Good morning.
Index futures drift green but nobody’s leaning on the tape — 30Y kissed 5% and pulled back, DXY at six-day lows, Friday’s jobs print is the whole game. SNOW dominates the premarket again: +24% DAY TWO, hiked FY product revenue guide to $6.07B vs $5.86B street, product revenue +37%, and AI drove ~50% of the growth acceleration. Soitec’s the only tech name on Europe’s movers board — raised guidance to ~50% constant-currency growth, same bottleneck story one layer earlier in the stack. AVGO’s cash-session selling persists: FY27/FY28 AI books ($115B/$230B) are now base case and the market’s anchoring on Q4 revenue trailing the Street. Asia closed messy — Korean memory sold off on Lutnick tariff chatter, though TSM’s ~20 fabs under construction and equipment demand nearly doubled since year-end remind you capacity, not demand, is the binding constraint. Macro: the yen’s run is pure BOJ repricing — no intervention prints in Tokyo’s accounts — and the market’s pricing a 25bp hike this month with acceleration optionality. That makes carry unwind the tax on gross leverage into Friday.
Three threads. Software first: SNOW’s cash-session hold decides squeeze vs repricing — hold better than a one-third giveback and software migrates from funding source back to ownable. TDC’s CEO compressed the bear case for legacy DB moats into one quote (migrations now in 3 quarters vs 2-3 years) — MDB/ORCL carry that overhang today, and the gross-margin dilution read-through punishes narrow SaaS (CRM/NOW/ADBE/HUBS) harder than the scaled platforms (MSFT/GOOGL). Second, semis entered the expectations game: AVGO is the first crowded long forced to beat the buy-side, and the bear wedge is dual-sourcing — MediaTek on TPU training, Marvell-Google — while on NVDA, BofA's bottoms-up (~2x demand vs supply) puts FY28 +70% as the floor. Third, the bottleneck moved from who-has-the-chip to who-has-the-power-and-politics: KEPCO wants Samsung and SK hynix to prepay KRW 25T (~$18B) of electricity to fund the grid — power is now an allocation resource, and that extends the semicap and electrification complex (AMAT/LRCX/KLAC, VRT/ETN). Memory confirms the same inflation at end demand: Acer’s CEO says PC units decline this year on storage-price-driven BOMs — ASP up, units down — while SNDK tries to rebrand NAND as an inference-tier hardware play. Watch CXMT at 10% DRAM share and the tariff chatter for cracks in that pricing.
We'll hit up SNOW and AVGO first — the two positioning battlegrounds — then get to semicap/power: TSM, ASML, AMAT, VRT.
AVGO just did what megacap semis almost never do: put a hard two-year AI revenue stack on the table and dare the street to catch up. The numbers — FY27 AI semiconductor revenue $115B (raised from >$100B) and FY28 at $230B, implying total FY28 revenue around $285B and FY28 EPS "exceed $30" — are 10-15% above consensus before the print even gets modeled. At $367, that's roughly 12x the company's own FY28 number. Not the street's.
The quarter itself was a side show: July EPS $3.32 vs ~$3.21 consensus, revenue $29.59B (+86% y/y), record FCF $13.7B. October guide ($34.8B revenue, 66% op margin) basically in line. So no, the stock isn't moving on the print. It's moving — or not moving — on the durability of the FY28 target.
GOOGL's TPU shift to MediaTek was known. Google's share of the incremental custom ASIC program falling from ~80-85% to ~60-65% was known. The stock already corrected ~25% from its high on that exact narrative, while the SOX is up ~59% YTD. That risk is priced.
What's new is the customer mix under the hood of the $230B. Google is no longer the anchor tenant. By FY28, AVGO expects Anthropic to be its largest XPU customer, OpenAI second (5GW+ of Jalapeno and successor), Meta third at ~3GW, and Google — still a "multi-tens of billions" per year TPU customer through the decade — effectively drops to #3 or #4 in the lineup. That's the real de-risking event. Single-customer concentration just became a six-customer story.
Also new: actual gigawatt disclosures. Demand is 10GW in CY27 and 20GW in CY28, per management. The guide is supply-constrained, not demand-constrained. Limiting factors are land, power, shell readiness, substrates, CoWoS, and HBM — not willingness to buy.
The bull case is brutally simple. AVGO is the only vendor shipping custom XPUs at hyperscale across multiple architectures — TPU v8i, Jalapeno, MTIA — and the FY28 number is backed by named customer commitments, not hope. Cantor puts the stock at 17x its CY28 EPS estimate of $36, calls it a Top Pick, and says the constraint is purely physical.
"Severely supply-constrained in fiscal 2027 across substrates, wafers, CoWoS, and HBM." — Cantor Fitzgerald
Macquarie made the boldest move — upgrade from Neutral to Outperform at $490 — and they did it for the most honest reason: the Google bear case is now fully public, partially confirmed, and sitting in the price. Their model has Anthropic alone buying >$40B from AVGO in FY28, which more than offsets the Google loss.
The bear case is about who's actually paying. Anthropic and OpenAI — two non-public, loss-generating companies — drive the marginal $100B+ of that FY28 guide. TD Cowen said out loud what every PM is thinking:
"Those two customers driving a significant share of fiscal 2028 growth is likely to be a source of skepticism given concerns around circular financing in AI infrastructure build-outs." — TD Cowen
The counter: the hyperscalers funding Anthropic and OpenAI are the same ones signing the lease agreements. And 20GW of demand doesn't disappear because one customer's financing structure gets messy. But if you think the $230B is partly "pretend revenue" that the ecosystem is financing into itself, this is where you fade the stock.
Post-print targets span $475–675, with the dense cluster at $490–600. Low end: TD Cowen and Raymond James at $475 — both still Buy/Outperform, so even the "cuts" are really just valuation discipline. High end: Baird at $630, Cantor and Rosenblatt at $600. Bernstein at $575, Mizuho $530, Barclays $500, Morgan Stanley $505.
Not one note out this morning sits below Overweight. Macquarie came up from Neutral. The debate isn't direction — it's how much of the FY28 number you believe, and what multiple you apply to it. Morgan Stanley's $505 is arguably the most nuanced: they think calendar 2027 AI revenue expectations are too high, but AVGO's share of its serviceable market is too low. Over 80% share over time, per their industry checks.
This print is a gift to the custom silicon complex. AVGO just validated a 20GW custom XPU TAM for CY28 — that's a direct read-through to ALAB and MRVL, and a warning to anyone who thought custom ASICs were a single-customer fad. The zero-Google narrative is dead. The story is now multi-tenant, multi-generational, and physically constrained.
