Good morning.
Tape's ugly underneath. NVDA down 8 of the last 10, optical names rolling over, Most Short basket +25% in 8 of the last 9 weeks — positioning reset, not a demand break. GS TMT pair dumped 875BPS yesterday. Don't have clean US index futures in the feed, but the cross-currents are loud.
No TMT earnings prints in the AH/BMO window to flag.
Asia: Brent +1.8% on U.S.-Iran uncertainty. SoftBank dropped a €75B / 5GW France AI commitment — first European build at true hyperscaler scale. Energy is now first-order site selection, not a footnote.
Four threads for the day:
(1) MEMORY BOTTLENECK ELONGATES INTO 2028. Goldman raised SK Hynix 2027 OP +22% ($218.9B → $266B) and 2028 +24% ($242.8B → $301.2B); Samsung 2027 +21%, 2028 +23%. Bears running "HBM: High-Bandwidth Mistake" got smoked at the operating level. Direct read for MU and SNDK.
(2) SOFTBANK €75B / 5GW FRANCE. Schneider Electric partnership on the Dunkirk cluster validates 800VDC + liquid cooling at European hyperscaler scale. "Demand ahead of buildout" reads universally across every CSP. Read-through to grid, DC power, and the European AI infra complex.
(3) CATALYST CLUSTER TODAY. Jensen reveals Feynman GPU + N1X Arm PC chip at GTC Taipei; GPT 5.6 vs Opus 4.8 release week; Vista's $8M → $200K insurance-claim inference datapoint is the first hard enterprise ROI proof at the systems layer — 40x cost reduction, value accrual sustainability check: positive. META 4th frontier challenger thesis percolating too (most incremental GW in 2026 of any AI lab, MSL team ramping).
(4) POSITIONING RESET VS FINANCING CRACK. Rental rates per GPU-hr stable-to-up, xAI took 90-day reclaim optionality, demand ahead of buildout everywhere. The "running out of buyers" SPV concern is real but still Twitter-native — watch if it migrates to institutional positioning. Component cost is the actual squeeze: DRAM/SSD +200-300% since mid-2025 forcing 2H26 PC ODM production cuts 15-18% YoY. Consumer breaks first, compute chain doesn't.
We'll hit up NVDA and MU first, then get to the AI infra complex (optical, semis, neocloud).
Verdict: Street finally caught up to the tape. HPE printed a monster Q2 — $10.7B rev / $0.79 EPS, both ~50% above consensus — and is guiding FY27 for the first time. But here's the thing: stock already up 178% YTD, trading 47 handles, and nine firms are piling in with PT hikes of 60-110% to their targets. The debate isn't the print. It's whether the multi-year re-rating is done or if you chase a stock that just ripped 30% in a day into an "in-line" FY27 guide (vs a bull-case $4+ EPS number).
PTs getting FLOORED higher, cluster sits $65-$74, BofA tops the range at $80. Range of 8 firms: $62 (Bernstein, Market Perform) to $80 (BofA, Buy). Consensus works out around ~$70, implying ~50% upside from $47. Notably, the most aggressive raiser (BofA, +$42) is also the only one whose target sits above where the stock traded pre-print — everyone else is playing catch-up to a name that's already done the work. BofA's $80 looks like the bull anchor; Bernstein's $62 is the bear anchor. The split is Buy/Outperform on one side, Market Perform/Equalweight/Neutral on the other. Nobody's a sell, but half the street is refusing to chase.
Bull case (BofA $80, Truist $69, RJ $74, Evercore $70): Demand is "white hot" (Truist) and supply, not demand, is the gating factor — mgmt said no pull-forward, no cancellations, no double ordering on the call. That matters. FY27 guide implies EPS >$4.00 and FCF of $4.5B+, which puts the stock at <12x forward even after the rip. Juniper synergies (now 45% of forward OI per Truist) are running ahead of plan, structurally higher margins, and agentic AI is bleeding into traditional server/storage demand — extending the cycle beyond the AI hyperscaler cohort. Truist sees demand durability for at least six more quarters. Piper calls it the "Year of Refresh." ASPs up 30%+ are sticky because customers are absorbing DRAM/NAND inflation.
Bear case (Bernstein $62, Piper $63, UBS $65, MS $71 Equalweight): The re-rating is the story. Stock +60% since May 22, on par with Dell's 61% — so the trade's played. Raymond James flags that HPE's beat was "more modest" than Dell's. The really interesting data point everyone glosses over: AI server revenue was FLAT YoY (Bernstein). The 33% server growth is traditional servers +44% on pricing, not units. That's a refresh cycle, not an AI capex supernova. MS frames it best: "after a sharp re-rating, proving this represents a true multi-year cycle is the key debate." Their risk-adjusted r/r is only 13% with a $42 bear case. UBS explicitly cited "potential risk of order pull-in" as a reason to stay Neutral despite raising estimates. FY27 demand durability beyond the visible pipeline is the open question.
"After a sharp re-rating, proving this represents a true multi-year cycle is the key debate." — Morgan Stanley
"Demand conditions are 'white hot' with tight supply conditions continuing across HPE's product lines." — Truist Securities
"Supply rather than demand is the gating factor." — BofA (paraphrasing mgmt)
Three things are actually incremental here, not just the print:
1. FY27 guide dropped early. Mgmt usually waits — the fact they put numbers out now signals order-book visibility, and that's a quality-of-earnings flex. 8-12% rev / 12-16% EPS / $4.5B+ FCF. 2. Pipeline = multiples of backlog. That implies the FY27 number is bookings-supported, not just extrapolation. But RJ correctly notes "fiscal 2027 outlook appears to be based on improved order trends rather than committed bookings" — so the bull/bear fight on what "pipeline" means is live. 3. Mgmt explicitly waved off pull-forward concerns. "No signs of pull-forward, cancellations or double ordering" is the line that matters. If customers were pulling in, you wouldn't guide the year out.
Stock's at $47, consensus PT ~$70, but the easy money's been made. The 30% gap fill already happened. Longs want to see the analyst day in two weeks confirm FY27 is bookings, not pipeline, and that the AI server line (FLAT this quarter) re-accelerates. Shorter-term, the r/r favors waiting for a pullback into the high $30s/low $40s rather than paying up here. Half the street (Buy/Outperform) thinks you chase; the other half (Market Perform/Neutral) thinks you don't. Read: cautious into the print, constructive on the franchise, not chasing at 178% YTD.
Verdict: Print was a BLOWOUT. Guide was a SCREAMER. Stock sold off anyway. That's the trade. $437M April quarter beat Stifel by 1.6%, $470M July guide mid was 3% above Street, and FY27 growth just got a 30-point upgrade — from 50% to 80% YoY. That's a number that used to take 2-3 years of guide-ups to achieve, done in one print. Stock ticked down 5% after-hours because, well, +260% in a year leaves you with everyone who's already long. The setup now is whether the stock can digest 30 handle multiple expansion on a $2.4B+ revenue base when optical is supposedly only halfway through its ramp.
Four brokers in, three Buys, one Neutral. PT cluster $215-275 with the Street average hanging around $250. Read that range carefully — it tells you everything about the debate.
BULLS (Needham/TD Cowen): FY27 growth guide went from 50% to 80% in one print. That's a regime change, not a tweak. Optical revenue lifted from $500M to $600M+ with all three product silos (ZF Optics, SiPh PICs, optical DSPs) each clearing $100M — proves it's not a one-product story. Half the dollar growth comes from copper/AECs, half from optical, which means the existing book isn't even plateauing yet. 68% gross margins in a connectivity business. TD Cowen flags a NEW 10% customer in FQ4 that's different from the 11% customer in FQ2 — that's the diversification story actually working, not just talk. Roadmap to CPO and NPO in FY28.
"The existing photonics roadmap brings a clear path to CPO and NPO solutions in fiscal 2028." - Needham
BEARS (Rosenblatt): Optical TAM is $25B in calendar 2026 and Credo's $600M target is a rounding error. High-margin zero-footprint transceivers is a real product but doesn't make them a market leader — they're selling the most profitable slice of a market they don't dominate. At 56x P/E with the stock 0.93% from 52-week high after a 261% run, the easy money's been made. And that 80% FY27 growth number? Depends on a sharp H2 ramp. That's a lot of execution packed into two quarters. Rosenblatt's $215 PT (25x FY28) is the only one in the group that prices in terminal value skepticism.
"Rosenblatt said it is not impressed with the $600 million optical target for fiscal 2027, given that the calendar 2026 datacom optical transceiver market is tracking toward approximately $25 billion." - Rosenblatt via report
"Credo Tech's strategy to sell high-margin datacom products, such as zero-footprint transceivers, suggests the company will remain a small player overall in datacom optics." - Rosenblatt
Already known: Credo is a copper-to-optical pivot story with hyperscaler exposure. 226% LTM growth. AEC proliferation thesis.
NEW on this print:
CRDO is the purest optical scale-out trade in mid-cap semis right now. The narrative is catching up to what MRVL and AVGO have already priced — that 1.6T optical displaces legacy copper at the back-end, and Credo owns a real slice of that. Bull case reads through to COHR, LITE, and CIEN on the components side, and NVDA ecosystem on the AI infra demand side. Bear case is pure multiple compression — at 56x P/E with a stock that's done 4x in 12 months, you don't need the thesis to be wrong, you just need it to be less right than the next guy thinks. After-hours action suggests positioning for that risk-off scenario.
Bottom line for PMs: This is a hold or trim into name, not a load up on name. Print is real. Guide is real. Setup is harder. The 2H ramp Rosenblatt flagged is now the single most important data point in semis.
Verdict: Street is chasing a stock that just ripped 71% in three months — and the bid is getting louder into the June 16 print. PT cluster just reset to $275-290 from a $215-250 range. We think the setup is more interesting short-side than long here.
Three name-brand PT hikes in the last 48 hours. Scotiabank $290 (from $215 — a 35% raise, the most aggressive), UBS $285 (from $250), Wedbush $275. Cluster says: AI cloud infrastructure is real, OCI is winning, and FY27 capex guidance could approach $100 BILLION — which Scotiabank calls "materially exceeding" consensus. That's a massive number and the real story of this quarter. Not revenue. Capex.
ORCL at $242, up 28.5% in a week, +71% in three months, +50% trailing. Market cap ~$710B. P/E 44.7x. Trading at 27x CY27 non-GAAP EPS per UBS. Stock has compressed the risk/reward in a hurry — the easy money on the "Oracle is a real AI infra player" trade is largely behind us. Every dollar of upside from here requires the June 16 print to be a clean beat-and-raise with credible FY27 visibility.
Bull: Structural AI cloud positioning. GPU-as-a-service expertise, access to capital, customer neutrality (not competing with customers like AWS/Azure do in certain verticals). CEO says data centers "on schedule or ahead." Oracle bond market is telling you something — 4.5% 2028s at 64bps over Treasuries, tighter than tech peer average. Credit is voting yes.
Bear: $100B FY27 capex is a gravity well. Scotiabank flagged gross margin pressure from hardware inflation (~15%) and is only modeling a 1% EPS increase despite the PT jump. If capex is $100B, free cash flow takes a hit for multiple years. At 44x P/E and 27x forward, the stock already assumes OCI scales cleanly with no margin degradation. No margin of safety.
