Good morning.
Tape’s coiled into a pre-jobs knot. S&P/Nasdaq futures flat, gold pinned, crude creeping higher on Hormuz — nobody wants to own a position on either side of the 8:30 print. Europe showed the only conviction: AI up, software down, and Wednesday’s SNOW squeeze isn’t refilling.
Taiwan’s overnight revenue prints carry the real signal. ASPEED: RECORD NT$16.26B AUG REVENUE, UP +118% YoY, Q4 PRICES GUIDED UP DOUBLE DIGITS. UMC: 47-MONTH HIGH, +30.7% YoY — mature nodes with AI content. Pricing power has migrated from HBM to BMC to interconnects, and suppliers aren’t done taking it.
Macro has a new layer. Norway’s $2.4T fund proposing to shed $75–80B of Treasuries turns the debasement podcast into a structural supply event. JOBS CONSENSUS +53k: hot print = de-grossing 2.0, software catches the knife first; cold print = 2Y leads and AI duration bounces hardest.
Three themes frame the day. One: the supply chain owns pricing power — ASPEED’s double-digit hike validates HPE’s supply warnings, don’t fight the pass-through. Two: SAMSUNG CLOSED THE HBM SHARE GAP TO WITHIN 10 POINTS FROM ~50 POINTS A YEAR AGO — SK hynix’s sole-supplier premium is depreciating, and the 2027 LTA is the next repricing event. Three: DEEPSEEK’S 160K-HUAWEI-ACCELERATOR CLUSTER makes China compute independence an engineering fact — NVDA’s China re-opening option loses a notch, and token deflation gains a second cost curve.
We’ll hit up MU, SNOW, and NVDA first, then get to the memory/HBM complex — SSNLF, SKHNY, UMC — and grind through the Taiwan chain prints.
Snowflake just made the AI-in-software debate look silly. Product revenue accelerated for the THIRD straight quarter — 37% — and AI products drove roughly half of that step-up. Then management raised FY27 guidance by $230M, about three times the size of the $75M beat. You don't do that without visibility. Snowflake is no longer an AI story. It's an AI P&L.
The problem is the tape already knows. Stock ran 84% in six months into the print, then popped another ~20% to $373. Now the Street is chasing: at least a dozen PT hikes hit before lunch, cluster at $425-450. This is not a "is the story real" moment. It's a "how much is left" moment.
Product revenue +37% YoY — third straight quarter of acceleration (30% → 34% → 37%) — and a 5% beat vs guidance. Roughly one point of that came from the Observe acquisition, so organic is closer to 36%. Still a beat machine. Total revenue $1.547B vs $1.483B consensus. Non-GAAP EPS $0.62 vs $0.45 expected. NRR 126%. Operating margin 15% vs 11% a year ago.
The Q3 guide is 37-38% — a FOURTH consecutive acceleration. FY27 product revenue guide raised to ~36% from ~30.5%. Piper and Scotiabank both think Q3 prints north of 40%.
The margin story is more nuanced than the headline. Product gross margin slipped ~100bps to 75% — AI workloads carry lower gross margin, call it a mix tax. But management RAISED FY27 EBIT margin guidance 100bps to 14.5%. Headcount discipline is the offset. Growth accelerates, gross margin slips, operating margin expands, EPS beats by 35%. That's the model working.
CoCo is now in 9,100+ accounts — better than 60% of the installed base. That means the AI attach story is still in early innings. Two-thirds of the customer base left to adopt, and every adoption pulls core compute along with it.
We knew AI was helping. We didn't know AI was HALF the story. Cantor, Truist and Rosenblatt all land at roughly 50% of the growth acceleration coming from CoCo, CoWork, AI Functions and AI-assisted migrations.
That last piece is the sleeper. Faster legacy data warehouse migrations are pulling revenue recognition forward — for the SECOND straight quarter. Management frames it as AI-assisted migration driving earlier consumption. That's a calendar pull-forward, not just pipeline. It shows up in the guide. If it's structural — and the guide raise by 3x the beat says management believes it is — then FY27 estimates are still too low.
Management's own framing matters here: core workload strength and AI usage are "compounding" — AI products are a net-new consumption layer on top of existing workloads, not a replacement. That's the crux of the bull case.
Bull case: This is the cleanest large-cap AI monetization story in software. Three consecutive accelerations, guidance implying a fourth, EPS crushing by 35%+, and full-year guidance raised by more than the beat — for the second straight quarter. Stifel moved its base case from 25%+ sustained product growth to 30%+ for coming years. Mizuho calls it a top pick. CoCo at 60% of the base means the attachment story has room, and every AI workload pulls core platform consumption. The operating leverage is the kicker: even with gross margin dilution from AI mix, EBIT margins are going UP. That's a compounding machine, not a one-print wonder.
"Snowflake has proven it is operating in a league of its own." — Scotiabank
"The AI flywheel is accelerating at scale." — RBC Capital
Bear case: You're paying for the AI winner it might become. Guggenheim holds Neutral at 18x EV/NTM recurring revenue — and SNOW isn't cheap on forward estimates either. Bernstein moved to a 15x multiple (from 11x) but still sits at Market Perform, wanting more proof that Snowflake's AI differentiation holds against Databricks and friends. The subtle Guggenheim point cuts deeper: this could be the broader software consumption recovery showing up first in usage-based models — not Snowflake-specific share gains. Fair. So is the gross margin erosion. And if AI-assisted migrations are a catch-up wave — migrate once, then normalize — the 30%+ growth math gets real shaky for FY28.
"We need to conduct additional work to understand the differentiation of Snowflake's AI offerings and the sustainability of growth before adopting a more constructive stance on the stock." — Bernstein
"The results support our view that the software sector's new annual recurring revenue growth will return to normalized levels of 15-20%... Snowflake is seeing this trend first as a consumption model." — John DiFucci, Guggenheim
Range: $370 to $488. Low end is Rosenblatt at $370 (still a Buy), high end Citizens at $488. Morgan Stanley sits at $470. The CLUSTER is what matters: eleven firms at $425-450. Big single-day moves — Scotiabank $320→$440, Piper $320→$450, Canaccord $325→$450, Stifel $350→$450. Bernstein moved $250→$377 and STILL has a Market Perform — that tells you how much of the re-rate is multiple, not estimates.
No sells. Two neutrals. The debate is not whether Snowflake executes. It's what multiple you hang on a company growing product revenue 37% with AI tailwinds and 75% gross margins. The market is paying ~20x forward sales-ish for that. Historically fair for best-in-class software compounding at 30%+. Brutal if growth normalizes to 20% in FY28.
Watch the consumption-model names next. DiFucci's framing — usage-based businesses see the recovery before subscription models — makes DDOG, CFLT and the rest of the usage-based complex the next place to look for acceleration. They're all listening to the same CIO budget conversation.
SNOW is also the best public read on Databricks. These two are fighting for the same lakehouse workloads, and Snowflake winning faster AI-assisted migrations says the data platform wars are driving real takeaway, not just category growth.
The bigger thematic: AI spend in the data layer shows up as incremental consumption on existing platforms — not new greenfield projects. That favors platforms with installed base, workload gravity and usage-based pricing. Snowflake just became the template. Every software platform looking for an AI demand signal is going to be measured against this print.
Clean print. Conservative guide. That's the recipe for the Street to raise PTs into the $200-$225 zone. And they did — 10 of 11 firms sit at Buy/Overweight, with targets from $200-$225. The lone holdout is Morgan Stanley at $165 (Equalweight), who won't grant multiple expansion against a 4% net-new ARR guide. The bull/bear split is whether you read that guide as a setup or a warning. The stock sits around $177.80 — a discount to the group on EV/FCF, and the after-hours singe on the guide's optics was only ~2%.