Watch the supply chain names more than the semi comps. CoWoS, HBM, power delivery, and shell capacity are the binding agents. If AVGO is right that demand runs ahead of a $230B guide, the bottleneck trades — not the chip designers — are where the incremental alpha sits.
Bottom line: at 12x the company's own guide, with the Google overhang mostly washed out and every sell-side desk still modeling below management, the risk/reward skews long. You just have to stomach two years of quarterly noise to collect.
SNOW put the AI-monetization debate to bed last night. Product revenue +37% y/y, third straight quarter of acceleration, first-ever back-to-back >5% product beats, FY27 guide raised ~500bps — and management guided margins UP in the same breath. The 84% six-month run means this is a crowded long; the easy money sits in the book. But the rate of change is still the strongest in software, and the Street's revised target cluster leaves room for the tape to run.
Total revenue $1,547M, ~4% over consensus. Product revenue +37% y/y (THIRD CONSECUTIVE QUARTER OF ACCELERATION, off the 30% trough in FQ4'26). EPS $0.62 vs $0.45. NRR: 126%. Solid. Not the story.
Margin is the story. Non-GAAP op margin ~15%, +3.4pp q/q — the LARGEST SEQUENTIAL INCREASE in over three years. Product gross margin 75%, and management guides ~100bps of FY pressure from lower-margin AI workloads — yet they RAISED FY27 op margin guidance 100bps to 14.5% anyway. Revenue upside, disciplined hiring and AI-enabled productivity more than offset the mix drag. That's operating leverage you can model, not hand-wave.
Guidance did the heaviest lifting. Q3 product guide: 37-38% — MORE THAN 6% ABOVE THE STREET — another sequential step-up. FY27 product revenue guide: $6.07B, +36% (includes ~1pt from Observe). Implied Q4: ~35% midpoint; if SNOW keeps beating the way it just did, the exit rate prints ~40%. Growth re-accelerating AND margin guide rising in the same quarter is a rare combo in software. SNOW just posted it.
Ten firms hiked post-print. PT range: $370 (Rosenblatt, still a Buy) to $488 (Citizens). Heavy cluster: $425-470. Morgan Stanley swung hardest — $470 from $300. Guggenheim holds the lone Neutral; they concede the quality, they won't chase the multiple.
The collective thesis is unified — and it's a positioning change, not a data-point change. AI is now roughly half of the acceleration, tied to named products: CoCo at 9,100 accounts, CoWork at 5,800, both bringing NET-NEW users into the platform. Core data cloud consumption does the other half, with AI-assisted migrations pulling forward legacy data warehouse displacement. New logos AND faster existing-customer expansion at the same time is the flywheel the bulls have been waiting for.
"AI adoption is creating a flywheel effect at Snowflake, with a meaningful step-up in AI revenue complemented by sustained core data platform strength." > — Needham
"Product revenue outperformance of 5.2% exceeded the trailing three-year average of 3.9% and marked the first time the company posted back-to-back quarterly beats exceeding 5%." > — Deutsche Bank
Bull: SNOW just converted from AI narrative to AI model. Two engines firing — new AI users landing net-new workloads AND AI-assisted migrations accelerating core platform displacement. Then there's the margin delivery: largest sequential op margin expansion in three years, with guidance up another 100bps. On MS's math, $470 is ~22x EV/CY27 revenue — a discount to the high-growth AI software cohort (CRWD, NET, PLTR), not a premium. If the Q4 exit approaches 40% with 15% op margins, these PTs are waypoints, not ceilings.
Bear: The tape front-ran the story. +84% in six months means everyone already owns the reacceleration — the incremental buyer has to underwrite $425-470, not $300. The comp math flatters the acceleration (lapping a 30% trough quarter). NRR is 126%, steady, not expanding. ~1pt of the guide is Observe, not organic. Product gross margins go the wrong way — down ~100bps as AI mixes lower — which is exactly the opposite direction you want from a stock at 20x+ forward revenue. Guggenheim's Neutral is the valuation conscience here: spectacular operating quarter, but the market is being asked to pay for several years of flawless execution.
New: First-ever back-to-back >5% product revenue beats. FY27 guide up ~500bps. Margin guidance UP despite AI gross margin drag. Named-product AI attribution (CoCo, CoWork, AI Functions) with hard account counts. AI-assisted legacy migrations — a displacement vector, not just an upsell. The combination of growth reacceleration AND margin expansion in one print.
Known: The reacceleration narrative was already in the tape (30% → 37% over two quarters). The AI product suite was already a bull case. The Observe contribution (~1pt) was already announced. The +84% move told you the market believed.
What's genuinely incremental: this is the first print where SNOW showed AI-driven growth and AI-driven operating leverage simultaneously. That combination is why PTs went from a $300-350 cluster to $425-470 in one night.
SNOW is the data-infrastructure proof point for the AI software complex. PLTR proved AI demand. CRWD proved AI security spend. SNOW just proved AI data consumption — and it's accelerating, not just growing. That's a positive read-through to consumption-model software broadly and a negative for legacy on-prem data platforms losing share to faster AI-assisted migrations. MS's explicit framing — SNOW at a discount to CRWD/NET/PLTR — means if SNOW holds its post-print multiple, the whole high-growth AI software group gets a bid.
PANW keeps doing the one thing that matters — printing numbers that back up the platform narrative. FQ4 revenue $3.41B (+34% YoY), non-GAAP EPS $1.02 (four cents above the top of guidance), NGS ARR $9.1B (+63%, CyberArk inflation included). The street response is near-unanimous: 10 of 11 firms we track are Buy/OW. PTs now span $384-475, with the bulk clustered at $400-425 and RBC the outlier bull at $475 (up from $434). Stock at $329.66 — after a +90% year. The print was the easy part. The FY27 guide is the battleground.
But the guide has a crack in it. UBS calls FY27 NGS ARR growth of 22-23% "slightly below street expectations." BMO says management guided above consensus. Both can't be right — unless the debate is really about the organic/inorganic split. DA Davidson flags it directly: implied organic NGS ARR growth in the FY27 guide looks ~flat y/y, and that's "the main pushback from investors." Management didn't break out inorganic contribution. That omission tells you exactly how they view that debate.