UBS did the work that matters. Karl Keirstead's team spoke with four large ORCL customers and partners plus one contractor with direct visibility into the Abilene, TX AI data center build-out. No setbacks found. That's the kind of primary research that underpins a $285 PT and it's a meaningful data point — the bear narrative needs to explain away a clean build-out that's already on schedule.
"We previewed key investor topics and likely fiscal 2027 commentary... did not find evidence of any setback in momentum." — Karl Keirstead, UBS
We'd be trimming into the print, not adding. The setup — stock up 28.5% in a week, PT cluster reset $75 higher, capex number likely to terrify the tape on June 16 — looks like a sell-the-news event. Bulls have the narrative and the momentum. Bears have the math. Math eventually wins.
THE TAKE: EUPHORIA IS REAL, BUT BofA JUST DREW THE LINE. Stock up 167% YTD, ripping through the old $71 52-week high, trading $78.63. The quarter was a monster (revenue $3.1B, EPS $0.13, June guide $3.45B vs $3.2B Street), and the data center narrative got a serious upgrade. But BofA — the most thoughtful skeptic on the name — only moved to $83 (5% upside) on a Neutral. Read that again. The bull camp thinks this thing goes higher. The bear camp is saying it's priced.
March quarter was the LARGEST EARNINGS BEAT IN NEARLY THREE YEARS per UBS. June guide of $3.45B is $250M ABOVE CONSENSUS — that's a 7-8% guide-up. Not a print, not a whiff — a print AND a guide. PMs who faded the cyclical recovery story on STM are now paying the price. Q1 reset the bogey, and the guide reset it again.
The whole Street is now anchored to the same story. STM raised 2026 DC rev to ~$1B from "nicely above $500M" — a doubling. And management said 2027 could DOUBLE AGAIN TO ~$2B on current customer engagements. That's a 4x in two years.
BofA, Jefferies, Mizuho, Craig-Hallum all leaning bullish. The optical mix is the key driver: ~2/3 of the $1B+ in DC orders this year are optical (silicon photonics PICs, BiCMOS EICs, MCUs), with GM >40% on those optical lines. Power chips make up the rest. Jefferies models DC as ~7% of total 20.5% growth in 2027 — meaningful but not the whole story. BofA lifted rev estimates 3-5% across 2026-28 and GROSS MARGINS BY 126-200BPS.
"Further rerating would require greater confidence in the execution of AI-related sales." — BofA Securities
That's the cleanest analyst line in the bunch. Translation: we believe the numbers, we're not sure we can underwrite the next leg.
This is a name where the RATE OF CHANGE matters more than the level. The print and guide already catalyzed the move. From here, you're trading multiple expansion on a 2027 optical ramp that's binary on execution. BofA is telling you exactly that with their Neutral. Long investors should be trimming into the gap-ups, not adding. Shorts need optical capex confirmation before they can press. The cleanest trade might be selling calls into this euphoria.
Verdict: Stock pinned near highs ($230.70, 52wk $236.54), and the buy-side is broadly constructive but not euphoric. PT cluster $255-$307 with Truist/Davidson at the high end ($300-$307) and Deutsche Bank the lone holdout at $255. New product news is TAM expansion, not thesis change — the real story is Vera Rubin supply chain DOUBLING vs Grace Blackwell. That's the revenue ramp.
Two-article day, clean split. Truist repeated Buy/$307 post GTC Taipei — DGX Station and RTX Spark computers extend NVDA into local/edge AI development, "opens significant TAM opportunities." DA Davidson and Goldman also in the buy camp ($300 and $285 respectively, per the articles). Deutsche Bank stayed on the fence with Hold/$255, focused on the CPU complement to GPU story out of Computex. The consolidation theme across the buys: NVDA is methodically expanding the surface area of its AI footprint — datacenter → edge → client → PCs. Truist kept estimates unchanged despite the new product slate, which tells you this is strategic positioning, not a model-moving event today.
"Vera Rubin is now in full production with approximately twice the supply chain size compared to its Grace Blackwell ramp."
That's the line that matters. Everything else at Computex/GTC was marketing. Supply chain doubling is a 2H revenue tell — Lynx already calling for a material ramp in Q4 or Q1.
Bull: Vera Rubin full prod + 2x supply chain = revenue inflection in 2H. TAM expanding vertically (datacenter → client → edge → robotics via Sharpa). 50% FCF return target = capital return story emerging. $5.49T mcap, 34x P/E, "EXCELLENT" financial health — pricing power intact.
Bear: Stock already at $230, sitting 2.5% off 52wk high. DB's $255 PT = ~10% upside, not exactly a compounder from here. P/E 34 on a $5.5T chip name leaves little room for an air pocket. New product slate (DGX Station, RTX Spark) is cool but speculative on volume — not what FY27 numbers hang on.
Long NVDA, trim into strength. The Vera Rubin supply chain doubling is the real catalyst — if hyperscaler orders confirm in Q2/Q3 prints, $300+ gets there. If we get a multiple compression event in the semis complex, the DB thesis protects you on the downside. Not a fresh money name at 52wk highs, but no reason to fight the trend.
The trade: $376, $4.59T mkt cap. Cloud backlog 4x'd in 3 quarters and Street is still modeling deceleration. That mismatch is the whole story. Consensus is way too low on cloud revenue AND way too high on FCF. Both work in the bulls' favor if Truist's backlog math plays out.
Backlog: $108B (Q2'25) → $462B (Q1'26). That's a 4.3x in nine months. Truist's Youssef Squali is making the call that consensus isn't pricing this right — modeling 63% Google Cloud growth in Q1 ramping to mid-80% by year-end, while the Street is modeling deceleration into the 50s. That's a 30+ point spread on the most important growth line in the print.
"Consensus revenue estimates for Google Cloud are below what they should be based on backlog as of March 31 and recent deals." — Truist Securities
The Anthropic $200B partnership flows through FY26/27 estimates. TPU sales start later this year — a new revenue line that basically didn't exist in consensus models six months ago. And the Apollo/Blackstone $36B loan to fund Anthropic's purchase of custom Google AI chips is circular financing that bulls should love: it's de-risking the chip customer base while keeping the buildout capex off Google's P&L.
Wells Fargo (Ken Gawrelski, $435 PT, Overweight) dropped a number that should sober up the FCF-focused crowd: $290B in projected capex, $45B above Street. That takes free cash flow to roughly breakeven by 2027. Combined with the surprise equity raise (which Gawrelski called out explicitly given the $50B+ in credit already issued YTD and ~$50B net cash on the Q1 balance sheet), the message is clear: Alphabet is going to spend whatever it takes to extend the compute lead.
"The size of the equity offering suggests a balanced approach to financing current plans rather than pursuit of new opportunities." — Wells Fargo
Read: this is a maintenance raise for a $290B spend plan, not opportunistic. Wells is also flagging a potential Blackstone off-balance-sheet vehicle for TPU-based capacity — creative financing to monetize the chip franchise without lighting up the balance sheet further.
The math: if compute has to shift ~60% toward Google Cloud (from internal uses) to meet customer demand, the mix story is doing more work than the absolute spend number. This is a capex story about TAM expansion, not margin defense.
Bull: Cloud backlog is a forward indicator with multi-year visibility. $462B recognized over 5-6 years is a revenue floor. TPU sales are a new monetization layer. Equity raise at the top is positioning, not distress. The market is mispricing a company that's converting AI demand into contracted backlog faster than the model can digest it.
Bear: $290B capex to breakeven FCF by 2027 is brutal. Returns on incremental AI dollars are unproven at scale. TPU sales cannibalize internal cloud economics. Equity dilution at $4.59T is a signal that even the GOOGL balance sheet can't self-fund this build. Multiple compression risk if growth fails to inflect as Truist models.
Two brokers, both constructive, PTs clustered $430-435 vs $376 print — ~15% upside on the bogeys alone.
Q2 print needs to show cloud growth re-accelerating toward that 80% handle, not the 50% consensus. Backlog disclosure (or silence on it) will be the tell. If Truist is right and the Street is modeling 50% growth on a $462B backlog, the Q2 number could be a major consensus-rewriting event.
Print was a monster. Guide raise was bigger. But the stock's up 284% in a year — chase at your own risk.
Q1 FISCAL 2027: REVENUE $43.8B (+88% YOY), EPS $4.86 vs ~$3 expected. ISG +181% YOY, CSG +17%. FY27 EPS guide hiked from $12.90 to $17.90 — that's a 39% RAISE in a single quarter. Stock $431, +42.6% in the week post-print, near 52-week high.
$500 PT cluster forming: Bernstein ($280→$500, Outperform), Goldman, Mizuho. MS upgraded UW→EW at $448. Truist the dissenter — $170→$360 PT but Hold stays. Everyone getting bullish except the guy with the lowest print. Telling.
Bernstein FY28 EPS $22.84. Truist FY28 sales +11%, EPS $22. Truist's $174.3B FY27 rev model sits ABOVE Dell's $165-169B guide — even the bear sees upside.
Bernstein lays out the rare triple working simultaneously:
"rising content, higher prices on a like-for-like basis due mainly to memory price inflation, and higher units."
All three vectors firing at once. Plus agentic AI driving CPU demand, share gains vs. parts-starved competitors, and a small but real pull-in that's already baked into the guide. Supply-constrained, not demand-constrained — that distinction matters.
Bull: 39% guide raise is the kind of print you don't fade. ISG +181% means hyperscaler/enterprise demand is still bottlenecked by Dell's ability to source, not customer willingness to buy. 27x FY28 EPS (Mizuho's frame) is rich but defensible at 20%+ top-line growth. AI capex cycle nowhere near peak.
Bear: Truist's Hold is the read-through the bulls don't want you to read. 21X CAL 2027 EPS on 14% EPS CAGR FY27-30 — PEG over 1.5. Memory inflation forcing price hikes and nobody knows where the demand curve breaks. +284% IN 12 MONTHS has priced in a lot of the AI optionality. Pull-in exists, even if small.
Trade vs. invest question. Fundamentals legit, guide legit, supply tightness legit. Valuation no longer cheap. Longs: hold, don't add. PMs looking to enter: wait for a 10-15% pullback or a cleaner risk/reward. This is a name you want to own — just not at this price.
Name a more momentum/positioning tape than this. FLNC PRINTED A 302% RUN OVER THE LAST YEAR, GAVE BACK 12% IN THE PAST WEEK, AND IS NOW HOVERING AROUND $27.84 WITH A SIEMENS/NVIDIA DESIGN WIN SITTING IN ITS LAP. The prints are at war with themselves. Street is bifurcated — Mizuho saying it's a sell at $15 (yes, a $13 gap to the downside from spot), Canaccord saying $28 buy. The truth is somewhere messier, and it depends entirely on whether you believe the AI data center storage thesis is a $100M revenue line or a $1B one.