11 firms, targets from $165 (Morgan Stanley) to $225 (Stephens, Cantor). The consensus cluster sits at $200-$215 — Wells Fargo, JPMorgan, Needham at $215; Guggenheim $214; Canaccord $210; Stifel, BMO, Scotiabank at $200. The median is $215, right at that 28-31x EV/FCF band.
Valuation math, for context: Wells Fargo at 28x EV/NTM FCF calls it a discount to peers on cash flow "temporarily affected by component cost spikes." Stephens at 7.5x EV/FY28 revenue. Canaccord at ~8x EV/CY27 sales. The Street isn't paying up — it's paying fair value for a security franchise with a discernible bottom and a documented AI catalyst list.
JPMorgan framed the beat best:
"Beat with a slightly better outlook for revenue and ARR for FY27. Zscaler delivered revenue and ARR above consensus by a greater magnitude than we've seen year-to-date." — JPMorgan
Scotiabank does the math on what a normal beat means: FY27 new ARR growth of 10-18%, assuming the company beats like it typically does. That's a materially better picture than the 4% headline implies.
Bull: The beat RATE is expanding — larger beats than anything ZS produced earlier in the year — and guidance embeds zero credit for the AI security products that are demonstrably inflecting. Z-Flex carries a "nearly 30% ARR uplift" on adoption and record large-deal activity supports it. Management just raised guidance for a second straight look, which is a credibility signal after the sales leadership mess. Add the September 9 Agentic SOC launch and the first analyst day in FIVE YEARS on October 6, and you've got a catalyst deck that most $200 names can't claim. The FY27 guide being conservative isn't a bug — it's a floor.
Bear: Sub-5% net-new ARR growth is the headline number, whatever the explanation. Sales force transitions historically don't resolve in one quarter — the RIF and new hires suggest leadership itself is still shaping the org. Morgan Stanley flags Palo Alto Networks showing up in competitive processes and longer sales cycles on transformational deals. FCF guidance came in below consensus — capex and component costs are real margin drags, and the 77% gross margin doesn't offset a guide that steps down net-new ARR by two-thirds. When the company has to tell you the numbers are conservative, that usually means the numbers aren't good.
Cantor's take is the bull case in its cleanest form:
"Cantor views the guidance as conservative and enters fiscal 2027 more constructive than before the results... shares fell approximately 2% in after-hours trading following initial fiscal 2027 guidance that implies a step-down in net-new annual recurring revenue." — Cantor Fitzgerald
ZS's print tells you security budget is bifurcating: AI-related initiatives get funded (Security for AI +50% Q/Q bookings) while legacy SASE refresh cycles stay lumpy. That's a tailwind for anyone selling AI security with a platform story — think CrowdStrike, SentinelOne — and a complication for point-product SASE vendors without an AI governance answer.
PANW is the named bear case (Morgan Stanley cites competitive pressure explicitly), so watch for ZS mentions in their next earnings call commentary. The read-through to the whole space: AI security attach is becoming the marginal dollar in cybersecurity budgets, and the vendors winning it are the ones with visibility into user and application traffic. That's ZS's narrative to own right now. Whether the market pays for it at 4% net-new ARR growth optics — that's the debate.
Mixed print. Monster guide. That's AVGO's setup in four words.
The July quarter (FQ3) beat — $29.6B revenue, +33% QoQ, +5% vs FactSet — but the near-term was dirty. Software missed modestly. October AI guide ticked up but came in below the whisper. Implied Q4 EPS slightly shy of the Street. Stock settled near $362, down ~1.5% after-hours.
Don't anchor on the fade. Anchor on this: FY27 AI REVENUE RAISED TO $115B FROM $100B. FY28 GUIDED TO $230B — A DOUBLE, THEN A DOUBLE AGAIN. That's 10GW of XPUs, then 20GW. Management says demand is ahead of the guide. The constraint is power, shell readiness and physical deployment — not end demand.
Price targets span from RBC's $400 (Sector Perform) at the low to Baird's $630 (Outperform) at the high. The action clusters $490-$600: Macquarie $490 (upgraded from Neutral), Barclays $500, Morgan Stanley $505, Truist $520, StoneX $545, KeyBanc $575, BMO $575 (raised from $455), Cantor $600, Rosenblatt $600 (new Buy initiation). Raymond James sits at $475.
Direction of travel says it all. One high-profile upgrade. One $120 PT hike. One fresh bullish initiation. Exactly one PT cut — Truist trimmed to $520 from $550 on the software blemish and FY27 EPS to $21.12 from $22.35, but stayed Buy at 24.5x (a 5.5x discount to data center peers). RBC is the lone adult at the party. Most targets imply 40-70% upside from the mid-$360s. Either this street is anchoring on an unverifiable two-year guide, or AVGO is washed out.
The Google insourcing overhang is old news. Rosenblatt quantifies it — AVGO's incremental share of the TPU program steps down from 80-85% to 60-65% in FY27. Macquarie's entire upgrade thesis is that this risk is PRICED IN after a 24% drawdown from the $495 high. Google's direct investment in MediaTek confirmed the diversification a long time ago.
The new piece is the quantified offset, and it's bigger than Google:
Bull: The two-year guide is beatable, not aspirational. Even Truist — the one shop cutting estimates — admits gigawatt-scale project sizes make the numbers look conservative. You have multiple funded, non-Google customers with named programs and capacity math. BMO is significantly raising FY27/FY28 estimates. And the entry point is the kicker: AVGO +6% YTD VS SOX +59%. The derating already happened. You're buying a two-year AI compounder with a PEG of 0.48 after a 24% correction. That's a r/r worth respecting.
Bear: RBC's $400 target is the honest counterweight. LLM companies hold supply and financing agreements with MULTIPLE chip vendors and CSPs — no silicon allocation is guaranteed until it ships. At 18.5x CY27 earnings (a >30% premium to NVDA on a stock-comp-adjusted basis), the market is paying up for an asset losing its anchor customer. And the fade on a beat-and-raise tells you the buy-side wanted the near-term guide raised, not just the out-years. When a stock can't rally good news, respect the tape.
"Demand remains ahead of guided revenue, with the outlook constrained primarily by power, shell readiness and physical deployment rather than end-demand." — Rosenblatt
That's the whole bull case in one line. The constraint is physics, not competition.
"Solid Results, Mixed Guidance; FY27 AI Outlook Raised $115B; FY28 Expected to Double Again to $230B." — KeyBanc
The setup, in a headline.
"Considering the size of projects involving gigawatt deployments, the guidance appears beatable." — Truist
The skeptic, being honest.
Three things to watch in the complex:
1. The NVDA TAM debate just got a sparring partner. AVGO is guiding to a custom-ASIC TAM that doubles two years in a row. If that's real, merchant GPU share of incremental AI compute shrinks — and every strategist making the "NVDA is the only AI semis trade" argument has to answer the custom silicon question.
2. MRVL is the second-derivative play. Every TPU socket Google shifts from AVGO toward MediaTek is potential share for other custom ASIC shops. Macquarie thinks the insourcing risk is priced into AVGO — but the optionality on the same shift isn't fully in MRVL.
3. Power and shell are the binding constraints, not silicon. Management said it explicitly. 10GW then 20GW of XPUs is a tax on the grid and the data center shell. Read-through to the power/cooling/electrical infrastructure stack is direct and immediate.
Watch the tape into the open. A beat-and-raise that fades tells you there's positioning to clear. But now the market has a credible FY28 anchor — and the next six months of data points (Anthropic scaling, OpenAI ramp, TPU v8i allocations) will decide whether $115B/$230B was a floor or a ceiling. I'll take the long side of that debate.