Bull: This is a beat-and-sandbag machine operating inside a demand supercycle. DA Davidson's kicker: initial FY26 guidance implied just 10-18% organic NGS ARR growth, and PANW printed ~29%. History says FY27 guidance is conservative too. RBC expects "consolidation momentum to build" with upside to consensus as the year progresses. Post-Mythos security demand adds near-term fuel, cross-sell with CyberArk and Chronosphere is ahead of plan, and a 40% FCF margin target by FY28 gives the multiple a path to reason.
Bear: The 63% NGS ARR headline is M&A math. Organic is 29% — solid, but the guide implies that rate goes flat, and nobody will tell you the organic number straight. UBS is the one firm willing to say the FY27 outlook missed street. The valuation debate is almost comical at this point: 289-314x trailing EPS. Even StoneX — a Buy — flags InvestingPro's fair value framework showing the stock overvalued at current levels. The stock ran +90% BEFORE this print. The r/r at $330 is fine. The r/r at $475 (RBC's PT) assumes everything goes right for two more years.
Ten buys, one neutral (UBS at $390). PT range: $384 (JPM, the most conservative bull) to $475 (RBC, the outlier). Consensus lands ~$410-415 — roughly +25% from here. The old PT cluster was $345-355; the new one is $400-425. That's not an incremental move — that's a repricing of the platform thesis in a single print. JPM keeps it simple: ARR and revenue accelerated on CIO urgency, platform momentum, and M&A synergies ahead of plan. Every one of these PTs assumes management sandbags. History says yes.
"Broad Strength Continues; PANW reported a solid quarter against high expectations. NGS ARR upside was similar to last quarter, and we highlight that both legacy/core NGS solutions and CyberArk/Chronosphere contributed to the outperformance. Further, management is guiding to FY27 NGS ARR above our prior estimate/consensus, and we envision upside tension driven by strong post-Mythos security demands and cross-sell opportunities with CyberArk and Chronosphere." — BMO
"F'4Q was a strong finish to the fiscal year, with upside across all major financial metrics and continued progress across integrating the recently acquired businesses." — Piper Sandler's Rob Owens
Three themes spill over to the group from this print:
AI security attaches to existing platforms, not new ones. Prisma AIRS at ~$120M ARR in short order validates the CRWD/ZS AI narratives — but also validates the "bigger vendors win" thesis. Watch the pure-plays.
Post-Mythos demand is real and event-driven. BMO is explicitly modeling "upside tension" from security spend after the incident. That's a sector-wide budget flush, not a PANW-specific story.
Consolidation is now a strategy that compounds. CyberArk (identity) + Chronosphere (observability) pulling ahead of plan proves the platform trade. Reads through to FTNT and the broader "security spend consolidates to fewer, bigger vendors" narrative.
The honest take: PANW is the best house in the security neighborhood, and the guide is built to beat. The only real question left — organic vs. inorganic quality — is the question management is least interested in answering. Today isn't "is PANW good." It's "is good enough at 289x trailing earnings?" So far, yes.
VERDICT: GOOD QUARTER, HORRID TAPE. MDB beat, raised, and still got taken out back — stock down ~16% to $380 after printing Atlas at +28.9% YoY. Fifth consecutive quarter of ~29% Atlas growth. The problem isn't the quarter. The problem is the market priced this for ACCELERATION and got CONSISTENCY. When you pay a re-rating multiple for an AI inflection, stability feels like failure.
The bogey was clear going in. Mgmt had flagged a "200-300bps potential upside" scenario vs internal expectations for Q2. Atlas beat by ~290bps — HIGH END OF THE RANGE — and it still wasn't enough. Atlas growth decelerated ~50bps from 29.4% last quarter. The tape read that stripe as a ceiling, not a floor.
Everything else was fine. Actually better than fine. TOTAL REVENUE +25%. NON-ATLAS SUBSCRIPTION REVENUE +36%, LIFTED BY NEW AI FUNCTIONALITY. CRPO ACCELERATING. NRR IMPROVING. FY27 GUIDANCE RAISED. 73% GROSS MARGIN. NET CASH. AI LABS AND AI NATIVES CALLED OUT AS BUILDING MOMENTUM. Then the stock falls 16%.
(Context: MDB had run ~34% in six months into the print, brushing $473 — the 52-week high. Beat-and-raise in AI land normally works. But when the core metric doesn't inflect, you de-rate. Fast.)
Nine firms printed post-earnings. FOUR CUTS, THREE RAISES, TWO REITERATED. PT range $410-$540. CONSENSUS ~$472 — which, notably, sits right at that 52-week high the stock just backed away from. At $380, the stock is implying ~24% upside to the average target. That gap IS the whole debate.
The cuts mostly came from the high PTs, and they were valuation calls, not thesis breaks: RBC $515→$465, BMO $500→$450, UBS $460→$410, Stifel $475→$465. The raises came from the impatient and the patient alike: Piper $400→$470 (biggest dollar raise), Monness $460→$500, Bernstein $449→$484. Cantor holds the high water mark at $540. Davidson sits dead middle at $465.
The UBS cut to $410 Neutral is the bear floor: good business, stable growth, no AI boost in the numbers.
Bulls say the selloff is the setup. Five quarters of ~29% on a growing base is durability, not stagnation. Atlas outperformed mgmt's own internal scenario by 290bps. The leading indicators — cRPO accelerating, NRR inflecting, AI labs expanding use cases — point to re-acceleration in the coming quarters. Non-Atlas +36% proves the AI feature set monetizes outside the core engine. The "when, not if" crowd is doubling down at $380. Piper is the most explicit:
"We view the inflection in Atlas growth as a question of when, not if. We would be buyers of the pullback." — Piper Sandler
Bears aren't shorting the company, they're shorting the multiple. Five quarters of ~29% with a 50bps decel doesn't justify an AI re-acceleration premium. The beat-and-raise was real but undersized relative to the expectations mgmt and the street had built. BMO says it best — while keeping an Outperform, interestingly:
"The size of MongoDB's quarterly beat and guidance increase may have been below elevated investor expectations." — BMO Capital
Hyperscaler database alternatives keep improving and macro is still choppy. UBS is the cleanest expression of the bear case: stable, no AI boost, Neutral.
Nobody is negative on MDB the company. They're negative on MDB the stock at the growth multiple without proof of the AI step-function.