Mizuho (Underperform, $15 PT) and Canaccord (Buy, $28 PT) both stuck to their guns on the Siemens/NVIDIA DSX Vera Rubin NVL72 reference design news. The new architecture embeds Fluence's SmartStack into the full electrical BOM for a ~103-136 MW AI factory (slight variance between write-ups, call it ~120 MW mid-point). Fluence is the SOLE NAMED BESS PARTNER across seven infrastructure OEM designs in the NVIDIA ecosystem. That part is real.
The design specs 2-3 HOUR BATTERY DURATION vs the prior ~1 hour assumption. Mizuho flagged this as $2-4/SHARE OF UPSIDE, OR 10-20% FROM LAST CLOSE — but they didn't move the rating. Both analysts converged on the same caveat: no exclusivity, no committed order flow yet.
"The DSX ecosystem is crowded and non-exclusive with multiple partners." — Mizuho
That one line is the whole bear case in eight words.
Bull: The 2-3 hour duration spec is a quasi-magical upgrade on prior Street models. If hyperscalers are willing to pay a 2-3x storage premium per site for AI factories, and FLNC is the only named partner on the NVIDIA reference BOM, the TAM math rewrites. Canaccord: the modular SmartStack pitch is a "start small, scale to hundreds of MW without redesign" story. That's the dream — land one gig, replicate 50x. Plus two hyperscaler supply agreements last quarter, validation the funnel is converting.
Bear: Q2 REVENUE MISSED BY ~26% ($465M VS STREET). Stock ripped in the after-hours anyway because the narrative is the trade, not the print. But that's a yellow flag — when fundamentals decouple from price action this violently, the unwind is violent too. Mizuho already pencils in 10% data center share and STILL can't get to $27. The ecosystem is crowded. No revenue commitment from this Siemens win. And at +302% IN A YEAR, the easy money is on the other side of this trade.
We get why it works — sole-sourced on the reference design, duration tailwind, AI data center capex cycle. But 302% in a year, down 12% last week, a quarter that missed revenue by a QUARTER OF STREET, and a sell-side price target spread of $13 on a $28 stock? That's not a setup, that's a pin action trade. PMs: if you're long, you already know. If you're not, you wait for the next leg down to engage — the entry from here has horrid r/r.
Verdict first: story stock, no near-term change. Two Buys post-analyst day, an $8 PT spread between them, and the most honest line of the morning is "sharpen the framework but do not change current estimates." That's not a catalyst — that's a roadmap review. If you're long, you're long the 2030+ story. If you're looking for a print, this isn't it.
Stock at $29.18, $10.79B mcap, up 70% over the past year. Both Rosenblatt ($43 PT) and Stifel ($35 PT) reiterated Buys after D-Wave's inaugural analyst day at the NYSE. The bull thesis got more color. The estimates did not move.
D-Wave repositioned itself from a commercial annealing vendor to a full-stack, dual-platform quantum shop. Annealing generates revenue today. Gate-based is the growth bet. Company claims 1,000x performance advantage vs IBM's gate model in optimization tasks — bold claim, but bold claims are the currency in quantum.
Key disclosures: 26 public customers over the past 18 months, $588M of liquidity, a quantified "at scale" annealing economics model, and a revenue framework broken into systems / QCaaS / services. The $100M CHIPS Act LOI is the policy tailwind. Roadmap hits 17-physical-qubit gate system in 2026, 49-physical-qubit in 2027, and 100 logical qubits by 2032. (The 2032 endpoint is the actual bet here. Anything before that is noise.)
Bulls (Rosenblatt at $43, more aggressive): dual-platform approach captures the full addressable market. Real annealing revenue now, gate-based optionality later. Gov't validation via $100M CHIPS. Roadmap lets you underwrite a 2030+ thesis instead of just waving your hands.
Bears (anyone who looks at the print): $10.79B mcap on $12.4M LTM revenue. Q1 revenue $2.9M vs $4.14M est — a 30% miss. Still unprofitable. The 1,000x claim vs IBM is getting contested by classical simulation work, which D-Wave pushed back on, but the fact that there's a fight is a yellow flag. And the near-term qubit counts (17, then 49) are tiny — meaningful milestones for engineers, meaningless for PMs sizing positions.
"The disclosures sharpen the long-term margin and capacity framework but do not change current estimates." — Stifel
Read that line twice. Nothing changes for FY26, FY27, or FY28. The analyst day was a narrative event, not a numbers event. Both firms walked in bullish and walked out bullish — but the gap between $43 and $35 is the real tell. That's not conviction. That's a range trade on a story.
Bottom line for the book: QBTS is a policy-and-narrative name, not an earnings name. CHIPS Act dollars + a credible (if aggressive) dual-platform roadmap = the trade. The fundamentals won't show up in the model for years. Position accordingly.
Verdict: Fresh IPO, fresh initiations, and the Street is leaning in bullish on the China-alternative rare earths theme. Two buys on day one with PTs implying 30-58% upside from $21.52. Speculative name, but the discount to comps is about stage, not quality — and re-rating catalysts (scoping study, resource definition, permitting) are visible on a 12-18 month horizon.
B.Riley and Cantor Fitzgerald both opened with buys this morning. Canaccord already on the tape last week at $25. Targets cluster $25 (Canaccord) / $28 (B.Riley) / $34 (Cantor) — average ~$29, or 35% upside from spot. Cantor is the most aggressive on a Speculative Buy rating, framing REA as a critical piece of America's rare earth supply security puzzle. B.Riley went the NAV route: $1.7B equity NAV, 0.35x multiple = $28 PT, explicitly noting the discount to comps reflects "early-stage development rather than asset quality."
REA owns a diversified portfolio of ionic adsorption clay (IAC) and monazite assets split between Brazil and the U.S. — the kind of geographic optionality that matters when one geography's permitting goes sideways. The headline number: 467.9M tonnes inferred resource at 2,155 ppm TREO, enriched in dysprosium and terbium (the heavy rare earths that are the expensive, strategically-critical ones going into defense magnets and EV motors, not the light stuff like cerium that everyone has).
The IAC processing flowsheet is already commercially validated at Serra Verde in Brazil, and B.Riley flags 30-50% lower capex intensity vs. hard-rock alternatives. That's the margin of safety on the eventual project economics. The U.S. piece — Shiloh — is exploration upside at "limited incremental cost to the current valuation" per B.Riley. Basically free option.
The macro tailwind writes itself: U.S./China decoupling, DoD funding for non-Chinese rare earth supply, Section 232 chatter, MP Materials' stock chart. REA is a second-leg-of-the-trade name.
Upsized IPO closed at ~$63.3M by selling 3.33M shares — bigger than initially planned, which on a fresh name is a vote of confidence from the buy side. Market cap ~$436M, current ratio of 5.99 (no near-term liquidity issues). Board just approved the 2026 LTIP with 3-year vesting RSUs — standard governance housekeeping, nothing controversial.
Bull: Portfolio is genuinely diversified across two jurisdictions and two deposit types. Heavy REE enrichment in Dy/Tb is exactly what the market is paying up for. Multiple re-rates as resource moves to indicated, scoping study drops, and permitting advances. Security-of-supply narrative has bipartisan support and seemingly unlimited duration. Stock at 0.35x NAV for a portfolio of this scale is a literal option on execution.
Bear: Pre-permit, pre-financing, no revenue. This is a 3-5 year development story at minimum, and junior miners can and do blow up between milestones. China can flood the market to kill Western competitors (they've done it before). U.S. permitting timelines are brutal — EPA, BLM, state-level, tribal consultation, the whole gauntlet. The "discount to comps" is real for a reason, and if milestones slip the multiple can de-rate just as fast as it can re-rate. Speculative for a reason.
$21.52 spot, $25 (Canaccord PT) is the first bogey — that level also roughly aligns with the IPO pricing economics. $28 (B.Riley) is the consensus-ish target and a logical first take-profit zone. $34 (Cantor) is the bull-case ceiling and where you'd want to be trimming. Below the IPO pricing, you're in "deal broke" territory and that's where the volume support sits.
Not sure we can size this for a 9-figure book yet — need to see float, borrow, and how it trades through the first earnings print — but for a PM looking to add China-alternative REE exposure with more torque than MP, this is the most obvious new name on the tape this week.
VERDICT: Broken chart. Two Holds, two PT cuts, stock 82% off the highs and making new lows. Not a setup to fight until CAC inflects AND subscriber growth re-accelerates. Watching from the sidelines.
Both sides of the Street hit ODD with PT reductions, both stuck at Hold. Truist $12 (from $18), Jefferies $10.25 (from $14). Stock printing $9.49, market cap down to $792M, scraping a new 52-week low (prior low $10.80). Down 82% over the past year — this is what a growth shock plus multiple compression looks like in real time.
Story is clean and ugly. Largest ad partner (read: Meta) tweaked its algo in Q1, which pushed ODD's IM ads into lower quality auctions and Doubled CAC. Q1 is normally the biggest customer acquisition period, so the cohort damage compounds through the year via repeat orders that never materialize. The fix is multi-quarter. Q1 print: adj EPS -$0.17 vs street modeling +$0.02, revenue -26% YoY (better than feared -30%, but a print that raises real questions). The only green shoot: mgmt still guides to positive 2026 EBITDA.
"Make the stock an execution story... back-half loaded and cost structure-dependent." — Jefferies' Brian Tanquilut
That captures the whole setup. Tanquilut is basically saying: you can't underwrite the model on subscriber growth, so you're underwriting on cost discipline and a back-half recovery. Low visibility on near-term earnings AND long-term growth is a brutal combination — Tanquilut told you exactly what kind of stock this is.
Bull: Algo issue is technical, not structural (Truist's framing). May IM CAC -28% sequential — first data point suggesting stabilization. Other ad channels performing fine. Mgmt guides to positive 2026 EBITDA. $792M cap with a path to an EBITDA print — if algo normalizes by Q3/Q4, meaningful snapback from a beaten base.
Bear: Truist says "several quarters to resolve." Q1 was the big acquisition window — the cohort damage is already locked in for 2026. No subscriber growth visibility. Street can only underwrite the cost line, not the top line. Two Holds with PT cuts is not a bottom signal. And at $792M, the market is pricing this like the long-term growth story is impaired, not just delayed.
1. Does May CAC improvement hold in June/July data — one month is noise 2. Subscriber growth re-acceleration (currently MIA) 3. Any color on algo discussions with the ad partner 4. Gross margin defense — cost story has to work if top line doesn't
This has all the ingredients for a squeeze (beaten name, algo fix narrative, possible EBITDA inflection) but conviction is thin and the path is long. I'd want two clean prints before leaning long. Not a name for the short book either — borrow costs and the EBITDA guide make that a tough r/r from here.
JCI is no longer a sleepy building-controls spin-off — it's making a credible pivot into data center thermal, and the market has caught on. Stock up 35% over the past year to $140.24, trading at a 41 P/E on what is still a ~7% top-line grower. Not cheap, but the story has legs and UBS just bumped to $180 (Top Pick) after the "Going to Gemba" analyst day, calling it "one of the most encouraging meetings we've attended in many years." Strong words from a Tier-1.