DOCU GAVE THE STREET A CLEAN BEAT, A GUIDANCE RAISE, AND A PRODUCT MIX SHIFT THAT'S VISIBLE — IAM AT 15.1% OF ARR VS 14.4% EXPECTED. AND THE BEST THE STREET CAN DO IS ONE OUTPERFORM ($86) PLUS SIX SHADES OF NEUTRAL. THE STOCK'S 43% RUN IN SIX MONTHS HAS ALREADY EATEN THE RE-RATING. TODAY'S PT RAISES ARE ANALYSTS CHASING THE TAPE, NOT LEADING IT.
REV $875.7M, +9.4% AS-REPORTED. But the internals matter more: constant-currency growth was 8.1%, up from 7.1% in Q1 — so FX flattered the headline. Strip out digital add-on hardware and the core ran ~9.7%. Acceleration is real, just messier than the top-line suggests.
EPS $1.16 vs $1.08 consensus. Beat quality, though, came in below DOCU's own recent standard — 80bps over the high end of guidance vs a TTM average beat of 130bps (Wells). Fine, not heroic.
IAM is the story. Penetration jumped 250bps Q/Q to 15.1% of ARR. Citizens pegs IAM ARR at $529M vs its $474M estimate. Management now expects IAM to exit FY27 at 18–19% of ARR, up from ~18%. That mix shift, plus a 25bp raise to FY ARR growth guidance (to 8.5–9.0%), is the fundamental story.
Everything else backed it up: DBNRR +1pt to 103%. Customers paying $300K+ ACV now ~1,300, +14% YoY — accelerating 2pts sequentially, second straight double-digit print in the cohort that matters. Op margin 31.6% vs 30% est. FCF $296M, 34% margin. $307M buyback. Digital sales strongest in several quarters.
PT range $60 (Wells) to $86 (Citizens). Mean ~$72 against $65.97 spot — roughly 9% average upside. The ratings are the more honest signal: six neutral-ish voices (MS Equalweight, Baird Neutral, Citi Neutral, Wells Equal Weight, Evercore In Line, Needham Hold) against one lone bull.
The PT moves tell you more than the PTs themselves. Baird went $55 to $72 and stayed Neutral. Citi went $54 to $72 and stayed Neutral. Those are tape-alignment raises — acknowledging the run and the improving ARR trajectory without endorsing the multiple. MS moved only $69 to $75 and used the quarter to make a point: IAM is progress, not proof. Sixteen analysts revised estimates up post-print — the usual reflex after a beat, not a strategic re-rating.
The bull case. The mix shift is real and compounding. IAM at 15.1% of ARR on a guided path to ~18.5% by FY27 exit — if that keeps ramping, blended ARR growth crosses into double digits without heroic assumptions. The $300K+ ACV cohort growing 14% with a 2pt sequential acceleration says upmarket land-and-expand is working. Gross retention is improving, digital sales are the strongest in several quarters, DBNRR ticked up. The AI moat widens too — Iris now trains on 300M+ private consented agreements, up from 200M+ last quarter. Nobody else has that corpus. And the model prints: ~80% gross margin, 34% FCF margin, $307M buyback in the quarter. You get paid to wait for the growth inflection.
The bear case. This was a beat that didn't change the growth regime. FY ARR guidance moved 25bps to 8.5–9.0% — still high-single-digit, not the double-digit breakout the multiple needs. Management still hasn't quantified IAM pricing uplift or disclosed how much IAM is incremental spend vs migration from legacy eSignature. If IAM is mostly re-badged consumption, the TAM story doesn't change. Wells called the quarter "relatively down the middle," and the beat was smaller than DOCU's own trailing average. At ~10x FY28 FCF (Baird) or 12x EV/FCF (Citi), the stock is fairly priced for a high-single-digit grower — no edge. The 43% run already paid for the good news.
"The second quarter showed incremental progress toward the IAM growth case, though not significant progress." — Morgan Stanley
"The product cycle may have some staying power." — Citi
"Relatively down the middle for the period." — Wells Fargo
DOCU is the cleanest test of the "mature cloud + workflow/AI layer = re-rate" trade. The playbook: incumbent category leader uses its installed base and proprietary data (Iris, 300M+ agreements) to sell a higher-value product into the enterprise. Read this across BOX and DBX — both are trying to sell AI/workflow layers into stagnant content/collaboration bases and getting similar skepticism until growth actually inflects.
The DOCU tape says the market will pay up in advance, but only for products with visible mix penetration — not PowerPoint. Every quarter is a prove-it moment now. If IAM gets to ~18.5% of ARR and total growth breaks to double digits, Citizens' $86 looks like the floor. If it stalls at high-single, the multiple gives back the gain. Watch the $300K+ ACV cohort and DBNRR — those lead ARR by a couple quarters.
The deceleration trade just got run over by a keynote. CRWD management pre-announced fiscal 2028 Net New ARR growth of AT LEAST 20% — $1.626B+ — when the Street had 11-12% in the model and the biggest bull at Rosenblatt was only at 14%. That's an ~800bps guide-over gap, and it retires the "post-incident recovery stock" narrative in one slide.
But don't confuse a great conference with a great entry. Shares sit at $214.97 after a +111% six-month run and +83% YTD. Market cap is ~$214B, roughly 40x trailing revenue. The active PT cluster implies ~15-20% upside from here. I'd rather add on the next dip than chase this print.
Consensus is a Buy and nothing from Fal.Con disturbed that. The active post-conference PT cluster runs $235-$265:
The headline is the early FY28 guide: Net New ARR of at least 20%, or $1.626B+, crushing consensus at 11-12% and Rosenblatt's 14%. Management also pulled forward its long-term milestones by a year — $10B by FY30, $20B by FY35. It reframed that $20B as just 3.5% market share of a $565B CY34 TAM, which is a smart way to kill the "how big can this really get" question.
Product news does the heavy lifting:
Bull case: The guide resets the growth debate. If CRWD can print 20% NNARR growth next fiscal year while layering in SafeMind token revenue and a price increase via Falcon Guardian, this is a share-taking platform with genuine pricing power. Scotiabank spent two days at Fal.Con, met 15+ customers, and heard the same three themes — CRWD as the agentic AI security vendor of choice, accelerating stack consolidation, and an elevated threat environment.
"CrowdStrike is becoming a vendor of choice for agentic AI security needs, customers are accelerating consolidation of security stacks onto the CrowdStrike platform, and the risk environment is elevated though without evidence of budget pull-forward or increased endpoint spending." > — Scotiabank
Attendance was so strong analysts couldn't get into sessions. That's a pipeline tell. SafeMind also neutralizes the bear's favorite "OpenAI/Anthropic eats the security stack" narrative — CRWD is now a model owner, not just a model consumer.
Bear case: Price and timing. At $214B market cap and ~40x trailing revenue, the market already paid for the AI security narrative. The "elevated risk environment" with no budget pull-forward cuts both ways — if threat levels normalize, the urgency driving agentic AI deals could cool. SafeMind's token-pack consumption model is unproven at scale, and it's not in FY27 numbers. The 20% guide is a confidence game, not a revenue print. And if this AI security opportunity is as big as CRWD claims, the model labs and cloud giants have balance sheets to compete.
"The conference demonstrated CrowdStrike's approach to addressing threats in 'the new autonomous, agentic always-on world.'" > — Piper Sandler
This is the strongest signal yet that agentic AI security is a real budget line. The platform consolidation message pressures point-tool vendors — anyone selling a single security control plane now has to answer the Fal.Con keynote. PANW will need to match the narrative at its next event. S and Zscaler have to articulate an agentic AI story or get repositioned as niche exposures. Also a quiet read-through positive for NVIDIA and CoreWeave — when a $200B+ software company builds its frontier security models on your stack, that's another proof point for AI infrastructure spend. CRWD just set the agenda for the group. Now we watch whether the tape holds $215.