1. Non-Atlas is the sleeper narrative. +36% YoY, driven by new AI search/advanced features. Mgmt calling non-Atlas ARR "durably double-digit" is new language — that's an AI monetization path that doesn't run through Atlas. 2. AI labs and AI natives are now named drivers. That's incremental. These are workload wins with real consumption curves — land-and-expand that hits the P&L two to three quarters out. 3. The guidance raise was Atlas-driven. So the second-half setup hinges on Atlas holding ~29%, at minimum. Any dip below that band in Q3 guidance and this stock tests lower. 4. Upcoming catalysts: Investor Day Sept 29, Mongo.local NYC Sept 30. Back-to-back. Watch for a refreshed AI TAM framework or forward-looking pipeline disclosure. That's the next narrative inflection point.
MDB is the latest casualty of the tape's AI bifurcation. Infrastructure with AI revenue IN the P&L is printing 35-50% growth and re-rating. Software with "AI optionality" — MDB today, SNOW and DDOG at various points this year — gets punished for not inflecting NOW, even on a beat-and-raise. The market is no longer paying for when-not-if. It's paying for what's already in the numbers.
The longitudinal setup here is actually decent. If Atlas re-accelerates in Q3 or Q4 — AI labs scale, non-Atlas keeps compounding — this stock goes from $380 to $470 quickly. The street targets are already there. The brokers who cut to $465 this morning will be raising again the moment the inflection shows. Until then, expect chop. The Sept 29 Investor Day is the next real signal.
EIGHT FIRMS RAISED TARGETS. ONE REFUSED. STOCK DOWN 8.4% ON THE WEEK ANYWAY. Dell's Q2 FY27 beat-and-raise triggered a wall of PT hikes from $499 (Morgan Stanley) to $735 (Melius), median around $588 — and the stock still slid. When a record backlog print meets a weekly selloff, that's positioning, not disbelief. This is a $282B market cap company up 256% in a year. The marginal buyer has left the building. (For now.)
The revision wave clusters in three tiers:
The bulls own $617-735. Melius at $735, with gross margin 390bps ABOVE consensus — the standout number of the cycle. Bernstein at $650 on storage +26% and record profitability. JPMorgan at $635 (from $565) on AI momentum. Raymond James at $617 on AI sales doubling YoY. All four are paying for durability — multiyear contracts, broad demand, backlog visibility into FY28.
The middle — Piper Sandler at $558. Solid print, guidance above expectations, customers reallocating budgets to modernize IT workloads. No fireworks. No panic.
The skeptics cluster $499-505. Morgan Stanley, TD Cowen, Truist. Truist's move is the cleanest tell: it RAISED its PT 40% to $505 and still kept Hold. That's not business skepticism. That's multiple math at $435.
"A lack of material upside to the new target keeps its Hold rating in place." > — Truist Securities
"The 4% revenue beat was smaller compared to the previous quarter and consensus estimates had already embedded a beat into expectations." > — Brandon Nispel, KeyBanc
The non-AI business is doing the heavy lifting on profitability. AI brings revenue. Storage and traditional servers bring margin. KeyBanc questions how long enterprises sustain this spend level. And the 8.4% weekly slide says the market is asking the same thing. These bears aren't short the company. They're short the multiple. At $435 with 256% of last year's return in the tape, that's a legitimate position.
Dell tripling its AI-server guide is a supply chain event. That $74B number is GPU content, memory, storage, networking — Dell doesn't guide that high without allocation visibility upstream.
Best quarter in four years and the stock is up 72% in six months — that's the setup in one sentence. GTLB posted a 4.9% beat vs. guide, its largest percentage beat in four years, then raised FY27 by two points to 19% growth at the high end. And for the first time since 2024, DOLLAR-BASED NET RETENTION ROSE SEQUENTIALLY. Not bad for a company that was cutting headcount last quarter.
The problem: the tape already knows. Stock sits at $51.50, basically ON the low end of the new street range, near the $52.38 52-week high. This is now a valuation debate, not an execution debate.
Revenue $286M, +21.3% y/y, ~86% gross margin. Net new ARR accelerated over 40% y/y. Premium seats grew at the fastest clip in a year as AI coding agents pulled dev teams to scale. Ultimate tier upgrades hit their strongest level since 2022 on code cybersecurity anxiety. U.S. federal — the historically weak spot — came back a full quarter early, with renewals arriving ahead of schedule and above prior-year levels.
Then management did the honest thing: told you not to annualize it.
"After its best-ever pipeline conversion quarter, [management] does not anticipate a repeat of the second quarter and expects a more normalized pace of bookings." > — TD Cowen
That's the single most important sentence in the whole coverage. The beat is real, but part of it is pull-forward.
Broadly constructive, narrowly priced. PTs span $50 (UBS, Cantor) to $70 (Canaccord) with a cluster at $55-60. Median is roughly $55 — the stock at $51.50 is only ~7% below that. The actual actions this morning:
Bull case: This is the first clean proof the "one GitLab" post-RIF strategy works. Seats stabilized, SMB price sensitivity is easing, and AI usage (Duo Agent Platform, Flex pricing) is pulling premium content through. Net new ARR up 40%, Ultimate strongest since 2022, DBNER inflecting despite Flex accounting noise — the direction of travel is acceleration, not stabilization. Early Flex/consumption traction plus Fed improvement gives FY28 a real reacceleration setup. TD Cowen called it one of the strongest quarters since the new CEO took over nearly two years ago, with early dividends from AI-generated code landing on the DevSecOps platform.
Bear case: Management just told you Q2 pipeline conversion was a peak — and Piper expects no revenue growth acceleration until FY28 at the earliest. Flex pricing is a 3-point cRPO headwind and up to $13M of revenue recognition noise this year. Headline growth (21.3%) is still decelerating toward the 19% guide. The stock already trades at a premium to growth peers after the 72% move. At $51.50 vs. a $55 median PT, the r/r for fresh longs is mediocre unless you truly believe the DBNER inflection compounds from here.
"Further multiple expansion likely depends on acceleration in total revenue growth, which [we] expect to occur in fiscal 2028 at the earliest." > — Piper Sandler
Two things. First, the Fed beat: GTLB's public sector business had been the overhang through the budget mess — it improved a quarter earlier than expected. That's a genuine incremental positive, not just noise. Second, the Flex repricing math: management turned what could have been a discounting headache into a usage accelerator, but the 3-point cRPO headwind means reported metrics stay noisy. The DBNER inflection happening DESPITE that headwind is the cleanest signal the consumption story is real.