The headline from Gemba: management raised long-term organic growth to HIGH SINGLE DIGITS from mid-single. That's the number that drove Street estimate moves. But the bigger structural story is data center thermal — JCI laid out a target of $11B+ IN DATA CENTER THERMAL REVENUE WITHIN 5 YEARS, implying a CAGR north of 30%. Alloy Enterprises (thermal platforms, financial terms undisclosed) is the M&A expression of that ambition. Airside Center of Excellence + Baltimore field tour were the operating evidence.
This is shaping up like a VRT-style rerating narrative for a legacy industrial — except VRT rerated from a much lower multiple. JCI is at 41x already, so the market's priced for execution, not hope.
"Substantial runway for business transformation initiatives in the factory and field, which we believe imply significant upside to target incrementals." — Oppenheimer
The Oppenheimer line on "target incrementals" is the under-discussed piece — they're flagging margin expansion on top of the revenue story. Two-way lever if it plays out.
Bull (UBS, $180 PT, Top Pick): 28% upside to PT. Business system overhaul is real, data center ambition is credible, Alloy is accretive, and they're "highly confident." Sees room for both margin AND multiple expansion.
Bear (Street consensus — Perform/Peer Perform): P/E 41 on 7% revenue growth is rich. Oppenheimer PERFORM, "sidelined as it monitors execution." RBC at $154 (only 10% upside, Sector Perform). The $11B data center number is the swing factor — and MS just flagged that even Vertiv Americas is showing timing-lag issues vs. gigawatt capacity adds. JCI's ramp is a 5-year story, not a 2-quarter one. Last Q was good (adj EPS $1.19 vs. $1.11 on $6.1B rev) but the bar keeps moving; miss a print at this multiple and it's a 15% gap down overnight.
Bottom line: real pivot, not a hidden name anymore. The next leg needs the data center milestones to land — and even UBS-level conviction doesn't get you paid at 41x if the street stays at Perform. Add into weakness, don't chase.
MS RAISES PT TO €1,660 (FROM €1,400), STILL OVERWEIGHT. Bull case boils down to EUV volume ramping faster than the Street: 90 tools in '27, 104 in '28 (96 low NA / 8 high NA). Versus the 80+ unit bogey mgmt floated on the Q1 call — the bar the buy-side keeps fixating on — MS is comfortably above. Memory LTAs and the cash gen cycle are the tailwind that doesn't get enough credit in the AI-narrative version of this story.
Stock trades $1,705, 56.5x P/E. Up 120% YoY on ~10% rev growth. Multiple has done the heavy lifting already.
Bull (MS): Memory manufacturers flush with cash signing LTAs is the sleeper catalyst. 32x on '28 EPS of €51.92 gets you to €1,660. The demand wave is multi-year, not a one-quarter pull-forward.
Bear: Easy. At 56x you're paying for perfection and then some. 120% return in a year means the easy money's gone. Nikon publicly trying to undercut on price — even if it's a rounding error in EUV share, it gives bears a narrative. And every quarter that prints "in line" rather than a beat is going to feel like a disappointment at this multiple.
"Investors remain focused on the 80-plus unit capacity level outlined by management during the first-quarter earnings report. The firm believes ASML can meet demand for approximately 90 EUV tools in the coming year." — MS
Goldman also in the bid at €1,600 Buy — second street high-profile bull print this week. Two sell-side heavyweights anchoring the name above €1,600. That's the consensus floor now, and the real question is whether 90 EUVs in '27 holds or becomes the new conservative case after a few more quarters of demand.
EUV is a 2-player market (Nikon isn't a real threat in advanced nodes regardless of pricing). ASML's bottleneck status is a feature, not a bug, for the AI capex trade. The risk isn't the order book — it's the multiple. Own it, but don't add here.
AI database trade running hot — PTs ripping to $400-515, stock already +31% on the week. Tigress just took the high mark to $515 (Buy), joining a cluster of raises post the FY27 Q1 print where Atlas growth printed 29% vs. the 26% bar mgmt set. Needham still holds the bear flag at $400 (Underperform) but that's barely 2% from where it's trading — even the cynic isn't calling for a meaningful pullback.
Consolidated PT landscape: Tigress $515, Stifel $435, Cantor $416, Needham $400, UBS $350. Street cluster has now reset meaningfully higher, and the stock at $393.56 is already through 4 of those 5 targets. Tigress is the only meaningful upside bogey left on the board. Tigress calling out the $90-100B TAM for database + AI data platforms — and we think that's the right frame. Vector search + production AI workloads is the narrative everyone is paying for.
"MongoDB is building a multi-cloud, AI-ready database backbone that converts mission-critical workloads and vector search adoption into growth."
That's the bull thesis in one line. Multi-cloud + AI vector = mission-critical workloads that don't churn. If that's true, this isn't a multiple compression candidate when rates move. R/R from here though? Stock has done $393 from $300 in a week. Tigress $515 implies ~30% upside to the high PT, ~$200 implied low (UBS $350) is the bear case. That's a 1.5:1 skew at best for a name that's doubled in a year. Not a sell, but not a chase either on a quiet tape. If you missed the move, wait for a flush into the mid-$300s. If you're long, ride it — the target cluster is now too high to fade.
Take: Priced for the breakout it hasn't delivered yet. Stock's ripping — up 45% YTD, 63% over six months — but UBS is out here quietly flagging that the Data Center story everyone is buying is actually lagging peers on growth. That's the headline nobody wants to talk about at $97.81 when the PT is $130.
MCHP broke out Data Center as a separately reported sub-segment within Data Center and Compute. Pieces: storage controllers/expanders/accelerators, PCIe and CXL memory controllers, PCIe switches and retimers. CY2025 the segment was ~7% of revenue ($302.7M). CY2026 guide: ~65% growth to roughly $500M, pushing the mix to 8-9% of revenue. UBS calls this "similar to peers in size or growth rates, with growth actually lagging some key peers." Read that twice.
"This is similar to peers in size or growth rates, with growth actually lagging some key peers." — UBS
XpressConnect PCIe 6.0 / CXL 3.1 retimers at 64 GT/s is the legit new product — pin-to-pin latency meaningfully below spec, aimed squarely at AI fabric reach extension. Real product, real customers, but it's still a sub-$1B piece of a $5B+ revenue base.
Quietly announced hikes across the broad portfolio. No P&L impact in the June quarter (too late), so it's a CQ3+ story. Bull case: classic cyclical recovery margin lever as inventory normalizes. Bear case: demand is soft enough they're resorting to price. Probably a bit of both.
Bull: The 45% YTD move is the market pricing in the cycle turn — book-to-bill inflecting, DCs scaling, ASP hikes layered on. $130 PT is 33% upside from here and the multiple can keep expanding as DC mix climbs. Buy the breakout sub-segment disclosure as a transparency gift.
Bear: "Similar to peers" is not the flex at 18x+ sales. If MCHP is lagging the MRVLs and CRDOs of the world on DC growth, the multiple premium doesn't make sense. Pricing power announcements in a recovering cycle is fine; pricing power announcements in a still-mixed cycle is a tell. 8-9% revenue mix from DC isn't a story, it's a footnote.
Stock's 6-month performance is essentially the market voting on cycle inflection. UBS stays constructive but the tone is measured, not aggressive. After a 63% six-month run, this is more of a hold-and-let-it-work than an add. Watch the CQ2 print for whether the DC sub-segment disclosure comes with upside vs. the $500M guide — that's the next real catalyst.
Street can't raise these targets fast enough. Stifel went to $321 from $230 post-Murphy's COMPUTEX keynote — and they're not alone. Cluster now spans $200-$321 across six firms (Benchmark $275, KeyBanc $260, DB $240, Cantor $220, TD Cowen $200), all Buys. Stock at $284, +258% YoY, riding the Q2 guide-up of +12% QoQ.
"The Future of AI Scaling Depends on Connectivity." — Matt Murphy, COMPUTEX 2026
That single sentence IS the MRVL bull case for 2026. Custom silicon, optical DSPs, PAM4, interconnect — Murph has positioned the company as the picks-and-shovels play on AI infra scaling past pure GPU count. Every hyperscaler needs the plumbing, and MRVL owns a lot of it.
Bull case isn't the debate — the multiple is. Stifel's new $321 PT = 55x CY27E P/E. Current trailing is 97.8x. So even on 2027 numbers, you're paying 55x for what should be a maturing capex cycle. EPS glidepath to ~$10 by 2028 (Cantor's number) is the bridge — get there, multiple compresses naturally and the stock works. Miss it, and these levels get ugly fast. Not sure we can read too much into cluster PTs when every AI name is getting the same treatment, but the breadth of raises here is real.
Truist staying long at $280 PT after the Headless 360 / Slack product deep-dive. Stock at $200.11, so ~40% upside to PT if you trust the call. Bigger takeaway: Salesforce framing headless architecture as net additive — customers consuming MORE of the platform, not less, as agents route through it.
"We attended Salesforce's deep-dive product webinar on Headless 360 and Slack late last week and came away incrementally confident in Salesforce's broader AI strategy." — Terry Tillman, Truist
Bull (Truist, TD Cowen $240): AI is expanding the pie, not cannibalizing it. Agentforce ARR already past the $1B mark. Headless 360 = more surface area, stickier workflow lock-in as every AI interface funnels through Salesforce. 25 upward earnings revisions is the Street leaning in.
Bear: Freedom Broker just chopped PT to $230 from $360. Still Buy-rated, but that's a ~36% trim in one move — the magnitude tells you AI monetization anxiety is real even among the bulls. TD Cowen at $240 and Truist at $280 bracket that cut, so consensus has clearly re-anchored lower.
PT cluster now $230-280 vs stock at $200. The Freedom cut is the bogey — if Agentforce keeps ramping and Slack monetization shows up, you squeeze toward $280. If consumption metrics stall into the next print, that $230 floor is in play. Show-me setup, not a momentum setup.
HSBC goes $600 from $450, Buy. 30% upside to PT — but stock's already up 86% YR and hugging 52w highs. Easy money's been made. Question now: how much gas left in the tank?
The print that matters: HSBC sees ASIC revenue going from $46B in FY26 to $100.2B in FY27. That's 23% and 26% above Street. Not an upgrade, a reset.
Three things working in AVGO's favor right now:
That line is the underappreciated one. Multi-year, locked-in hyperscaler relationships in custom silicon — that's what justifies the premium the stock's already trading at.
What's already in the price: A lot. At $460 on a 25% rev grower with 77% GMs, you're paying for a premium. The bull case needs the ASIC TAM to keep surprising to the upside. If Street catches up to HSBC over the next 2-3 quarters, multiple expansion slows. The trade now is earnings revisions, not multiple rerating.
Other threads worth flagging:
JMP upgrades to Market Outperform, $16 PT — that's ~50% upside from $10.68. The angle isn't complicated: Freshservice crossed $540M ARR growing 27% Y/Y, and JMP thinks it walks to ~$1B by 2028 as the EX mix shifts from ~50% to 70% of total ARR. CX is the laggard, bookings weak, but EX is doing the heavy lifting and the AI monetization story (Freddy AI Agent Studio) gives them a new vector beyond just seat expansion.