Verdict: CIEN beat like crazy, guided up hard, extended visibility into FY28 — and PMs sold the stock anyway. Revenue $1.67B (+37% YoY vs $1.64B est), EPS $2.11 vs $1.73, op margin 22.5% vs 19.8% consensus. Then management dropped an FY27 revenue floor of $8.3-8.4B (+30%+, vs $8.03B street) with OM at 25-27%. Stock's response: down double-digits in the past week to ~$315 after a 274% run. The excuse: FY27 gross margin guidance of 45-46% (light for a stock priced for perfection) and a quarter flattered by one-time tariff refunds. This wasn't a narrative break — it's a positioning unwind.
Six houses updated post-print. Targets now stretch from $347 (B.Riley, Neutral) to $615 (Stifel, Buy) — on a stock at $315. Average is ~$500, but that's a barbell, not a consensus. The bullish cluster sits at $520-615 (Stifel, Raymond James $600, Rosenblatt $525, Needham $520). The cautious pair sits at $347-375 (B.Riley, Evercore). All six agree the demand story is real. The entire disagreement is price.
The rate-of-change tell is in the cuts: Rosenblatt RAISED FY27/FY28 EPS and chopped the target from $720 to $525 on "compressed sector multiples." B.Riley raised FY27/FY28 EBITDA to $2.21B/$2.84B and cut the applied EV/EBITDA multiple from 24x to 18x. Estimates up, targets down. That's the purest expression of a de-rating you'll see this cycle.
Bull case: supply is the constraint, and the constraint is a floor. Management framed the FY27 guide as component-supply-limited, not demand-limited:
"The guidance represents a floor set by component supply versus demand." — Ciena management, via Stifel
Backlog tells you the demand is banked: $8.5B exiting Q3, +$800M q/q (accelerating from +$600M prior quarter), >$10B by fiscal year-end — more than 1x FY27 revenue before the year starts. DCI +82%, pluggables more than doubled, CIEN secured its second multi-rail win, hyperscalers now ~50% of revenue. On margins, Rosenblatt thinks the 45-46% guide is a floor — HyperRail adds 100-200bps through FY27 and GM approaches 50% by FY28. Implied FY27 EPS ~$12 vs $9.68-9.75 street. At $315, you're paying ~26x FY27 and ~21x FY28 (Rosenblatt's $15.01). Not a crazy price if you trust the two-year ramp.
"The focus should shift from demand durability to execution and supply constraints." — Raymond James
Bear case: you're paying 105x trailing for a margin guide that is BELOW the quarter you just saw. Reported GM of 46.4% includes a 70bps tariff-refund kicker; clean GM is 45.7% and guides to 45-46% for all of FY27. That's a plateau, not a ramp. Evercore cut $550→$375 on exactly that. Same asterisk hits the op margin beat — 22.5% gets a tariff-refund assist, so the "270bps beat" overstates the run-rate. And supply cuts both ways: a supply-capped guide means FY27 revenue can't flex higher, so you're paying full price for an execution story with no near-term catalyst (Raymond James, a bull, said that). B.Riley's $347 target is the honest framing: great fundamentals, wrong price.
If CIEN is component-supply-constrained while DCI prints +82%, the pricing power sits one layer up the stack. Bullish read-through for the optical component chain — COHR, LITE, and the broader AI interconnect complex own the capacity Ciena is bidding for. But the systemic signal is the de-rating: analysts raising estimates and cutting targets is the signature move of a maturing trade. CIEN remains the cleanest pure-play on AI optical demand. Just know you're buying delivery, not discovery — and the market is no longer paying discovery multiples.
Best operational print in two years, and the stock is DOWN ~6% on the week. The Street is reading the wrong line.
Samsara printed NET NEW ARR OF $134M — the highest Q2 in company history, +28% CC y/y — and did it while lapping a comp that included its LARGEST-EVER DEAL plus ~$5M of tariff-related slippage. Revenue: $508.4M, +30%, ~$24M above the high end of guidance. Adjusted EPS: $0.20 vs $0.16 ests. Large customers broke the model: +242 net adds at $100k+ ARR, +20 at $1M+, both records. The $100k+ cohort now sits at $1.3B of ARR, UP 38% Y/Y. Gross margin: 76.3%. Then management CUT FY27 FCF MARGIN GUIDANCE BY 200BPS and handed the bears their only real ammunition.
That guide cut is the story of the tape. It's also mostly good news wearing cost's clothes.
Five firms, five PT raises. Range: $48-$58, consensus ~$53 vs the $38.75 close — about 38% upside. BofA to $55 (Buy), Wells Fargo to $58 (OW), Evercore to $55 (OW), Guggenheim to $51 (Buy), Truist to $48 (Hold). The lone Hold is the entire debate in miniature.
"Waiting for a more attractive entry point following the stock's recent outperformance." — Truist
That's not a broken-growth call. Truist raised $38→$48 (+26%) and still can't get there on valuation — 327x trailing P/E will do that. But let's be precise about what they're NOT saying: the story broke. It didn't.
The installed base momentum was known. The beat was largely bid in — this thing ran 35% in six months into the print. What's actually incremental:
"The company lowered its fiscal 2027 free cash flow margin guidance by 200 basis points, primarily due to inventory investments to support accelerating demand." — BofA
Bull: Core telematics penetration is still low, and the expansion story keeps widening — maintenance, asset tracking, workflow automation, cameras. The $134M NNARR print landed on the toughest comp in company history; that's genuine velocity, not an easy compare. Physical AI is becoming a real institutional theme heading into CY27, and Samsara owns the data layer of the physical world — every vehicle, camera, and facility running AI at the edge. If the market attaches a software multiple to 30% growth with 76% gross margins, Wells Fargo's $58 is a bogey, not a ceiling.
"Samsara has multiple levers to sustain durable growth as core market penetration remains low and emerging products create a growing expansion opportunity within the installed base." — Evercore ISI
Bear: You're paying 327x trailing earnings and ~13x EV/CY27 sales for a company that just guided FCF margins DOWN 200bps. Hardware and component inflation caps gross margin, and inventory builds tie up cash exactly when fleet customers tighten budgets in a macro slowdown. The 35% run into the print already priced the acceleration — the 6% fade this week is the market recalibrating. Truist's $48 says the risk/reward is no longer asymmetric after the move.
The market is selling the wrong line. The FCF cut isn't a margin story — it's a shipment-demand signal. You don't pre-buy IoT devices and eat supply chain costs because business is slowing. And the demand data underneath is the cleanest it's been in two years: new ARR growth re-accelerating off a brutal comp, record large-customer adds, a top-5 U.S. city win, record international.
Don't chase the open. The stock is $38.75, still up 35% over six months, and the tape has fresh supply to work through after this week's fade. But put this on the shortlist for any risk-off dip. The consensus PT math says ~38% upside even AFTER the selloff. If net new ARR holds anywhere near this pace next quarter, the de-rating narrative flips fast.
Watch the components. If supply chain inflation persists into Q4, the FCF guide cut deepens and the hardware-tied bear case gains real data points. If it normalizes — as management asserts — this week is the buy point.
Two signals worth carrying to other names.
1. Physical AI is becoming a software story. Evercore explicitly framed IOT as a Physical AI play heading into CY27. The market is starting to look past the GPU layer for beneficiaries with recurring revenue and proprietary physical-world data. Names with sensor/camera/workflow data moats should ride that narrative repricing.
2. Hardware-coupled demand is a capex tell. Accelerated device pre-buys and 28% camera growth signal real appetite from the physical economy — fleets, field service, public sector. That cuts against any broad enterprise IT budget freeze narrative. Samsara just printed its highest Q2 net-new-ARR ever against a monster comp. The physical economy is spending.
VERDICT: PATH PRINTED A CLEAN QUARTER — AND STILL TRADES ABOVE EVERY SINGLE FRESH PRICE TARGET ON THE STREET. That's the whole setup. Shares at $18.22, up ~65% in six months, near the $19.84 high. Four analysts just moved targets to a $14–$17 range, and not one will underwrite the stock where it trades. Canaccord downgraded to Hold on the back of a beat. The AI-trade reversion did the heavy lifting ahead of the fundamentals; now the street wants proof before it chases.