GTLB is a leveraged bet on AI code output. More AI-generated code → more code to review, test, and secure → more seats and consumption. That thematic lifts the broader DevSecOps complex — TEAM, CRWD, and the GitHub/Azure DevOps stack inside MSFT — plus the private Snyks of the world. The SMB stabilization and early Fed recovery also read through to software demand generally: the 2026 budget digestion appears to be ending sooner than feared.
Net: one strong quarter, honest guidance, and a stock that's already re-rated. This is no longer mispriced — it's fairly valued with a proof point. For PMs: own it on dips, don't chase it at $52 when management just told you the pipeline conversion normalizes from here.
HPE is not a buy-the-dip yet. The quarter was spectacular — revenue, profit, gross margin, FCF all records — and shares still faded after hours on a beat-and-raise. That's the tell. Q3 FY26: ADJUSTED EPS $1.11 VS. $0.92 ESTIMATE, revenue $12.2B vs. $11.93B bogey (+34% YoY), non-GAAP EPS north of $1 for the first time in a single quarter. After a 132% run, the market isn't asking whether AI demand is real. It's asking at what EBIT% the backlog finally converts.
BULL. Deutsche Bank raises to $68 from $62 (Buy) on RECORD AI ORDERS OF $3.1B and a $7.6B BACKLOG. Goldman reiterates Buy at $75. Raymond James pushes to $86. The shared logic: this is a supply-constrained demand story, not a demand problem.
Orders are coming in 3.5x faster than revenue. Guidance reflects supply-side constraints, not demand limitations. Memory availability is the primary supply constraint. > — Deutsche Bank
Guidance supports the read: FY26 revenue growth of 34-37%, non-GAAP EPS $3.75-$3.85, FCF NORTH OF $3.75B, networking +73-74%. Management raised the FY27 revenue framework to 13-17% from 8-12% — a secular re-rating, not a quarter's worth of momentum. Guidance doesn't assume any AMD Helios ramp. Add the $3.5B hyperscaler inferencing win plus the Oracle gigawatt-scale Juniper deployment, and HPE is becoming an AI networking story — richer margins, deeper moat — not just a GPU box shifter.
BEAR. Wells Fargo cuts to $54 from $67 (Equal Weight) and asks the right questions: revenue conversion pace and underlying EBIT%. Intelligent Edge growth normalization is taking longer than expected — customer inventory digestion. Translation: the high-margin networking side pauses to digest just as the lower-margin AI mix ramps. If revenue recognition wobbles, FY27 has a growth scare built into the setup.
PT spread after a beat-and-raise: $54 (WFC) to $86 (RJ). WFC sees fair value near the tape — InvestingPro says $52.08; the stock sits at $51.83. The bulls see a backlog feeding the P&L for quarters. Both are right. The argument is entirely timing and EBIT%.
DOWN 4.7% ON THE WEEK and fading post-earnings — that tells me positioning is heavy and the next catalyst is a quarter out. Not a short — I won't fight a $7.6B order book. But chasing here is poor r/r when the margin mix debate needs another print to resolve. The entry gets more interesting a few dollars lower, after the forced sellers clear.
DA Davidson keeps Buy / $245 PT after Fal.Con 2026 — rating unchanged, but the conference takeaways are the real signal. Customer and partner chatter points to strong AI Detection & Response demand plus continued vendor consolidation onto CrowdStrike. Shares at $207, so the Buy is not exactly a stretch — the catalyst is the investor briefing, where the firm expects management to pull forward long-term ARR targets, potentially with an early FY28 NNARR growth guide. FQ2 already printed $332.8M NNARR (+51% YoY) on 26% revenue growth, so the consolidation story has receipts behind it.
"Conversations with customers and partners at the welcome reception and throughout the day were positive."
Weak-sauce quote, but it's the whole ballgame for a conference-check note: AI security demand is pulling spend forward and the platform is the consolidation winner. Watch the briefing for the FY28 guide. If they frame NNARR growth as durable rather than pulled-forward, the stock's re-rating debate shifts from "can they sustain" to "how much is AI security worth."
MS bumps NTSK to $16 from $14, stays Overweight. Not a big PT move — but the signal is the timing: net new ARR landing AHEAD of third-quarter expectations. Sales cycles haven't accelerated yet, which is the bull case for more to come.
Q2 print beat bogeys clean: -$0.03 adj loss vs -$0.07 expected, $221M revenue vs $214.2M. Guide raised. The 31% LTM growth and 69% gross margin get the airtime, but the company STILL can't print profits — and the stock's +34% six-month run already pays for a fair amount of the optimism.
MS hangs the narrative on AI. Two vectors: development efficiency on the roadmap side, new inroads on the customer side. If that compounds, the TAM expansion is real. But the $15–$28 sell-side range tells you where the debate lives — MS's own $16 target sits at the LOW end of the strip.
"Net new annual recurring revenue growth is arriving earlier than third-quarter expectations."
That's the ballgame. Early ARR either becomes durable acceleration — and the low-end PT cluster breaks higher — or it's pull-forward, and $16 is fair. Light coverage, so no new counter-thesis here. The risk is that everyone already reads the AI story the same way at $5.5B cap.
JPM trims CRDO to $310 from $335, keeps Overweight — call it optical ramp optics, not a demand problem. FQ1 printed clean: $1.20 EPS ON $479M REV (+115% YoY, +10% QoQ), 67% GM. Management raised FY27 to >85% growth. Stock sold off anyway because the market wanted a bigger optical number and got a reiteration of the >$600M target instead.
Cardoso's read: the raised outlook may disappoint buy-side expectations on magnitude, specifically on the optical ramp. Copper remains the clean compounder — optical is the swing factor and the timing is murky. Street PTs are all over the place, too: Stifel at $350, JPM $310, Needham and BofA $275, Rosenblatt $235. Wide dispersion on a 165% LTM grower says the valuation debate revolves around optical execution, not end-demand.
Truist fires the first meaningful buy call on Vishay 3.0, and they're not messing around — $42 PT, 37% upside, BUY. The whole thesis rests on operating leverage: 9% sales CAGR '25-'30, opex growing at half that pace, and gross contribution margin expanding to 40-50% from today's horrid 20% gross profit. That's not a cyclical recovery call, that's a transformation story two years in — and Truist admits the details are messy but the direction is clear.
The math is aggressive but coherent. They model $1.83 EPS in 2027 and NORTH OF $4.30 BY 2030 — that's a 2.4x step-up off an already-improving base, helped by lower taxes and a higher share count. The $42 PT is just 23x the 2027 number, which is a peer multiple, not a growth multiple. Fair, given VSH trades at $30.61 and the Street's target range is $26-$45 — so Truist is at the high end but not alone.