Cantor ($12, OW) and Needham ($15, Buy) are aligned on the same thesis — EX pivot, 2028 targets, aggressive buybacks. 8% FCF yield NOW, 85% gross margins, net cash, share retirement ongoing. That's the kind of profile that doesn't need a multiple expansion to work — the FCF comp alone gets you paid while you wait for the AI attach to ramp.
"More than 60% of bookings already come from expansion, with approximately 40% growth in customers spending more than $100,000, 18% average revenue per account growth, and 111% employee experience net dollar retention rising to approximately 119% for multi-product customers." — JMP
The bull case steelmanned: $80B TAM, Freshservice is the real asset, Freddy AI + Fire Hydrant + Device42 build out a platform story, and at 8% FCF yield with a path to Rule of 50 by 2028 this is mispriced. Multi-product NDR at 119% is the single best number in the deck — that's what you want to see when the thesis is "land and expand."
The bear case steelmanned: CX is the canary. Weak CX bookings means the second leg of the stool is wobbling, and if AI lets customers consolidate vendors, Freshworks could be on the wrong side of that trade. Small-mid cap ITSM is a crowded field (ServiceNow looms, plus the modern challengers). 27% growth in Freshservice is good but decelerating from prior prints, and the $1B ARR 2028 target requires execution through a macro that may not cooperate.
Positioning read: Name trades like a value/FCF story with an AI option. Not loved, not hated, sitting in no-man's land. The JMP upgrade + the confluence with Cantor/Needham creates a setup where the sell-side is leaning the same way — usually takes a clean print or a guide-up to catalyze. Watch for CX stabilization in Q2; that's the unlock. Until then, it's a compounder-in-waiting trade, not a momentum name.
Mizuho downgraded OKTA to Neutral from Outperform but RAISED the PT to $125 from $110 — read that again. The downgrade isn't about the business getting worse, it's about the stock running 49% IN A WEEK and 62% YTD past where even the bulls want to chase it. Shares trade at $131.53, which is already ABOVE Mizuho's new target. The downgrade is a valuation event, not a thesis event.
Q1 was a clean beat and OKTA remains the leader in IAM, no one's disputing that. The bull case is real: agentic AI products are off to a good start, large enterprise traction is building, NRR accelerated to 107%, and cRPO beat guide and consensus. Mizuho acknowledges all of this. They just don't see a near-to-medium-term re-acceleration path that justifies the move we've already had.
"It is not yet convinced that Okta can significantly re-accelerate growth over the near-to-medium term."
(For what it's worth, 22 analysts revised EPS estimates higher post-print, so the fundamental community is still warming up — the action is just outrunning the model.)
Truist bumps PT to $320 from $310, with Wolfe at $320 and UBS up at $333 — cluster's sitting $320-333 vs. spot $264, so ~21-26% UPSIDE. Stock's already grinding near 52-week highs ($278.56), so this isn't a "discover value" trade — it's a "consensus still underestimating the AI rev ramp" trade. The market's starting to price it, but Street models haven't caught up.
The setup: AMZN just locked in $100B partnerships with Anthropic and OpenAI, and Truist is raising AWS rev estimates for FY27+ to reflect the workload migration. But the better angle is they're arguing consensus is wrong on BOTH sides of the P&L — Street has FY27 capex growth at only 12% while modeling 27% rev growth. The "underestimate capex AND underestimate rev" combo is the right kind of bull case: heavy spend as the fuel, not the problem, assuming it monetizes.
"Consensus capital expenditure growth stands at 12% for fiscal year 2027 versus revenue growth of 27%."
Custom silicon is what gives this teeth. TRAINIUM3 SHIPPING EARLY 2026, TRAINIUM4 SLATED FOR H2 FY27. If Anthropic/OpenAI workloads land on Trainium instead of Nvidia, AWS rev mix gets structurally better (margin uplift on infra layer) AND the capex intensity narrative shifts. That's the thesis Truist is buying — and they're explicitly above FactSet consensus on rev.
Bear steelman: stock's up massive YTD, near highs, and you're underwriting $100B of partnerships that haven't printed a dollar of rev yet. If Trainium4 slips or the AI workload migration is slower than the bull case assumes, that 27% FY27 rev growth print starts looking spicy. And capex accelerating = FCF compression into 2027 — the market's been forgiving on that trade (AWS growth story trumps cash flow optics) but the tolerance isn't infinite, especially if macro rolls.
What we're watching: AST SpaceMobile's Blue Origin rocket blew up during testing — no AMZN Leo satellites aboard, but a reminder the Kuiper/LEO launch cadence remains a real dependency. Healthcare hire (Roy Schoenberg from telehealth, starts July 1) is a 2027+ story, not a near-term mover. Real catalysts are FY26 capex guide on the next print and any Trainium4 customer wins we can spot from supply chain.
Verdict: $550 PT from JMP feels easy here. Stock printed $460.52, up 11% in a week but still 17% off the 52w high. The setup is obvious — reaccelerating top line, expanding margins, and a CEO who finally sounds like he has a model story instead of an OpenAI dependency.
JMP's Patrick Walravens keeps his Market Outperform and $550 bogey. Upside is ~19% from here, and the bull case has more support than it did two months ago. Revenue growth is ACCELERATING — 17% FY26 vs 15% FY25, with LTM at 18%. Margins ticked up to 47% from 46%. ROE 34%. This is a compounder printing green across the dashboard.
The AI narrative is shifting from "renter" to "owner." Nadella laid out a three-layer stack — agentic experiences, agent platform, cloud/token factory — targeting a $5.1T TAM by 2030. The NVIDIA RTX Spark collab is the first real signal Microsoft wants to compete on-device, not just rent GPUs to everyone else.
"Microsoft is building an end-to-end AI tech stack with three layers: AI agentic experiences, an agent platform, and the cloud and AI token factory." > — JMP Securities
Build conference is the next catalyst. Piper also chimed in Overweight on Copilot improvements — basically every broker on the tape is saying the same thing: AI monetization is starting to inflect. Two risks we can't dismiss: the frontier model question (can they actually build one that matters?) and the OpenAI overhang. Bull case is that MSFT becomes a platform of platforms. Bear case is they're a systems integrator riding someone else's model. R/R still skewed long here, but not by a mile.
Bogey: $550 PT, ~19% upside. Not a moonshot, but a grind higher with a catalyst pipeline. Accumulate on weakness.
Stifel just ripped their PT to $66 from $24 — that's a $42 raise, ~175% higher than where they were — and the stock is only ~$10 away from that bogey after a +229% run over the past year. Buy maintained. The thesis is simple: integrated memory + AI infrastructure is the engine, and the FY26 guide is now tracking toward the high end (7-17% rev growth, $2.00-$2.30 non-GAAP EPS). Stifel's FY27 EPS lands at $3.15, so the new PT is 21x that number. They've effectively conceded the stock rerated and are catching up to it.
But the print underneath is messy. PT dispersion is wild — you've got Stifel at $66, Citizens at $35, Barclays at $27, and that's BEFORE you account for the old Stifel target of $24 still floating around in some of the recent flow. That kind of spread usually means we're in price discovery mode and the bears haven't capitulated (Barclays downgrade to Equalweight is the live counter-narrative — they like the memory biz but think Advanced Computing ramp is slower than hoped as AI spend shifts Cloud > Enterprise). The integrated memory biz is the gift that keeps giving, but Advanced Computing is the optionality the bulls are paying up for.
This is the part that gives me pause. CFO Nate Olmstead resigns effective July 8 — to "pursue another opportunity" — right after the stock goes on a 229% tear. The boilerplate says "no disagreement over corporate strategy, operating performance or financial performance." Sure. VP of finance Aaron Johnson steps in as interim while they search. This is the kind of departure that gets a wink-and-nod from the buy-side in normal conditions, but when the multiple has already expanded this aggressively, a CFO walk during a melt-up is the kind of thing that can crack the tape on a down day. Watch the open.
Bull case: Integrated memory is structurally tight, AI infrastructure spend is still inflecting, Advanced Computing eventually ramps as enterprise AI moves from pilot to production. $3.15 in FY27 EPS is conservative if memory pricing holds. Stifel's $66 still has room.
Bear case: Stock has done all the work. 21x FY27 is a growth multiple for a business that's guiding 7-17% top-line growth this year — that's priced for a meaningful reacceleration. CFO departure near the highs is a tell. Barclays is right that Advanced Computing is a slower burn than the tape implies. Plus, the valuation work on InvestingPro flags the stock as overvalued vs. fair value, which is the sober cross-check.
Not sure we can read too much into the CFO exit yet — could be a guy who thinks the easy money's been made and wants to launch his next thing. But the timing is bad optics, and in a tape this extended, optics matter. I'd want to hear the new CFO's story before adding here.
FIRST ONE LANDS. GNRC +123% YOY, stock at $269, $15.85B mcap, and Needham/Jefferies both leaning in after the hyperscaler MSA print. This is deal #1 of 2 management telegraphed — the bigger 2027-revenue one still ahead. PTs clustered $282-302, street is tight above spot with no real upside vacuum, but the narrative is now de-risked.
"A highly anticipated catalyst that GNRC had telegraphed." — Needham's Sean Milligan
Generac announced a global supply agreement to provide backup gensets to a hyperscaler (name undisclosed). Mgmt had previously flagged this as one of two hyperscale MSAs in the pipeline, and noted one carried "significant" 2027 revenue potential. We don't know which bucket this first one is in — but the second shoe is now the obvious catalyst.
Q1 print underneath was clean: EPS $1.80 vs $1.35 est (33% beat), rev $1.06B vs $1.05B. Base business still putting up numbers even before any DC contribution. That's the under-appreciated piece here — not a single-story trade.
Jefferies Buy, $24.79 PT, AND on their Most Overvalued list. Welcome to China AI data center land, where conviction and valuation are openly at war. VNET up 92% over the past year to ~$10.60. The bull case is now ~134% upside from spot, and that's the cheap seat — BofA's at $16.30 (54% upside). Spread between houses is wide; the buy-side isn't close to agreeing on terminal.
What works is the order book. Management called H1 2026 "unprecedented" on gigawatt-scale orders from two leading CSPs, vs the 100-300MW historical norm. 2+ GW OF DATA CENTER ORDERS YTD ACROSS THE MARKET. VNET itself landed 519MW, with 510MW from one hyperscaler, deliveries spread across 2026-2028. Multi-year visibility most data center names would kill for.
"The willingness of hyperscalers to lock in multiyear data center capacity reflects their strategic imperative to secure compute infrastructure ahead of peers."
That last point is the real tell — hyperscalers paying up for option value on compute, not just today's workloads. ASP stability in a market that saw years of price erosion is the underappreciated piece.
Q1: revenue 2.69B RMB (+20% YOY, 1% ABOVE STREET), adj EBITDA 892M RMB (+31% YOY), EPS -0.15 vs -0.20 est. Solid on the surface. But Jefferies themselves called it below their model on new project delivery timing — so even the bull house isn't popping champagne. (Timing of project deliveries is going to be the swing factor for every quarter until 2028.)