Q2 FY27 revenue $410M, +13% reported (~+11% ex-FX — ignore any claim of 16% ex-FX; an $8M FX tailwind doesn't square with that math). Beat consensus by ~$12M and came in $10M ABOVE THE HIGH END of guidance. Non-GAAP op income $89M, $14M ahead of guide. Op margin 21.7%, ~290bps better than expected. 83% gross margin — pricing power was never the issue. Fourth straight GAAP-profitable quarter, EPS $0.15 in line.
The metric that matters: NET NEW ARR $37M vs $31M a year ago, a $4M beat. Sounds fine until you realize the comp was PATH's second-lowest NNARR quarter ever as a public company. Backing out FX, it's $36M vs $30M. NRR stuck at 109%. AI showed up in 18 of the top 20 deals — that's the bull evidence, and it's real.
Guide: FY27 revenue raised $13M to $1.789–1.794B. Non-GAAP op income raised to ~$445M from $430M. FCF dipped to $31M on tax timing, full-year $425M unchanged. The tell is in the back half: implied 2H net new ARR of ~$130M, ROUGHLY FLAT YOY. That is not reacceleration. That's stabilization with a tight opex leash.
Also in the release: deputy CFO Hitesh Ramani gets the CFO seat, Ashim Gupta stays focused on COO. Buybacks slowed to 2.4M shares from 20.4M in Q1. Noted, not flagged.
Targets: Mizuho $14 (Neutral), Wells Fargo $15 (Equal Weight), TD Cowen $16 (Hold), Canaccord $17 (Hold — downgraded from Buy). Range $14–$17, midpoint ~$15.50. Consensus is a collective shrug with a valuation warning. No one wants to stand in front of a 65% momentum move, but no one will extend PTs above the tape either. Translation: the street views this as a show-me story into the 9/22 Investor Day.
"Shares have rallied significantly over the past three months due to AI trade reversion and software sector recovery ahead of tangible growth improvement." > — TD Cowen
That's the whole setup in one sentence. Multiple expansion first, fundamental proof later.
The firm maintained it believes the company is navigating an AI supercycle. > — Canaccord, downgrading to Hold anyway on valuation and "an increasingly fluid competitive environment"
When a bull cuts to Hold after a beat, the reason is always the same: the easy money is gone.
Bull case: Execution is stabilizing, and the business model works at scale — four straight GAAP-profitable quarters, 83% gross margin, 21.7% op margin with room to flex. AI is pulling deals, not killing them: 18 of top 20 had AI components. The agentic disruption narrative looks great on X but hasn't shown up in the numbers yet. Flat 2H NNARR guidance is conservative, and the 9/22 IA Day could validate AI monetization and a path to genuine reacceleration. At $18.22, you're paying for that optionality — not a crazy price if the event delivers.
Bear case: The quarter beat, but the one metric that matters — NNARR — is $37M against PATH's second-worst quarter ever as a public company. NRR flat at 109% means the installed base isn't expanding. Management is guiding 2H NNARR to roughly flat YoY and calling it good. Meanwhile the stock has run 65% in six months and sits ABOVE EVERY PT ON THE STREET. Canaccord is right — the competitive environment is increasingly fluid, agentic AI is coming for legacy workflow automation, and you're now being asked to pay up before the reacceleration is proven. This is a "sold on the news" risk, not a buy-the-dip setup.
PATH is the pure-play barometer for the "RPA is dead, long live agentic AI" debate, and the quarter argues the obituary was premature. AI is showing up in the deal flow. But the flat 2H guide argues the reacceleration isn't here. This matches what we're seeing across application software: AI trade reversion has re-rated the whole group ahead of tangible growth improvement. PMs should treat this rally as multiple expansion, not fundamental inflection.
The swing event is 9/22 Investor Day. If PATH validates an AI monetization path, the stock has momentum, narrative tailwind, and a low bar to clear — it runs. If the event underwhelms, you're holding a name priced past every target on the street with a bearish overweightedness in the analyst community. The r/r into that event is binary, and the street has already picked its side.
NTSK gave the Street the one thing it wanted: proof the AI security pipeline actually converts. Three brokers — BMO, Piper, RBC — all raised targets Tuesday on the FQ2 print. Piper and RBC land at $20, BMO at $18. The collective message: NNARR inflected, guidance went up, and the agentic-security pitch now has bookings behind it.
FQ2 WAS CLEAN: REVENUE $221M VS $214.2M CONSENSUS. ADJ LOSS -$0.03, HALF THE -$0.07 BOGY. ARR +27% Y/Y, beating expectations. BMO says the performance improved "across the board" versus the prior two quarters — that rate-of-change matters more than the level.
The beat flowed straight into the model. FY27 revenue growth guide raised to ~25.5% from 24%. Management doesn't do that after one good sprint. They see what's in the pipe.
Three hikes, one thesis. BMO to $18 from $13 (call that catch-up — biggest absolute jump), Piper and RBC to $20 from $18. All three maintain Outperform/Overweight. MS sits at $16 as the conservative of the bunch. Nothing negative anywhere. Consensus hovers near 1.5, high target at $28.
RBC frames the setup best: security AI products plus a still-ramping sales force = favorable into 2H FY27. The kicker — one-third of the AI security pipeline is already in proof-of-concept. Piper says early customer evals "progressing toward production deployments." That's revenue with a timestamp, not a narrative.
Bull: NNARR beat bigger than last quarter, POC-to-production conversion underway, sales reps approaching full productivity, and shares at UNDER 6X CY27 REVENUE. The market prices NTSK as an also-ran SASE vendor, not an agentic-security leader.
Bear: Growth still decelerates — 31% LTM revenue to ~25.5% guided, no profitability modeled this year. Zscaler and Palo Alto own the enterprise security relationship. NTSK fights for the leftovers. 6x sales only looks cheap if the growth reacceleration holds.
Piper's line beats anything I can add:
"We believe the current valuation does not fully reflect NTSK's differentiated positioning in agentic security."
That's the whole debate in one sentence. If agentic security is real — and the POC numbers say it is — $14.29 is cheap. If it's marketing fluff, you own a decelerating SASE vendor at 6x sales. Today's evidence skews bullish. The guide raise is the hard proof.
Restatement is a non-event — and that's instructive. MSFT did not reclassify to hide a decelerating number. The new two-segment structure breaks out pure Azure and hands PMs a cleaner read on Copilot. KeyBanc: no major revelations. Fine. We'll take the transparency.
STOCK TRADES AT $510 — $3.79T MARKET CAP. Street PTs cluster $530–$600: KeyBanc and BofA at $600 (OW/Buy), JMP at $550, Stifel at $530 (Hold, but raised from $450 — when a skeptic hikes ~18%, the momentum signal is real). Range spans $400–$870. Consensus: ~12% upside.
The segment news is accounting. Stifel's readout after management meetings is the actual signal: M365 Copilot has cleared proof-of-concept and is moving into full-scale deployments. Monetization is seat-plus-consumption, and management explicitly said outcome-based pricing is not coming near-term. Deliberate choice — protect the seat base while AI consumption layers on top. Plus management gets "opportunistic on pricing" where they have unique product (read: they know where pricing power sits).
"Accelerating momentum for M365 Copilot... as customer proof-of-concept work transitions into full-scale deployments." — Stifel's Brad Reback
Azure backs up the narrative. GROWTH ACCELERATED TO 43% IN FQ4 (from 39%), WITH FQ1 GUIDANCE AT 45%. The restated historicals do not distort that trajectory — KeyBanc notes Azure's growth is nearly unchanged from the old "Azure and Other Cloud Services" line. The messy aggregation is gone. Good riddance.
