He out-hawked the tape: the dots flipped to a 2026 hike, the 2-year ripped 16bp (the kill-switch we named), and the S&P fell to 7,420. The call is dead by our own rule — we own it. The honest twist: equities shrugged it off by Thursday. The grade, the lesson, and the refreshed book into Micron’s June 24 print.

SpaceX — initiation. A world-class moat, priced for a flawless 2030. Coverage initiation

The largest IPO in history — and the best operating business in space. Starlink compounds at ~50% on 63% segment margins while the launch moat widens in real time. The problem is arithmetic: a conglomerate that lost $4.9B last year, priced at triple-digit multiples of sales, with 82% of the votes in one founder’s hands and the biggest supply unlock landing before Christmas. Initiating at HOLD — fair value ~$115, target $135: sell the rich premium, buy under $120. Published July 5 with the tape already 26% off the draft-date high — the update note scores the call.

Peace trumped the print. Now Warsh inherits the rally. Week ahead

May CPI hit a three-year high at 4.2%, but a soft 0.2% core and a U.S.–Iran peace deal that collapsed oil to the low $80s sent the S&P to ~7,550. Wednesday, Kevin Warsh chairs his first FOMC — a hawkish hold the market has already priced. The non-consensus call: the surprise skews dovish-of-fear and the relief rally holds 7,450 through Friday. Three trip-wires refreshed; Iran flips from war risk to peace-cracking risk.

Dell — initiation. Start with what we got wrong. Coverage initiation

Initiating Dell at HOLD with a ~$340 probability-weighted reference. The stock re-rated from 9× to 24× forward earnings in five months — before the earnings did the work. Our first hypothesis didn’t survive the audit: the +33% print day was an EPS event at a flat multiple, and the market is paying mid-single-digit-margin AI servers as if the durability were already proven. The corrected math says the Street’s $484 average target is our bull case — the initiation walks the re-rate, the collar of kill-shots (memory as the #1 backlog constraint, the pull-forward air-pocket), and exactly what 20×+ requires.

Streak broke at nine. The thesis hits a tougher test. Week recap

The S&P broke its nine-week streak with a −2.6% week to 7,383.74. Broadcom’s record $10.8B AI quarter with a maintained 2027 outlook triggered the chip selloff Thursday. Friday’s 172K NFP (vs 85K expected) killed the cut hopes; 10Y above 4.54%. Dell tested the bottom of our consolidation zone. Today: Iran/Israel exchange + WTI gap above $94. Trip-wires refreshed, CapEx-raise call still open into June 25.

June is when Micron has to do the work. Single-name pitch

MU peaked at $1,079.57 on June 3 ($20 short of our second trim), then crashed 20% after Broadcom’s Q3 total revenue guide came in light of aggressive Street and a hot May NFP. The PHLX semi index posted its worst session since April 2025 on Friday. Monday June 8, Wells Fargo and Cantor Fitzgerald both raised PTs through the drawdown ($1,220 and $1,500). MU bouncing +8% intraday around $948. The first trim looks better in hindsight. The framework still holds. FQ3 prints June 24.

We were wrong on the math. The thesis broadened. Week recap

The S&P closed Friday at 7,580 — ninth straight up week, the longest streak since 2023. Dell put up its best single day on record at +33%. PCE came in at 3.3% Y/Y, 50bp above our hawkish scenario. The 10Y rallied 25bp from its peak. Three of our non-consensus calls broke (the mean-reversion math, the cross-asset divergence, the AI-multiples canary). What we got wrong, where the thesis still holds, and the pivot into June.

I bought Micron at $668. Here’s why I’m not buying at $881. Single-name pitch

UBS tripled its PT to a Street-high $1,625. Three weeks ago I was the lone voice at $1,100; now six of eleven sell-side firms are at or above me and two are at or above my $1,500 bull case (UBS, DA Davidson). MU closed Friday at $923.52 — my first trim level fired the day after publishing. PT stays at $1,100; ladder continues at $1,100 / $1,400 / $1,500 full exit. When the variant view becomes consensus, the discipline isn’t celebration. It’s trim.

Eight weeks up. Now PCE has to confirm. Week ahead

The supercycle passed all three tests last week. NVIDIA crushed it with a $91B Q2 guide ($4B above Street). Walmart U.S. comp +4.1%. The FOMC April minutes printed the most divided meeting since October 1992. S&P booked an eighth straight up week. But NVDA closed -1.77% on the print, the 10Y closed at 4.56%, and the hawks at the Fed want hikes back on the table. Friday's April core PCE decides whether the dovish framing holds or the rotation finally begins.

Three tests the supercycle has to pass this week. Week ahead

Wednesday May 20 is binary: FOMC minutes at 2pm (read on the most-divided April vote since 1992) and NVIDIA earnings after close (the $87B Q2 guide threshold) land in the same 8-hour window. Plus the consumer test: Home Depot Tue, TJX Wed, Walmart Thu. What to watch, what each outcome means, and the calendar that runs the week.

The macro broke. The market printed higher. Weekly outlook

Hot CPI (3.8%), soft retail (+0.5%), record-low Michigan sentiment (48.2), 10Y at 4.49%, Fed cut odds collapsed to 3%. S&P and Nasdaq still printed new ATHs. Cisco AI orders 4× YoY, NVDA H200 cleared for 10 Chinese firms, Micron +38% on the week (best since Dec 2008). What we’re watching today and into next week’s NVDA print.

The market is cooling under the semis Market recap

Markets head into Friday’s payrolls print with the economy clearly cooling — but equities have not broken. Semiconductors and AI infrastructure are doing the heavy lifting while the rest of the cycle decelerates. Six inline charts and the full sector positioning framework.

MU — Through-cycle long Single-name pitch Report

Filed in Reports · HTML pitch with reverse-DCF + scenarios.

Mid-cycle Micron at $750–$850; the market is pricing peak earnings at trough multiples. Reverse-DCF, the HBM mix shift, and the three observable kill-shots that would make me wrong.