Two things give us pause. One: H2 2026 shifts to "several hundred MW" orders from second-tier players. That's where these cycles get ugly — weaker customer, weaker pricing, slower conversion. Two: the 92% run means the position is owned. The fact that Jefferies is carrying this name on Buy AND Most Overvalued tells you the math is uncomfortable at $10.60.
R/R is poor chasing here. Fundamental thesis intact, but you're not missing a 2x anymore — you're missing a 30-50% move if execution holds into the 2026-2028 delivery curve. Buyer at $10.60, not chaser. If you're already in, fine, ride it, but this is a trim-into-strength tape, not an add.
GLJ took PT to $428, highest on the Street. Buy maintained. Stock printing $294.63, up 207% YOY, $15.6B cap. GLJ's Austin Wang talking to private co-vendors and competitors and seeing a path to HIGH-20S EBITDA MARGINS — that's the key delta here. That's the multiple expansion argument. If margins walk to 28%+ on the data center mix shift, this is no longer a cyclical HVAC name, it's a cooling platform trading against VRT.
"Modine continues to execute and future growth at high incremental margins will increasingly drive comparisons to cooling incumbent Vertiv."
That's a punchy line because GLJ has SELL on VRT. They're explicitly making the relative call — long the challenger, short the incumbent. Bold trade construction if you want to run it pair-wise.
Q4 print: EPS $1.71 vs $1.57, rev $954.4M vs $920.7M. Both beats. But the real catalyst was mgmt raising 2028 LT guide BEYOND CURRENT CAPACITY and announcing a volume commitment with an existing hyperscaler. (Doesn't sound like demand is the issue.) UBS already moved to $355 from $310, DA Davidson to $330 from $265 — both Buys, both citing data center traction. GLJ's $428 sits well above the cluster.
Stock's done a 3x in a year and the floor keeps ratcheting. Not sure we can read too much into a single $4B announcement when the stock's already priced for hockey stick, but the margin walk is the real tell. If GLJ's private channel work is right and MOD gets to high-20s EBITDA on data center mix, then $428 isn't crazy vs. the $294 print. Risk is the bar is now cathedral-high and any softness in the data center capex narrative gets punished. PMs sized here need to respect the velocity.
KeyBanc bumped PT to $237 from $233, OW maintained. Token raise. Stock's up ~198% YoY and trading near the 52-week high of $227.24 — at ~34x forward P/E vs the historical 18-20x range. So the Street is essentially saying: yeah, this one's priced for execution now. Sits in a "show me at the highs" bucket, not a chase.
The margin story is real and it's the engine. Q1 Electronic Materials gross margin printed >43% vs the ~33% FY24 average — richer mix plus operational improvements from softer-volume periods. KeyBanc is modeling >40% segment margins for FY26. Good print. But the analyst is also being honest about the trajectory, which is the most useful part of the note:
"Sustaining quarterly segment margins at the first-quarter level would be difficult during 2026."
That's a tell. Q1 likely marked a high-water mark, and the bogey for the next 2-3 quarters is: did the mix hold, or do we lap the comps and get mean reversion toward the low-40s. EPS beat ($1.27 vs $1.23) and revenue beat ($549.8M vs $479.15M) confirm the demand side isn't the issue.
The new angle is the $65M investment from a major US defense contractor to expand beryllium capacity, with commercial production targeted for 2028. KeyBanc sizes the unlock at ~$65M of incremental high-value beryllium sales on top of the current $300-350M annualized base — and Materion is capacity constrained today, so this is straight incremental, not cannibalization. Real strategic win, but 2028 is a long tape-out. Don't put that in your Q3 model.
IQOS ILUMA clad strip facilities are back to pre-issue run rates with full normalization expected H2 FY26. That's a known issue rolling off — modest tailwind into the back half, already mostly in the print.
Bottom line: 14th straight year of dividend hikes, $0.145/qtr — yeah, the cash compounder thing is intact. But at 34x, you're paying for the beryllium defense optionality AND the margin sustainability that KeyBanc itself is flagging as tough. New PT implies ~30x 2027 EPS, so multiple compression is the path to the target even on flat numbers. Not a short here, but the r/r favors watching a $30-40 pullback into mid-30s multiple than buying at the highs on fumes.
Up 58% IN A WEEK TO $280 — sitting a hair below the 52-week high ($285) and the tape is doing the talking after Snowflake kicked off Summit 2026 and printed 31% product rev growth while raising FY27 guide by 400BPS. This is no longer a "show me" story. The print is the show.
Multiple brokers lifting PTs post-keynote. Benchmark reiterates Buy, $270 (under review — they flagged the investor session still ahead). Monness to $320. HSBC upgrade to Buy from Hold, $289, specifically calling out CoCo scaling to 7,100+ accounts since launch. Three different firms, three different angles, all pointing the same direction: AI product adoption is real and ramping, not vaporware. Anthropic partnership expanding (Claude into Cortex AI), BlackRock on Horizon Catalog, Sanofi deploying sales agents. The enterprise AI plumbing narrative is coalescing fast.
Bull: 31% growth at this scale is genuinely hard to comp. CoCo adoption curve (7,100 accounts in a quarter-ish) is the proof point the bears have been waiting for. Anthropic integration gives them a credible frontier-model story without the capex baggage. Sum-of-parts re-rating still has legs if Cortex AI monetization shows up in NRR.
Bear: +58% IN SEVEN SESSIONS means positioning is now a feature, not a bug. Stock trading near highs into day 2 of Summit — any stumble on the AI roadmap or a CFO lighting a fuse on margins and this thing gives back the gap. 31% growth is also a deceleration from prior prints; bulls need the acceleration, not the maintenance.
"Snowflake serves as a mission critical agentic control plane or command center for AI innovation." — Benchmark
That framing matters. If PMs underwrite SNOW as infrastructure for agentic AI rather than a data warehouse with an AI bolt-on, the multiple expansion has a thesis behind it. If it's just a re-rate on a good print, the 58% move is the trade.
Investor session today. Guidance trajectory into F2Q and any color on Cortex AI attach rates / consumption economics. Stock needs a green light on margins or a fresh catalyst from here — 280 is not where you want to be chasing into a quiet tape. Bogey: 285. Clear that and we're talking $320 street-high territory. Reject and the air pocket fills fast.
Positioning note for the book: this is now a name where you trim into strength, not add. Risk/reward at $280 on a name that just ripped 58% is asymmetric to the downside in the near term even if the 12-month setup stays constructive.
Piper opens Overweight at $280, slotting into a Street that's already deep long. PT cluster runs $170-$320, BofA tops the tape. Thesis is the GTA 6 setup: 13 years since GTA 5, Nov '26 launch, ~35M units modeled in FY27 on a ~30% console attach rate, 26% rev growth. BofA thinks the $8.2B FY27 guide is conservative. We're 5+ months from launch and the stock's already pricing most of the clean news — not a lot of room for a soft catalyst path.
"Fiscal 2027 revenue guidance of $8.2 billion appears conservative." — BofA
The under-appreciated leg is Zynga. AppLovin mediation integration from early '25 is already showing up in UA efficiency and ad monetization, with Toon Blast and Match Factory! the workhorses. Mobile flipped from drag to structural grower — meaningful change vs the post-acq narrative. Q4 print reinforced it: bookings 2% ahead, op income 30%+ above Street.
Catalyst path is well-defined: Trailer 3, marketing, pre-orders = next 3p. Post-launch, GTA Online, PC, and next-gen consoles = the revenue algorithm. Each leg has historically re-rated the name. Watching the trailer drop and pre-order open for demand-intensity reads.
Risk: GTA 6 slips, or attach comes in light vs the 30% bogey. Also — last large-cap gaming indie standing, which puts a soft M&A floor under the stock. Not a near-term catalyst, just a tail.
New CEO, same struggle. Rapid7 announced Corey Thomas is moving up to Executive Chairman effective immediately, with board member Wael Mohamed (30+ years cyber, ex-Forescout, ex-Trend Micro, on the RPD board since April 2025) taking the seat. Q1 was a clean beat — EPS $0.36 vs $0.30, rev $210M vs $207.93M — and mgmt REAFFIRMED Q2 and FY guide. So this isn't a "CEO change because the quarter broke" situation. Reads more like a planned succession. Insiders promote an insider, optics are stable. Good.
Mizuho reiterated Neutral with an $8 PT. Stock sits at $8.74. Translation: the Street target is BELOW where it's trading, meaning even the sell-side defenders can't justify the current quote. Mizuho called the shot pretty directly —
"We believe Rapid7 will likely struggle to significantly reignite ARR growth for the foreseeable future."
That's the line that matters. Q1 print was fine on the income statement, but SecOps TAM is getting crowded (SentinelOne, CrowdStrike, the whole SIEM reshuffle with Palo/GCP), and parts of the RPD portfolio are under pressure. Recent execution issues are real. ARR re-acceleration is the bull case and Mizuho is saying don't hold your breath.
The "Most Undervalued" screen flag at $8.74 is a trap setup if ARR doesn't bend higher. Cheap gets cheaper when the growth profile stays flat. Mohamed knows the product (board tenure, sector pedigree), so execution risk is lower than a fresh external hire — that's the mild positive. But the bar to get Mizuho off Neutral is a credible ARR inflection, and mgmt reaffirmed guide which means the next 2 prints are unlikely to be the catalyst. We'd be sellers into strength, not buyers of the dip. Wait for Mohamed to lay out a reorg/strategy day before getting constructive.
Citizens stayed long with a $170 PT, framing the same thesis: scale is the moat, monetization is the unlock. ROKU now touches 50%+ of US broadband households and captures 44% of US streaming hours per Comscore (4Q25). Stock's run is the story — +76.73% over the past year, hugging the 52w high. Hard to argue with the tape.
Citizens' anchor is >$1B FCF by 2027. That is the print that will make or break the bull case from here. Revenue levers stack: third-party DSP partnerships, premium subs, the new Home Screen (first real refresh in 10+ years, rolling to 100M+ households), Roku Channel engagement, Ads Manager adoption, and international. Disciplined cost discipline layered on top. Jefferies ($150) and Piper ($148) echo the political ad / World Cup upside into 2H26. The cluster is now $148-$170 — a tight range that says the Street's largely aligned.
The new Home Screen is the under-discussed one. Quick Access + expanded Top Picks is a real product change, not a paint job — changes the ad inventory and engagement surface. Pair that with FOX One at $19.99/mo (including all 104 FIFA World Cup 2026 matches) landing on The Roku Channel, and you have a credible premium content wedge. Political ad dollars 2H26 are the near-term kicker if it materializes.
"meaningful upside in shares at current levels based on its unchanged core thesis on the company" — Citizens
At 76% YoY and near the high, the easy money is in. $1B FCF by 2027 is a high bar that requires both top-line acceleration and margin expansion holding simultaneously. Streaming ad market is getting crowded and CPMs are not guaranteed. Not sure we can call this anything but a momentum name into 2H catalysts — if political spend disappoints, the multiple compresses fast.
Unity ripping — UP 20% IN THE PAST WEEK, now $32.17, and Piper just initiated at OW/$40 (24% upside). Not a solo call though — Needham ($40), Oppenheimer ($38), and BTIG ($43) all moved PTs higher in recent weeks. The cluster is $38-43, Piper slots in mid-pack.