Bull: Clean disclosure, visible Azure acceleration, Copilot converting to deployment. Data center efficiency gains unlock incremental capacity without a linear capex drag. $600 targets imply ~18% upside on a compounder that finally shows its AI P&L clearly.
Bear: Stifel stays at Hold despite the hike. Management's own language — "acceptable returns" on ramping capex — carries some hedging. If seat-plus-consumption monetizes slower than the capex cycle turns, this is a $3.79T software story paying AI prices for a 43% growth line that inevitably decelerates. The $400 handle on the Street range is that tail risk.
Bottom line: Not chasing at $510. But the restatement plus Copilot conversion data points upgrade MSFT on our radar. Give us a drawdown toward $480 and the r/r flips.
THE PRINT WAS FINE. THE MULTIPLE IS THE PROBLEM. GWRE beat FQ4 numbers across the board, stock still sold off after hours. That tells you everything: nobody questions the execution, they question the 109x P/E and whether FY27 guidance confirms peak growth. Shares sit at $202.86. WFC keeps Underweight with a $195 PT — BELOW the tape. Citizens doubles down at Market Outperform, $220. That's the whole debate in two lines.
Guidewire reported ARR OF $1.24B, +19% y/y cc with $95M NET NEW ARR — $5M above the high end. Total revenue $411M, a ~1% beat against an 11-14% growth guide. Subscription and support revenue +32% y/y, ahead of the 28-31% range. EPS $0.99 vs $0.93 consensus. Looked clean.
WFC's pushback is fair: the $5M ARR beat sits right in line with the trailing eight-quarter average of ~$6M above high end. In other words, no acceleration. License revenue still shrinking -18%. Stock already up 32% over six months. You're paying a premium multiple for a beat pattern that's basically mechanical by now.
Citizens frames the real question — did management sandbag the initial FY27 guide? They think yes:
"The initial fiscal 2027 guidance may prove more conservative than usual based on management's commentary that it has more upside potential than downside anxiety."
That quote matters because GWRE management is not in the habit of lowballing. They've beaten and raised consistently. Fully ramped ARR still grew 22% y/y cc. Pipeline commentary stays positive.
The kicker is new product traction. PricingCenter closed 8 of its 12 FY26 deals in Q4 ALONE, including Nationwide as an anchor logo. ProNavigator closed 14 of its 28 FY26 deals in the same quarter. Those are ramping faster than expected, and they're not in the old core P&C cycle. That's the bull case for durability.
Bull: The guide is a setup, not a signal. Management explicitly says upside risk outweighs downside. New products inflect, fully ramped ARR holds 20%+, and the 1% churn normalization headwind is already disclosed. Paying up for a compounder with 25% revenue growth and land-grab optionality.
Bear: You're paying 109x for a company whose beat streak is running on autopilot and whose own guide is designed to reset expectations. Churn normalization is a real 1% ARR drag. Large lumpy deals take time to close and often land with lower first-year ARR — so FY27 has embedded lumpiness risk. WFC won't touch it at this valuation, full stop.
The after-hours fade says the market wants to see the FY27 number before underwriting the next leg. Guidewire's history says that number will end up being conservative. Today's r/r is just messy in between.
AMBA beat FQ2 by a rounding error, guided Q3 above consensus, AND THE STOCK STILL FELL 10% IN A WEEK. That's a tape tell. When the market marks down an in-line beat with an up-guide, it's pricing the caveats — margin pressure, a prolonged ramp for the new channel deals, and memory scarcity that threatens FQ4. Stifel trimmed its PT to $90 from $106 on sector multiple compression, not thesis breakage, while keeping Buy. (Oppenheimer, for its part, just parks at Perform.) At $63.38 with a consensus PT range of $80-$120, this is a show-me stock that needs to clear the Q4 guide before I trust the low end of that range, let alone the high.
Solid, boring beat: REVENUE $108.1M (+7.7% QoQ), Auto at a record. EPS $0.18 vs the $0.17 bogey. FQ3 GUIDE MIDPOINT $119.5M (+10.5% QoQ), roughly 0.6% above Stifel's $118.8M estimate, driven by physical AI demand inside IoT. That's the bull's headline number.
Then the kicker: management sees little memory-related revenue impact in Q2/Q3 — i.e., it hasn't shown up yet — but flags "continued uncertainty into FQ4." Combined with margin pressure and a "prolonged timeline" for the Capgemini/Macnica partnerships to generate real revenue, that's why the tape can't hold a bid. Those partners are long-term positives — each expected to contribute $0.5B in sales "over time" — but over time is doing heavy lifting when this stock trades at $63 and the narrative needs a near-term catalyst.
Context: LTM revenue is +28%, but AMBA remains unprofitable trailing. Eight analysts revised numbers up for the next period and the consensus call is profitability this year. That's the pivot.
Bull: Physical AI/IoT is accelerating through the model. Auto set a record. SAM raised to $8.5B in FY27 and $22.9B by FY32 (IoT ~70% of the terminal year), and the Capgemini/Macnica channel deals scale the go-to-market beyond direct sales. Stifel's $90 target — 7.6x CY27E EV/Sales — is 40% above the tape even after the cut. If FQ4 caveats don't materialize, this is a gift.
Bear: A beat-and-raise that falls 10% tells you the Street hears the FQ4 warning. Memory supply is external and unmanageable — and if the caution persists into next quarter's guide, $63 is not the floor. The partnership timeline is long, and you don't pay 7.6x forward sales for "over time." You pay that for booked orders.
Stifel on the guide: "The FQ3 midpoint came in ~0.6% above our estimate, driven by physical AI demand within IoT."
That one line is the bull case. Everything now rests on whether management can clear the memory-scarcity overhang when it prints FQ4 guidance. If it does, the consensus range ($80-$120) becomes the playing field again. If it doesn't, wait for a lower entry. I'm not catching this knife, but I'm watching the next print like a hawk.
September 9 is a pricing event, not a product event. Shares sit at $328 — 37.6x, just off the $344 high — and the market has already banked an iPhone 18 supercycle plus a $2,000+ foldable. KeyBanc is the clearing bear: reiterates Underweight $250, says the launch is a NEGATIVE CATALYST. BofA stays Buy $380, looks through the App Store decel and says monetization is improving. The entire debate collapses into one question: how much pricing power does the consumer actually have?
Consensus sees iPhone 18 Pro/Pro Max ~$200 higher. The foldable "iPhone Ultra" starts ABOVE $2,000. That's a new high-water mark for the premium tier. It's also the first launch under new CEO John Ternus — no dry run.
KeyBanc frames the trade-off cleanly:
"A larger price increase could soften gross margin pressure but would likely impact unit volumes and bring sticker shock to consumers."
Translation: big hike = margins hold, units crack. Narrow hikes = margins squeeze and Apple needs another price round sooner. Either way, KeyBanc says history is against you — September events tend to print modest negative reactions. Underweight at $250 is a real drawdown call. DA Davidson (Neutral $270) is also watching the foldable price as the swing factor.
The App Store data is genuinely two-sided. AUGUST GLOBAL REVENUE -0.5% YoY — FIRST NEGATIVE PRINT SINCE FEB 2023. Downloads -3.8%, second straight negative month. Q4 FY26 tracking $6.3B (+0.1% YoY), decelerating from +0.6% at 35 days. Evercore flags six months of decel and gaming -10% YoY. That's not a rounding error; that's the engine losing RPMs.
BofA's rebuttal: dollars per download +4.2% to $1.01, and China revenue +7.9% YoY. Fewer downloads, more revenue per download — price over volume, the exact same bet as the iPhone side. Fine. But if the theory breaks on September 9, it breaks on both sides of the P&L.