"Truist described the evolution as numerous and complex in its details."
Translation: this is a multi-year hold-your-nose operational story, not a Q2 trade. And the tape confirms it — VSH beat Q2 (EPS $0.19 vs $0.14 est, revenue $918.6M vs $886.5M) and STILL sold off premarket. Market wants proof the margin story is real, not just guided. New flyback transformer series (IFBT) is a nice product drip but won't move the needle on its own.
Worth a starter position if you have 12-18 month horizon. The risk/r:reward only works if you trust management to execute on 3.0 — and the 20% gross margin says there's plenty of wood to chop.
RBC starts RNG at Outperform with an $85 PT — the crux is that the market is pricing this as a no-growth utility while the business keeps proving durability. Mid-single-digit top-line growth post-COVID, $2.8B ARR across ~600k businesses, and now the AI/contact center attach story gives RBC room to argue for a SLIGHT ACCELERATION from here. Stock's already up 141% in a year and trading at $72.58, basically at the 52-wk high ($72.85), so the re-rating is partly done — but the r/r still works if you buy the "low-to-no-growth" mispricing.
Q2 was clean: EPS $1.22 vs $1.16 est, rev $657M vs $650.55M, and they raised FY guidance on AI product demand. Needham also went to $85 after management meetings, so you've got two shops landing on the same number from different angles — one from valuation, one from growth visibility. That's a decent tell.
RBC sees potential for slight acceleration from current levels, while the market prices the stock at low- to no-growth levels.
The bear case isn't crazy — the easy money in the multiple is made, and mid-single-digit growth doesn't scream compounder. But with margin expansion and buybacks layered in, you don't need heroic assumptions for the equity to work. Decent spot for a name most PMs probably haven't looked at in two years.
DA Davidson says the 76% six-month rip is just the beginning. Raised PT to $40 from $33 (Buy) – a blink at 31x FY28 FCF, but they're buying the model, not the multiple. The firm nudged up FY27/FY28 revenue estimates ahead of the quarter after "constructive" exchanges with the company and a positive read-through from a key service partner.
The thesis: enterprises are re-scoping how they do customer engagement, and Braze expands FCF margins as it scales. That's a long-duration growth story with a clean leverage point. The new PT lands ~21% above the current $33.07 – not a stretch on a name that's already re-rated. Feels like "keep holding, more to go" rather than a fresh conviction call.
The firm views Braze as a long-term beneficiary of enterprises shifting customer engagement objectives with the ability to expand free cash flow margins as the business scales.
The crowd is already on board – TD Cowen at $36, BTIG $35, GS init Buy at $34 in recent weeks. Nobody is fading the AI adoption angle. The risk is the print: BTIG flagged the YoY growth comparison humps ahead. DA Davidson is effectively saying the comps don't matter if the FCF durability story holds.
Needham bumps MGNI to $30 from $25, Buy kept — but this isn't a multiple story, it's a share shift story. The Google behavioral remedies ruling handed Magnite a structural overhang on DV+ ad dollars, and the market is finally pricing it.
The math is stunning in its simplicity. Magnite's own disclosure: every 1% of Google DV+ revenue that migrates = $50M of ex-TAC net revenue. That's ~6% upside to Needham's $842M FY27 net revenue estimate, and it doesn't require multiple expansion or macro help — just a court order doing the sales work.
Every 1% share shift from Google's DV+ revenues to Magnite adds $50M of ex-TAC net revenues to the company.
Stock's already up 81% in six months, so the re-rating is underway. But the Q2 print keeps pace: EPS $0.26 vs $0.25 est, revenue $192.8M vs $179.2M est, EBITDA +30% YoY, CTV accelerating 36%. Scotiabank joined the party, $17 to $27 post-print — same thesis, confirmation of operating momentum.
Bull case: Google's hands tied, CTV SSP consolidation continues, and MGNI is the purest play on programmatic CTV's secular growth. Bear case: 81% run already front-runs a chunk of this, and the $50M-per-1% is an option value, not a base case — actual DV+ share leakage takes time, if it comes at all. This is a 2027 story trading like the shift is imminent. At $3.57B market cap against $742M LTM revenue, you're paying for the court ruling to be sticky.
Cantor bumps CXM to $7.50 from $6.50, stays Neutral — and with the stock at $7.34, the market has already banked most of that. Fair enough. The turnaround is real but measured: Q2 beat on subscription revenue, operating margin, and EPS, while total revenue came up slightly light purely because a large implementation project is winding down. That's the right kind of miss — lower-margin professional services rolling off the books.
NRR is the reason this isn't a Buy. Overall NRR slipped to 102% from 104%, and the $1M+ spend cohort dropped to 112% from 115%. Decelerating NRR caps the multiple no matter how clean the balance sheet is (cash > debt, and the stock is +32% over six months). Offsetting that: total RPO accelerated for a second straight quarter to 11%, highlighted by a $20M TCV five-year deal across 35+ sports-betting and gaming brands. That's the land-and-expand proof-of-life the bulls needed.
The catch is current RPO, which decelerated to 3% from 5% — the near-term conversion signal. So the narrative splits cleanly: pipeline quality is improving, but near-term revenue visibility is still soft. At $1.7B market cap with a 66% gross margin, the downside is protected. The upside needs current RPO to inflect. Show-me.
Barclays bumps ADBE to $295 from $250 but STAYS AT EQUALWEIGHT — the rating says "I won't fight you," the PT says "I'm not paying up for the fight." Print lands in 7 DAYS (Sept 10) with a new CEO/CFO in the room, and the setup is more interesting than the rating implies.
Barclays models Q3 net new ARR at $400M — a sequential decline on greater freemium mix — with real upside to $420M+ if web traffic and app downloads stay strong and residual pricing benefit lingers. Q4 steps up to $770M on seasonal enterprise strength, a clean 40/60 H2 split with essentially no pricing benefit left by then.
The part that actually matters is FY27. Barclays sits below the Street's $2.34B consensus. But strip out ~$480M of Semrush contribution, ~$250M of pricing, and the 53rd week, and the Street is implicitly modeling 20%+ ORGANIC GROWTH in underlying net new ARR. The bearish narrative is AI substitution; the model math says the crowd is already paying for acceleration underneath the headline noise.