This is a 2027 EBITDA inflection story. Piper models $776M EBITDA vs $71M LTM — that's a 10x ramp, and they explicitly flag top/bottom line as conservative. Vector (their AI tool) is driving results WITHOUT proprietary data, so new data sources = the next leg. Grow rebuild looks legit. AI risk = overblown in their view, and the stock is still down 20%+ from the Genie update in January.
"The rebuild of Unity's Grow business is impressive and we expect continued execution. Vector success is currently driven without proprietary data, making new data sources an area of interest."
That Vector angle is the underappreciated piece — they're getting paid on AI without the data moat yet. Once that data layer comes in, the moat narrative gets real.
Q1 was fine, not fireworks: revenue $508M BEAT $505M, but EPS $0.23 MISSED $0.24. Strategic revenue +35% YoY to $432M, adj EBITDA margin expanding. Mixed bag but the forward setup is what matters here.
For PMs who already own: the Genie-scare dislocation is healing, multiple analysts turning bullish simultaneously, and the AppLovin halo is real (Piper explicitly cites AppLovin's success as a tailwind — competition strengthens Unity's value prop for the long tail of publishers). The bear case is duration — the stock just ripped 20% in a week into a cluster of PT raises, and InvestingPro's fair value flag suggests overvaluation at current levels. Buying the rip into a 10x-in-2-years EBITDA model is the trade; just know you're paying for 2027 today.
Verdict: DBX is a capital return story now, not a growth story — and the tape is finally pricing it like one. $1.7B of total buyback authorization against a sub-$7B market cap is a real number. That's north of 20% of the float potentially coming back, and mgmt reiterated on the Q1 call that they plan to exhaust authorization in 2026. Stock +6% on the day vs IGV +4.8% — buyback headlines doing the work.
RBC stays Outperform, $32 PT, noting shares trade below InvestingPro's $35.01 fair value. Hard to argue with the math, and the street is going to have to mark up models for buyback accretion.
Q1 was clean but not the catalyst. EPS $0.76 vs $0.73 Street, rev $629.5M vs $615.92M. Solid beats, mid-single-digit growth — enough to support the capital return narrative, not enough to change the multiple on fundamentals alone. Buyback math is what moves the stock from here.
Andrew Houston is stepping aside as sole CEO. Ashraf Alkarmi joining as Co-CEO May 26, then sole CEO with Houston as Executive Chairman. Internal transition, not a blow-up. But the read-through is Houston — the founder, the vision guy — is moving to oversight. The bull case needs Alkarmi to keep the product roadmap humming. Watching the next 2-3 quarters of product velocity is the real tell.
Bull: FCF gusher returning capital aggressively, valuation reasonable, $1.7B authorization is real EPS accretion. Own the cash cow.
Bear: Topline is mid-single-digits. Multiple expansion is capped without a re-acceleration narrative — and a founder stepping back isn't typically what sparks one. This is a story about how fast they can buy back stock, not how fast the business grows.
Not sure we can read too much into a single session's outperformance on a buyback headline, but the setup heading into year-end (when authorization burn should accelerate) is constructive. Trim into rallies, don't fade the buyback bid.
Small-cap battery spec with three end-market shots (drones, smartphones, eyewear) all inflecting at once, and the stock just ripped 19% in a week on a Q1 beat + military FEOC angle. Trade, don't anchor a position.
The print: $7.6M rev vs $6.95E, non-GAAP loss of $0.14 vs $0.16E. Beat, sure — but on a base so small the beat doesn't really matter. LTM rev growth of 50% sounds great until you realize the run-rate is still under $35M. This is a narrative stock right now, not a fundamentals stock. (At least for another 4-6 quarters.)
What actually moves the needle from here is the catalyst path, and it's getting crowded in a good way:
Momentum trade with real optionality underneath. The 19% one-week move means chasing is dangerous — wait for a flush into the $7-8 zone if you want exposure. Don't size it like a conviction long at $30M run-rate rev. If the smartphone qualification framework gets adopted more broadly across OEMs, this re-rates fast. Until then, it's a 1-2% sleeve at most.
"the company has aligned with its lead smartphone customer on a silicon-specific qualification framework, and another customer acknowledged the limitations of the legacy 0.7C cycle-life test for silicon batteries"
That qualification reframing is the actual alpha here — if it goes industry-standard, ENVX has the inside track. 2027+ monetization story though, not a Q3 catalyst.
D.A. Davidson's Bison framework has BOX at 10/12 — they're calling it under-appreciated AI winner, $45 PT, +21% upside from $26.96. The math is straightforward: stock trades ~10x CY26 FCF, target ~18x. That's an 80% multiple expansion priced into a name that just printed a real quarter.
BOX didn't just beat — it accelerated. REV +10% CC, BILLINGS +13% CC (vs implied guide of 6-7% — big beat on the leading indicator). EPS $0.37 vs $0.36. Enterprise Advanced driving seat expansions AND pricing premiums, which is the combination you want to see in a SaaS comp.
Bull case (DAD, $45): Best-of-Breed Bison name. 56% GROSS MARGINS. Content + AI workflow moat. Reasonable valuation at 10x FCF for a name with double-digit growth accelerating.
Bear case (UBS, $29 Neutral): Same data, but UBS only took PT to $29 — implied ~12x FCF multiple. They're not buying the AI re-rating story, treating it as a normal SaaS comp.
The spread tells you everything — $16 of PT delta between two firms looking at identical numbers. DAD's framework is the differentiator (the Bison checklist). If you buy the moat thesis, you buy DAD's multiple. If you don't, you're at $29 with UBS.
BOX is small-cap, low-float, low-coverage. Q1 was the first real "credible AI" print, not just narrative. Worth a swing if you have the risk budget — the $29-$30 area is the bogey if this narrative builds. Risk is the print gets priced and the stock chops into the next quarter without fresh catalysts.
The trade is played. The call works because of what comes next.
POWI sits at $81.91 — basically AT Stifel's $82 PT after a 139% run in six months. So this isn't a price-target story anymore, it's a "do you believe the 800VDC ramp" story. Q1 was a clean beat ($0.25 vs $0.23, $108.3M rev vs $106.6M) but the stock still dipped aftermarket, which tells you the bar is high and the marginal buyer needs the DC narrative to keep paying.
Current per-rack content opportunity is ~$500 for aux + main power combined. The transition to 800VDC could push that 4-5x — call it $2,000-2,500 per rack over time. Stifel's framing was honest:
"This represents an opportunity set rather than committed revenue at this stage."
That's the right hedge. Supplier decisions land in 2027, meaningful revenue 2028 earliest, and management explicitly said winners get picked ITERATIVELY across successive 800VDC generations — so the supplier roster can churn iteration to iteration. Not exactly "land and expand" — more like "win the round, fight the next one."
Bull: Power semis are scarce, hyperscaler DC buildouts are accelerating, and even capturing a sliver of 4-5x per-rack content on 800VDC is massive relative to POWI's $4.56B market cap. The 400VDC socket is the nearer-term proving ground — aux power today, main power design wins in progress.
Bear: Stock is up 139% in 6 months, sitting at the PT, and the actual 800VDC revenue is 2+ years out with iterative supplier selection. Q1 beat got sold. The 400VDC main power design wins are the bridge — if those don't land, the 800VDC optionality trades off the table.
Tape's pricing in 800VDC already. Need to see 400VDC main power design wins materialize to bridge to the 2027/2028 setup. Otherwise you're holding a name that just did its move.
META — MOST INCREMENTAL GW IN 2026 of any AI lab. Combined with MSL team ramp, this is the first public signal META could be a legitimate 4th frontier challenger in 12 months. Re-rates the frontier lab market from 3-player to 4-player (structural, not trade). PJM/ERCOT queue pressure compounds — META's 2026 capex is the single most direct trade on US grid strain. $632 PT listed; positioning as both hyperscaler AND frontier lab is unique in mega-cap.
CRWD — Q1 EPS $1.10 BEAT, revenue beat, ARR guide raised to $6.53-6.56B, 4-FOR-1 SPLIT. POST-PRINT -9% ON BILLINGS MISS. Billings miss in SaaS is a harder forward signal than revenue miss — forward RPO at risk. AI attack surface thesis tested for first time: demand is healthy, just decelerating into longer pay-as-you-go cycles. Unit economics changing — clients getting more AI automation for same spend, forward multiple compresses.
GOOG — $80B → $85B equity offering UPSIZED, Berkshire $10B ANCHOR CONFIRMED. One of the largest tech equity raises ever. The real signal is why equity over debt — mgmt views stock as expensive AND capex too large for FCF. GCP gained 6 SHARE POINTS in 4 years (16% → 22%) while AWS lost 7 — most under-appreciated share gain in mega-cap. TPU program + Anthropic's Apollo/Blackstone $36B debt facility = vertical AI infra integration.
MU — PT RAISED TO $1,600. Squeeze-to-$1K by 6/24 earnings trade discussed. Part of dot-com 7 high-fliers up avg 158% YTD, +$1.7T combined mcap. GETTING CLOSE TO $1T MARKET CAP. DRAM/SSD contract prices +200-300% since mid-2025, MLCC hoarding, ABF substrate +20%, Intel CPU +15% — component cost cascade. 2H26 PC ODM production cuts 15-18% YoY. Cycle is bifurcating — AI pulls supply from consumer, consumer breaks.
AMD — MI300X secondary market OPENING: $32K eBay OBO, $1.99/GPU/hr rental, stablecoin payment accepted. Crypto rails as AI compute access rails = new market structure (fragmentation, demand-elasticity testing, possible over-supply at certain hyperscalers). China = ~20% of AMD revenue; Lisa Su keeping LOWER PROFILE in China vs Jensen. AMD is the SECOND-BEST China play in AI semis. MI300X DRAM demand proxy: 1M agents = 32 petabytes.
AAPL — iPhone-connected smart glasses DELAYED TO LATE 2027, $200-$500 mid-tier target. Apple TV/HomePod mini waiting on new Siri. iOS 28 = 'BELL' / macOS 28 = 'POPPY' ('BOPPY'). iOS 27 = 'RAVE' with Siri app syncing chats across devices — finally a frontier-class consumer AI interface, but ~2 YEARS LATE vs ChatGPT. Distribution advantage still in place.
CRWV — FIRST NVDA VERA RUBIN NVL72 RACK DELIVERED. 72 Rubin GPUs, 36 Vera CPUs, 3.6 exaFLOPS FP4 inference, 75TB HBM. CRWV is the anchor neocloud deployment for next-gen architecture. $108 PT. Anthropic closed 1GW gap with xAI via sublease — demand stacking, not digesting. CRWV is highest beta to AI capex squeeze with most leveraged rental economics.
SNDK — Q1 2026 revenue $5.95B (+251% YoY), production cost $1.288B (-2%), GM 78.3% vs 22.5% prior. Growth is ENTIRELY PRICE, NOT VOLUME. Cleanest print on memory pricing power. "Sell more, earn less" reversion expected — bull case needs NAND tightness through 2028 (per GS), bear case is structural GM compression to 40-50% as BiCS 8 ramps. 'When, not if' mean-reversion trade.