Bull case: Pricing power is intact — hardware AND services. The foldable re-rates the premium tier, China stays healthy, and 37x CY27 EPS ($10.32) compresses fast if the launch video hits. (BofA $380, Evercore $365 OW.)
Bear case: The consumer is hitting a wall, App Store growth already rolled over, and a new CEO's first move is asking $2,000+ for a phone. Risk/reward is bad this close to the event. (KeyBanc $250 UW, Rosenblatt Neutral $303 — though he did raise on the Ternus transition.)
Apple also filed suit against OpenAI over a stolen circuit schematic. Noise for now, but a sign the IP litigation posture isn't changing under new management.
Verdict: respect the event risk. The analyst dispersion — $250 to $380 — is the widest in years. That's not a stock with consensus; that's a coin flip on pricing psychology. If you must express a view, wait for the September 10 prints before chasing direction.
The PT raise is theater — MS goes to $210 from $207. Equalweight stays. The real signal is in the math: EPS estimate jumps to $10.98 (CY28) from $8.28 (FY28) on improved GPUaaS gross margins, yet the target multiple gets crushed to ~19x from 25x. Net effect: a $3 PT bump. That's a two-handed acknowledgment that margins are better AND they still don't trust the story enough to pay up.
"This represents a discount to Oracle's historical average of approximately 1.6 times due to execution risk and lack of free cash flow support through the forecast period given GPUaaS investment."
So you've got MS at $210 vs. a $154 tape — roughly 36% implied upside, but no conviction to go Overweight into September 9. Options are pricing ~10% one-way on the print. The bulls who do own it are holding firm: Jefferies at $290 (trimmed from $320 on AI data center financing/execution risk), Mizuho at $320 after the VA contract ceiling ballooned ~$17B with optional periods into May 2031, Bernstein still yelling third-largest-hyperscaler from the rooftops. Key bogies on the call: 115% OCI growth, 41% operating margin. The bull case is an investment cycle that inflects FCF positive later; the bear case is the market has to fund that cycle with no FCF support in the near term. Both are right. Stock's down 20% YTD for a reason — you're buying a promise, not a cash machine. At 0.9x PEG vs. 1.6x historical, MS is saying the discount stays until execution proves otherwise.
Post-print Street is climbing over itself — Truist to $70, GS to $75, RJ to $86, DB to $68 — and the aggregate thesis is simple: order growth is the tell. HPE printed 42% order growth in fiscal Q3 vs mid-teens pro forma revenue, which means backlog now provides roughly five quarters of visibility. That's why Truist lifted estimates 10-15% and sees agentic/inference AI pulling through both AI and traditional servers, with Networking synergies layering on margin upside. Not bad for a stock already up 142% in six months.
The quarter itself was a beat-and-raise: EPS $1.11 vs $0.92 consensus, revenue $12.2B vs $11.93B, +34% y/y, record non-GAAP EPS. Management raised FY26 to 34-37% revenue growth and pledged at least 75% of FCF back to shareholders starting next quarter. Shares still faded after hours on margin-mix concerns — that overhang is the Wells Fargo thesis ($54, Equal Weight, worried about revenue conversion and EBIT %). Fair pushback. But with the order converter working this well and five quarters of backlog stacked, the debate is whether AI-server mix eats the margin story — not whether demand is real. It's real.
Mizuho cuts INTC to $92 from $109 (Neutral) — and here's the kicker: stock trades $91.67, RIGHT ON TOP OF THE NEW TARGET. This isn't a price call anymore. It's a "valuation finally caught the story" call, and multiple compression is the bogey hitting agentic-AI names first.
The thesis underneath hasn't cracked. Mizuho still sees agentic-AI demand accelerating into 2027, driving token growth and a better CPU-to-GPU mix, plus enterprise server refreshes and a PC market running hotter than expected as corporate cycles kick in. The real torque: CPU supply constraints means INTC UNDERSHIPS DEMAND THROUGH 2027, with key customers facing shortages into next year. That's pricing power wearing a constraint costume.
The new businesses are getting numeric shape. Mizuho models Advanced Packaging at $3.5B by 2029 on EMIB-T for TPU and external foundry at another $3.5B on 14A. Small next to core, but it's the optionality bulls are paying for.
Fundamentals back the momentum: Q2 printed $16.1B rev / $0.42 EPS vs $14.4B / $0.21 consensus, client segment $8.9B on ASP strength. Bernstein at $110, Baird at $125. CEO Tan put $12M of his own into the $20B raise that saw ~1/3 of investors shut out — insider signal, demand signal, same thing. Bull case: tight supply, real AI pull, execution finally matching narrative. Bear case: at $92 you're paying full freight for the turn, and when agentic-AI multiples compress, no one asks which name has the best foundry roadmap.
Initiations don't get much more constructive than this. StoneX starts MPWR at BUY with an $1,800 PT (~48% upside from $1,219), slotting right into a target range that now stretches $1,575–$2,100 following the Q2 print. The setup: lean expense profile produces best-in-class semi margins (55% GM, 18.5% ROA), and the monolithic power design moat sits dead center of the hyperscale buildout. Q2 was a beat-and-raise — EPS $6.50, rev $980.6M, +22% QoQ/+48% YoY — with ENTERPRISE DATA +160% Y/Y, the single number that matters.
Post-print PT bumps from at least three firms (Truist, KeyBanc, Needham) all chased the same narrative: data center share gains are accelerating, and management raised the FY outlook accordingly. No bear case worth respecting near-term unless you think the AI capex cycle rolls over — nothing in today's data suggests that.
"The analyst cited the company's lean expense profile, which results in best-in-class margins among semiconductor companies."
Momentum is clearly on MPWR's side. The question isn't whether the story works — it's whether $1,219 with 160% growth in the key segment is still a reasonable entry. At ~48% implied upside to StoneX's number, they think yes.
The HF deal is a platform grab, not a needle-mover. Needham reiterated Buy/$300 after Nvidia announced its $12.93B acquisition of Hugging Face — $11.9B headline plus up to $1B equity retention for staff, closing 1H 2027. Couch-cushion math for a $5.52T cap at ~29x trailing. The point is positioning, not the P&L.
Hugging Face is the GitHub of AI — the default hub for open-weight models, datasets, and the developers who ship them. The tell: Nvidia keeps it an open platform. They're not buying this to shackle models to CUDA. They're buying the distribution layer and the workflow that sits on top of it.
Nvidia said it will keep Hugging Face as an open platform following the acquisition.
That sentence is the whole deal. Close the thing up and you torch the community overnight. Leave it open and you're embedded in the build pipeline of every AI lab on earth.
Bear case: open platform means zero lock-in — anyone can still deploy on AMD, Google, or custom TPUs. Bull case: Nvidia doesn't need lock-in when it owns the path of least resistance.
Rest of the tape is incremental: $35B Anthropic cloud deal via Lambda (Hut 8 building the Texas site), Rubin CPX reportedly pulled in to 1Q 2027 production, JPM Overweight, Baird Outperform. No one's hitting the sell button. Thesis unchanged — HVD (hit the bid? no. hold. whatever) — just consolidation while the market chews on hyperscaler capex digestion.
BTIG stamps the Street-high: $230 PT from $180, Buy maintained. Thesis is pricing power + AI monetization, and they did the homework to back it up. Ran seat-level scenarios (100/1k/10k) on Atlassian's new usage-based pricing calculator. Conclusion: Rovo credits alone can multiply a customer's spend by several turns. Cloud price increase lands October 13, meters go live December 3 — this is a deliberate monetization engine, not a price hike.
SHARES SIT AT $194.68, +153% OVER SIX MONTHS, ONE TICK OFF THE $198.60 HIGH. So the market already caught the first leg. BTIG's edge: the displacement risk everyone fears is actually the moat. Jira stays because auditors and regulators already bless it — even in an AI-native world. The strongest validation comes from outside the bull case:
Anthropic's blog post on AI-native software development highlighted that systems like Jira are difficult to displace because auditors and regulators already accept them.