Analyst complex is bifurcated: RBC at $315 Outperform and CLSA initiating at $300 Outperform on one side — Morgan Stanley at $240 Underweight on the other, citing AI substitution into Creative Cloud recurring revenue. Citi at $301 splits the difference: software multiple expansion doing the heavy lifting while flagging a 26% H2 NNARR decline. That's a wide dispersion into the print, which usually means positioning-driven tape, not fundamentals-driven.
The Saudi deal — 27M citizens getting free access to Adobe's AI tools, valued at $4B+ — is a nice demand signal but not a near-term model input. Use it for sentiment, not for estimates.
Biggest risk to the quarter isn't Q3 — it's the new CEO/CFO combo potentially setting FY27 guidance conservatively. That's the one thing that could make a beat feel like a miss.
KeyBanc is telling you to stop flinching and lean in. They reiterated Overweight with a $280 PT — that's ~59% upside from here, and they made a point of saying the entry is "highly compelling" on a pro forma basis ahead of the Performance Technologies spin-off.
VERDICT: This isn't a fundamentals debate right now. It's a positioning call dressed up as valuation. MOD is DOWN 45% FROM THE 52-WEEK HIGH ($323.25 → $176.45) and down 24% in six months. Sector sentiment is horrid. That's exactly where they want you to act.
The bull case hangs on the spin-off unlocking the data-center cooling story. The bear case is simpler: DC margins haven't proven out, and the Q1 print showed the fragility. Revenue beat the narrative but missed the number — $874.1M (+28% y/y) vs $891.14M consensus, with supply chain issues choking the data center business. Meanwhile EPS landed at $1.53, a clean beat over the $1.38 bogey. So the demand story is intact, the execution story is not yet.
KeyBanc's line:
Shares are too compelling to ignore, particularly on a pro forma basis.
They flagged improved margin execution as the necessary condition — "necessary" being the operative word. Near-term catalysts cited, but they didn't spell them out in the piece. The spin-off itself is the big one. Hunterbrook also dropped a bullish piece suggesting the DC cooling business is bigger than disclosed, though that one's self-dealing by design (they hold a long). Keep it in context.
Net: This is a high-conviction, high-risk buy-the-weakness setup. Aggressive buying is the explicit rec. But if you're a PM who needs to see margin expansion before paying up, MOD will make you wait a quarter or two. The r/r only works if you believe the spin-off clears the overhang and DC margins inflect. KeyBanc does. Strongly.
NVDA — Rubin Ultra's memory de-spec (HBM4E 12-Hi 384GB → HBM4 8-Hi 192GB) is the shortage talking, not demand cracking: memory ran to ~40% of accelerator TCO and the cut brings it to ~28%. THE LARGEST BUYER CUTTING MEMORY CONTENT CONFIRMS SUPPLIERS OWN THE PRICE. BofA sees FY28 +70% as a floor on ~2x demand vs supply; at ~16x CY27 and ~0.3x PEG, the stock already discounts a lot of bad news. Custom XPU displacement is now quantified (TPU 8i, Jalapeno beat merchant silicon on targeted inference), but that share loss runs slower than the demand pool grows. The $105B OpenAI guarantee with a six-partner backstop caps the cash-out option near ~25% — circular-financing fear looks overpriced at this multiple.
TSM — Building at 5x the historical pace — 13 fabs in Taiwan plus 5-6 overseas — and still can't meet demand. EQUIPMENT DEMAND NEARLY DOUBLED SINCE YEAR-END; construction labor now binds. Custom XPU growth doesn't bypass the toll booth: Broadcom's path from $58B FY26 to $230B FY28 AI revenue pulls TSM deeper into every node, interposer and advanced-packaging step.
ASML — Same TSM signal, cleaner angle: suppliers can't keep up with equipment demand, and ASML owns the litho choke point. Larger dies, chiplets and HBM stacks add litho layers per wafer, not fewer. Visibility is now the catalyst — but supplier capacity, not orders, is the limiting factor.
AMAT — TSM's near-doubled equipment demand runs straight through deposition/etch, advanced packaging and HBM integration. This is order visibility, not a spending forecast. If TSM can't build fast enough, AMAT still wins on process intensity per tool.
LRCX — TSM's buildout plus memory's HBM and conventional DRAM expansion feeds Lam's etch/dep book directly. The customer-side constraint is fab completion pace, not end-market demand.
KLAC — Process-control intensity scales with die size, chiplets, HBM stacks and interposer complexity — not just wafer starts. Even with decelerating wafer growth, KLAC rides the complexity curve. TSM's equipment doubling confirms the attachment.
AMD — Same custom-XPU squeeze as NVDA but without a CUDA-grade software lock-in. Hyperscalers now choose between NVIDIA and their own silicon — AMD sits outside that decision. MI455X's N2/N3P node split reserves leading-edge wafers for the transistors that matter, which raises packaging and substrate complexity. The offset is real demand for alternative accelerator capacity, but the structural position is a notch worse than the narrative implies.
MRVL — The Google deal headline (up to $120B through FY33) matters less than the dual-source tell: Broadcom going vague on Google's exact-GW reads as MARVELL GAINING SHARE INSIDE GOOGLE. Amazon and Microsoft also sit outside AVGO's XPU projections. Gross-margin % is the wrong lens for custom AI — dollar profit compounds even as the margin line bends.
MU — DRAM GMs ~90%, Nanya August +560.85% Y/Y, Rubin de-spec, Broadcom locking HBM for FY27-28. Every print says suppliers price, buyers queue. The shortage now spills into conventional DRAM. Lutnick's chip-tariff headline hit Korean memory, but US-built fab capacity earns the exemption — MU's American footprint is the relative winner if tariffs go real.
TDC — The highest-conviction negative software read in the feed. AI-assisted migration compressed a Teradata exit from 2-3 years of friction into under three quarters — an Australian bank moved a 17B-transaction financial-crime platform and queries run 10x faster. The moat was inertia; AI just monetized its removal.
ORCL — Snowflake's Hybrid Tables and Postgres push is a slow compound against Oracle's operational database base, not a current-quarter event. As agentic apps pick native, simpler foundations, the legacy estate erodes one workload at a time.
CRM — AI model costs will dent software gross margins before operating leverage arrives. CRM sits in the platform bucket with owned infrastructure, but the whole application tier is repricing — UI and implementation hours are the replaceable part.
NOW — Workflow scale absorbs AI model costs better than sub-scale SaaS, but agentic AI also compresses the billable implementation labor inside platform deployments. The margin squeeze lands before the opex benefit shows.
HUBS — Third-party model costs take a bigger bite at HubSpot's revenue scale, and AI-driven bundling shrinks the tool count customers need. Scale and owned infrastructure decide who eats the margin hit.