INTC — 18A yields rising fast. Panther Lake shipping at 7X VOLUME this Q on 50% yields. Internal demand rising so external supply LIMITED to Apple/Google/Nvidia/AWS strategic. AMD and other external foundry customers get SQUEEZED. If confirmed at Intel earnings, this changes the foundry competitive picture. CPU +15% cost pressure = bifurcation: cost inflation victim on consumer, pricing power winner on foundry.
TSM — NVDA >20% OF TSMC REVENUE 2026, CONFIRMED #1 CUSTOMER (overtook Apple 2025). Jensen confirmed on Jodi Shelton podcast. Scale-mover for NVDA-TSMC concentration risk — most concentrated single-customer exposure in semis. 18A yield 60-65%, improving 5-7% for 6 months. Packaging = $80B+/yr → $100B+ (hybrid bonding, 3D integration, EMIB = new Moore's Law). Counter: Huawei "logic folding" challenges EUV-centric moat.
LITE — InP revenue $600M (2025) → $9B (2030), 15X GROWTH. NVDA driving suppliers to ~12x capacity commitment vs 20x demand. STRUCTURAL SHORTAGE through 2030. LITE has scale, COHR has growth. Parabolic run ROLLING OVER — watch late summer/early autumn pullbacks to 40-week MAs. Positioning reset, not fundamental break.
COHR — InP revenue $125M (2025) → $4.3B (2030), FASTEST GROWING of InP names. 6" wafer + laser capacity = competitive moat. Report top pick for InP exposure. Quantum interconnect bottleneck shift = long-duration structural read-through to photonics from new end market ($2B CHIPS Act quantum foundry commitment). Highest-conviction, highest-growth name in TAM expansion.
AAOI — InP revenue $60M (2025) → $2.1B (2030). UHP CW CPO share ~10%. Report top pick alongside COHR. Smaller base but CPO exposure gives next-gen packaging optionality. Most volatile of InP names — highest beta to parabolic move. Watch 40-week MA pullback.
QCOM — Snapdragon C for laptops at COMPUTEX, $300+ price points, Acer/HP/Lenovo partners. Positioned as "Neo competitor" vs Apple M-series. Cristiano Amon keynote, 2026 = "year of agents." QCOM is the incumbent in ARM Windows. Debate: NVDA could do better given distribution + Jensen/Microsoft partnership, but ARM Windows needs ecosystem critical mass.
VRT — 2026 guide $13.75B revenue, +34% GROWTH. Heavy Field Services hiring = real-world deployment commitment. Direct read on AI DC power + cooling buildout. Schneider Electric/SoftBank Dunkirk partnership validates 800VDC + liquid cooling at European hyperscaler scale. $3.5M/MW AI DC anchor revenue. Closest pure-play to AI DC power bottleneck.
NBIS — Token Factory positioned around OPTIMIZATION, ORCHESTRATION, AGENTIC DEPLOYMENT, not basic GPU rental. Differentiation vs CRWV/IREN. $105 buy. Positioning for inference-bottleneck rotation. Margin compression risk from NVDA server rental costs is the structural headwind; Token Factory is the higher-value wedge.
IREN — $63 PT. BTC-mining-pivot-to-AI-compute play. Existing power infrastructure ahead of pure-rental neoclouds. SPV concerns same as broader neocloud trade. Higher beta than NBIS with less differentiation. Pivot narrative is the catalyst, not fundamentals.
LSCC — CEO: hyperscaler capex 2027 "GOING TO BE OVER $1 TRILLION... AND THIS NUMBER CONTINUES TO GROW." MOST DIRECT DATAPOINT on hyperscaler capex sustainability. FPGA play at the edge of AI server designs, less direct than AVGO custom silicon but real read on buildout pace.
TMHC — Berkshire buying Taylor Morrison for $8.5B all-cash at $72.50/share. Buffett's first BRK.B acquisition in a long time. Skeptical read: public homebuilder premium is a tough tape to time. Bull read: housing supply still constrained, builder margins attractive, potential Clayton prefab synergy on TMHC land.
ANTHROPIC — Confidential S-1 filed, Morgan Stanley + Goldman lead. Target valuation $965B — OVER OPENAI as world's most valuable AI startup. ARR $47B (vs $9B end 2025), $65B raise, profitable in Q2 expected. First trillion-dollar IPO before year-end alongside SpaceX. Apollo+Blackstone $36B debt for TPU purchase = first "AI lab as project finance" capital structure. Musk 180-day datacenter question is the first crack in revenue stability narrative.
PANW — First quarter with CyberArk integration. CRWD billings miss raises the bar but PANW = higher-quality comp. Platform consolidation thesis: PANW + CyberArk = most complete enterprise security platform. Needs to beat AND show synergies to extend the AI cybersecurity trade.
NOW — Bernstein $236, DA Davidson $190, Evercore $150 — WIDE DISPERSION = high uncertainty. Now Assist AI 2026 target RAISED from $1B to $1.5B. Enterprise workflow agent is a different product layer than Copilot. NOW positioning as "platform of platforms" for enterprise AI. $1.5B target = bull case anchor.
AKESO — Lung cancer drug late-stage trial: ~4-MONTH SURVIVAL EXTENSION, ~34% DEATH RISK REDUCTION. Major China-listed biotech catalyst. Healthcare cost deflation narrative is long-term headwind for AI healthcare trade, but AKESO data is near-term positive. Cuban read: "cell and gene therapy ain't cheap... 10x first before any deflation" — near-term AI healthcare trade intact.
RDDT — Reddit as AI INFRASTRUCTURE: human-generated conversation corpus as irreplaceable training data asset. Data-licensing revenue as PARTIAL HEDGE against AI disintermediation. Data moat vs OpenAI/Anthropic scraping. Thesis circulating per detailed thread.
NKE — Volume shelf setup per technical analysis. No fundamental catalyst in the feed. Watch list signal — momentum name with potential technical breakout.
ETH — At $2K. Stablecoin payment rail on AMD MI300X = AI-crypto link. Fed's Waller: "stablecoins broaden reach of US policy." Crypto rails as AI compute access rail thesis extends to ETH layer.
SPACEX — IPO priced $135/SHARE, $1.77T VALUATION, up to $86B raise. JUNE 12 LISTING. Musk >80% voting power. Top-tier institutional buyers (Altimeter, Sequoia, Apollo, Blackstone, Dragoneer). First trillion-dollar IPO alongside Anthropic. Musk 180-day Anthropic datacenter question = first crack in revenue stability narrative.
KOPN / MXL / GTLB / HQ / MTSI / GEV / HIVE / LIN / MRLN / UIS / VEEV / ACN / NN — No signal in the feed. Watch list only. (GEV ties to grid/EPC theme if anything prints; LIN to industrial gas for fabs; MTSI to RF/optical; GTLB to dev tools/AI agents; VEEV to vertical SaaS.)
Hearing META's MSL team is the same group that built Llama under friction — internal read is they're closer to a 4th-frontier challenger than the Street is pricing. If 12-month thesis holds, MSFT/GOOG/AMZN have to re-rate frontier exposure.
Word is Anthropic's ARR $47B is heavily concentrated in API + Claude Code — enterprise seat growth is the underappreciated line item. S-1 formal disclosure will be the test on customer concentration.
Channel checks suggest Apollo/Blackstone $36B TPU debt facility is structured as a take-or-pay with Google as ultimate credit support — not a true project finance, more of a vendor financing shell. If true, GOOG has more TPU revenue risk on the books than the equity raise implies.
Hearing META's captive/off-grid strategy is a forced response to PJM queue lead times exceeding 5 years in Northern Virginia — they're effectively bypassing the incumbent utility model. Direct read for behind-the-meter infra.
Word is NVDA's $32K MI300X secondary market pricing on eBay is BELOW what some hyperscalers paid on contract — implies neocloud/secondary arbitrage is real and hyperscalers may be over-paying for capacity they don't fully utilize.
Channel checks on Intel 18A: 50% yield on Panther Lake is at-pad yields, not high-volume manufacturing yields. Real external foundry yields likely 30-40%. If 18A external customers get squeezed, AMD/MediaTek/others will reroute to TSMC — TSMC 2027 mix shifts bull case.
Hearing xAI struck 90-day reclaim optionality on a short-term capacity lease — first public sign that even xAI is hedging on multi-year build commitments. Supply normalization is being priced in by the largest buyers.
Word is SoftBank's €75B / 5GW France deal is partially funded by Saudi PIF — if true, the "European AI sovereignty" framing is a cover for Gulf capital deployment. Schneider Electric partnership is the only verifiable domestic angle.
Hearing that Anthropic-SpaceX datacenter agreement is indeed ~180 days, not 3 years as S-1 implies per Musk's public pushback. If S-1 disclosure confirms, it's the first material miss in the Anthropic revenue stability story and casts shade on the broader "AI lab as utility" cap structure.
Channel checks: HBM tightness elongating into 2028 (per GS lift on SK Hynix/Samsung 2027-28 OP) is real, but it's SK Hynix-specific. Samsung qualification at NVDA is still behind — any memory trade should be SK Hynix pure, not basket.
Word is Reddit's data licensing revenue is ~$200M+ run rate and growing 100%+ YoY — not yet a hedge against AI disintermediation but enough to reframe the platform as infrastructure. RDDT thesis is early.
Hearing that TSM packaging backlog extends to 2028 — CoWoS-L and FOPLP both oversubscribed. This is the structural moat against Intel foundry encroachment. Glass core substrate at ~2028 production is the next packaging bottleneck (could be the answer when ABF/CoWoS plateaus).
Channel checks on optical parabolic rollover: LITE/COHR/AAOI are testing 40-week MAs after a parabolic run. Most institutional positioning is momentum-CTA, not fundamental. A 10% pullback would force de-grossing. Setup favors tactical short into late summer if the trend breaks.
Word is VRT's $3.5M/MW AI DC anchor revenue is conservative — Schneider Electric's Dunkirk cluster at 800VDC + liquid cooling is implying $4-5M/MW at European scale. EU energy cost premium shows up in DC revenue.
Hearing MU's 6/24 print needs to show DRAM contract prices still rising into Q3 — flat-to-down would break the squeeze-to-$1K trade. Trade is binary on the print direction, not magnitude.
Channel checks on QCOM Snapdragon C: $300+ price points for Windows on ARM is a small TAM today, but Acer/HP/Lenovo commitment is real. 2026 volume likely 2-3M units — not enough to move the needle, enough to validate the architecture for 2027 ramp.
Word is Berkshire's Taylor Morrison deal is at a meaningful premium to NAV and Buffett's team is targeting builder landbank for long-duration housing exposure. Prefab Clayton on TMHC land is synergy noise; the real play is the landbank.
Hearing NVDA's concentration at TSMC is now a Board-level risk item — Jensen has personally pushed for second-source qualification at Samsung 2nm. If Samsung 2nm qualifies in 2027, TSM has a 5-7% revenue at risk. Watch 2H27 earnings commentary.