MSFT — New segment reporting finally exposes Azure's scale: FY26 AZURE $101.9B, +40%, and 1Q27 guide accelerates to +45%. Total FY26 revenue $331.8B, +18%. OpenAI revenue recognition ($24.1B FY26 vs ~$102B Azure) now matters optically. Exclusive frontier attach fades as OpenAI spreads capacity across clouds, but Azure's scale internalizes the AI cost side better than anyone.
AMZN — Anthropic standardizing on Google-designed TPUs (16GW through Broadcom) is a structural knock on Trainium lock-in — a frontier lab chose Google silicon over AWS. Model alignments go multi-cloud, so AWS loses exclusive access to premium model demand. Near-term, SNOW's acceleration feeds AWS consumption revenue. The cloud layer wins either way.
GOOGL — Clean legal win: NO AdX SALE, no open-sourcing the final auction logic, no contingent DFP divestiture. The biggest regulatory overhang on the cash cow just lifted. Anthropic's TPU ramp (1GW Ironwood 2026, 5GW TPU 8i 2027, 10GW incremental 2028) validates Google silicon beyond GCP — Broadcom calls Anthropic its largest XPU customer by FY27 — and points straight at Trainium.
META — Back in the frontier fight with strong post-training capability, but the safety burden runs higher than Anthropic's or OpenAI's, and distribution favors GOOG/AAPL/MSFT. Off-balance-sheet commitments jumped $183B → $279B inside a $1.13T hyperscaler total. The tape compresses financing-risk multiples, not operating fundamentals.
AAPL — Memory supercycle punches the hardware BOM: Pro 256GB memory cost reportedly +~400% Y/Y in Q3 2026. ASP defense comes via 2nm foldables ($2,099-2,299, top configs potentially >$3,000), but unit volumes take the hit. The UK devs' £2B collective action adds a legal discount to Services — a multiple overhang, not an earnings event.
VRT — ~$1.5B upfront plus $1.15B in earnouts for a ~90-person company at ~13x expected 2027 EBITDA. That's a capability buy, not a revenue buy: Vertiv extends from cooling/power into grid interconnection and behind-the-meter design. Washington restricting Chinese-linked transformers only extends the runway. Source-to-chip verticalization.
ETN — Chinese-linked transformer restrictions plus multi-year lead times put Eaton directly in the AI grid bottleneck. Domestic electrical equipment gets the pricing and share first.
PWR — Restrictions lengthen lead times; someone still has to physically build the transmission and grid connections. Quanta owns the labor and construction capacity to close the gap — complexity and cost inflation work in its favor.
BE — Forward EV/EBITDA reset from ~85x to ~42x post-short; the debate moved from AIDC demand to delivery execution. Bloom models ~2.2GW of 2027 deliveries vs ~2.0GW consensus against a ~15GW North America power shortfall. Turbine supply won't normalize by 2030, keeping Bloom's time-to-power advantage intact. Risk: service organization scaling.
DDOG — The observability data plane gets stickier in an agentic world — coding agents need production feedback and pull Datadog in over MCP. The UI is contestable; the data plane is the moat.
SNDK — HBF stacks NAND like HBM and repositions storage from cold archive to an AI inference tier. SLC-based HBF consumes ~9x the NAND it displaces — a structural supply-tightening vector the market prices at zero. Watch hyperscaler adoption and Q4 NAND contract prints.
UBER — 3,300 layoffs with AI and robotaxis explicitly named as drivers. That's an automation-led cost reset and a pivot toward autonomous delivery economics — AI as labor replacement inside the platform, not just a customer feature.
TSLA — Cybercab launches tonight with almost nothing disclosed. Another binary on the autonomy narrative. France quietly starting FSD testing builds EU approval optionality. Options premium stacks around the event — an expensive single-day positioning trade.
DTEGY — Elliott built a position and publicly opposes the T-Mobile US merger. The activist no-vote dented the consolidation premium — Deutsche Telekom sits on today's movers as the market re-prices deal scenarios.
TMUS — Same Elliott item from the other side: implied deal synergies get delayed or diluted. This is a deal-probability reset, not an operational break.
KEP — Asked Samsung and SK hynix to PREPAY ~KRW25T OF ELECTRICITY BILLS over five years — KEPCO pays interest and uses the cash to build semiconductor-cluster grids. A national utility too big to fund its own grid capex is the purest power-constraint signal yet. Memory makers treating electricity like EUV allocation says the shortage is structural.
IFNNY — Infineon now formally lists AI data centers and physical AI as growth drivers. Europe's power-semis AI narrative is on the record. Second-derivative AI infrastructure, but the story finally exists.
SOTFY — Raised growth guidance from >30% to ~50% at constant FX on AI demand — the first European semi name with a credible AI revenue line. Photonics-SOI prints confirm the upstream strength already visible in optical and custom silicon chains.
PDD — The EU's €3 parcel fee generated €224M in Belgium alone in seven weeks — and import volumes fell. Washington and New Delhi tighten the same way. Cross-border e-commerce growth hits a slower institutional wall.
EQIX — The Nvidia deal positions colocation as an AI deployment venue against multi-gigawatt hyperscale campuses. Enterprises want lower-latency, flexible capacity. EQIX just found a path to monetize Nvidia systems inside its interconnection ecosystem.
INTC — A claim circulates that Washington wants Intel fabbing 50% of US domestic chip output. If real, that's a national-champion subsidy framework and foundry economics change. If not, Intel stays a high-capex foundry aspirant with execution risk.
QCOM — Unitika passed Qualcomm's T-Glass qualification. High-end glass fiber is a live supply constraint in advanced CCL and packaging. This validates the bottleneck basket more than it moves QCOM-specific numbers.
LITE — Soitec's photonics-SOI strength confirms what Lumentum's tape already shows. AI interconnect and custom-silicon scale-out keep optical demand bid. Confirming read-through, not new information.
COHR — Same optical echo: Coherent's AI-networking and co-packaged-optics content keeps climbing. Soitec's ~50% guide just validates the chain.
RKLB — Broke multi-year support at $62.5 with no fundamental datapoint in the feed. That's risk-off pressure in high-duration space names. Technical breakdowns in crowded momentum sectors usually precede further multiple compression.
CBRS — Back at the $185 IPO price. New AI chip listings meet a tape digesting carry unwind and rate repricing. Listing-supply signal, not a company-fundamentals story.