That's the lock-in story for the AI era. And the fundamental backdrop holds: Q4 printed EPS $1.87 VS $1.50 EST, RPO +44%, first GAAP operating profit in over two years. The $250M buyback doesn't hurt.
PT dispersion tells you who's paying attention: Jefferies at $200, BofA $175, FBN $170. BTIG at $230 isn't an increment — it's a different view on the pricing lever entirely.
Truist goes bottom-fishing with a Buy and a $42 PT — that's 37% upside from the $30.61 tape and near the top of a wide $26-45 street range. The bull case isn't end-market recovery; it's Vishay 3.0 margin math. Truist models a 9% sales CAGR from 2025 to 2030, gross contribution margin in the 40-50% range versus the current 20% gross margin, and opex growing at HALF the rate of sales. That operating leverage takes EPS from $1.83 in CY27E to north of $4.30 by CY30 — the PT is just 23x CY27 EPS, which Truist calls peer-equivalent. Two years into Vishay 3.0, this isn't a single-spark story. Truist describes the overhaul as numerous and complex in its details. The new IFBT flyback transformer series adds incremental proof the product engine is firing, but the prize is margin convergence, not the next transformer.
Q2 2026 printed a clean beat — $0.19 vs $0.14 expected, revenue $918.6M vs $886.5M — and the stock STILL faded premarket. Investor concerns center on future growth and macro conditions, not the quarter. At 23x a future-earnings number, one Vishay 3.0 stumble compresses the multiple fast. The wide target range tells you the Street hasn't signed off yet. Neither have we.
Stifel's back from swampUP and it's a Buy reiteration with $102 PT — stock at $90.61, +87% over the past year. The read-through: DevOps/software supply chain security is real, but the AI-artifact angle is still "early innings."
No FY27 target update at the event — consistent with expectations, so no headline catalyst. Customer conversations skewed toward self-managed and hybrid, with limited near-term cloud migration appetite. That aligns with management's commentary and keeps a lid on the cloud re-rate narrative, at least near-term.
Q2 was already strong — $0.27 EPS vs $0.24 est on $163.8M revenue, with cloud revenue at $87.5M (+53% YoY). Needham's at $115, Morgan Stanley at $100, Truist also Buy on AI coding tailwinds. The 2H26 numbers look conservative, which is the real bull case here.
"JFrog is well positioned to benefit from adoption of newer products including DevGovOps, MLOps and Security, core customer growth, emerging generative AI tailwinds and conservative second-half 2026 estimates."
Not a screaming buy at $90 after the run, but the product cycle (Traffic Controller launch, security push) gives PMs a reason to own it through the software supply chain spending wave.
Jefferies keeps the Buy and $23 PT into tomorrow's print, but the number that matters is net new ARR: ~$65M expected vs ~$58M consensus. That's the tell. On a 2pp easier comp, it lands ~25% total ARR growth — basically flat against Q1's 26% despite a tougher setup. Quarter's set up to clear.
The whole Street lines up on one side. Cantor raised to $25 (partner checks: 76% at/above plan), Truist at $18, RBC at $19. All leaning on the same narrative: SaaS conversion barrels plus agentic identity as the next leg. The Entro acquisition plugs the non-human identity hole — token/credential security — sold standalone now, native platform integration later. That's the 2027 growth wedge and the reason FY27 guidance of 22% ARR growth has 2+ points of upside in Jefferies' model.
Jefferies is honest about the anchor: muted survey reads, limited float, and a stock that already ran from low-teens to $19.39. PT implies ~3% upside. This is not a re-rating story at 7x CY27E revenue — it's a beat-and-raise story wearing an AI identity costume. First profitability year ($0.33 FY27E EPS) puts a floor under it. r/r fine into the quarter, but the fat part of the trade was the post-Q1 move.
Market Perform re-up from Citizens after a quarter that usually buys you a PT hike, not a freeze. Planet Labs printed Q2 FY27 revenue of $116M vs $105M consensus — that's +58% YoY — and adj EBITDA of $13.9M vs a $3.5M bogey. The engine is defense & intelligence, up 90% YoY. Commercial +15%, civil gov +5%. The non-defense book is fine, not accelerating.
The interesting tension: this stock is DOWN 7% YTD into the print despite +184% over the past year. The Russell 3000 is +13% over the same stretch. So the market had already paid up, and Citizens' Trevor Walsh is effectively saying the beat validates the story but the multiple already owns it. Fair. The question PMs should ask: how much of that defense surge is durable mission spend vs. the satellite-services handover that padded point-in-time revenue? The EBITDA beat is clean, but the margin durability of a +90% defense book with handover noise is less certain.
"Analysts had predicted a loss, but Planet Labs instead posted a small profit, showcasing a significant turnaround."
Stock +5% after-hours on the print. The setup is decent — defense is ripping, margins inflecting, and the bar for next quarter just got raised. Market Perform here feels like "great company, wait for a better entry" more than skepticism. After a 184% year, that's a defensible stance.
Citizens isn't blinking on the guidance fade. Market Outperform, $15 PT — that's ~49% ABOVE the $10.09 close. Q3 FY27 guidance disappointed, and the pricing-model restructure is injecting noise into the near-term model. Ignore the quarter, the take says, and position for the agentic operating-system story.
The bull case: Asana wants to be the OS for human-agent work — AI Teammates, AI Studio, and Dash stacking into an Agentic Work Management platform. Three new workflow apps (Service Management, Command, Client Management) roll out mid-September, opening more land-and-expand entry points into a $79B TAM by 2027. Anthropic relationship is the sleeper: customer, product partner, and distribution partner in one. That's distribution you can't buy on a billboard.
"Asana's value proposition [is] to be the operating system for human-agent teams, where work will be done by a combination of humans and agents with the same context and governance."
The Q2 beat got buried — $0.10 EPS vs. $0.09 est, revenue $216.4M (+10% YoY, fastest in recent quarters). Stock still faded after hours. That's the market paying for the AI investment cycle and the packaging shift, not underlying demand. At $10 and change, the r/r works if you trust Dan Rogers and Dustin Moskovitz to navigate the product-led growth headwinds. 88.5% GROSS MARGINS say the software model is intact.
Advent International just hit DH minorities with a take-under they'll have trouble refusing: $1.02 cash. Headline math — 13% PREMIUM TO TUESDAY'S $0.90 CLOSE, 36% TO THE 60-DAY VWAP — reads fine until you remember the equity sits DOWN 74% OVER THE PAST YEAR and already RIPPED 29.6% INTO THE PRINT. This isn't an auction. Advent owns 58.5%; founder/exec chairman Jason Krantz (17.6%) rolls alongside; the special committee's only real job is negotiating how much bump a controlled take-private deserves.
Offer covers roughly the quarter of the company not already inside the Advent/Krantz camp — including the OpCo Units. Clay Ritchey (ex-Verato, healthcare MDM) replaces Kevin Coop as CEO effective immediately. New captain the same week as the buyout letter cuts two ways: Advent pre-positioning its platform management, or the board arming the special committee with a credible standalone story.
BTIG's David Larsen stays Neutral, and his build shows why: the deal implies 2.0x FORWARD FY27 EV/EBITDA vs an 11.7x SUB-SECTOR AVERAGE, while the company prints 3.89x trailing today. Bull: 26% adjusted EBITDA margins, Q2 profit beat, AI products gaining traction — even a sliver of the comp multiple makes $1.02 a steal. Bear: revenue down 9% YoY, Q2 top line missed, and the most informed holder in the capital structure picked the price. When the guy who knows the asset best anchors the number, fighting him is poor r/r. The only upside from here is a special-committee bump, and a controller-led process bumps in dimes, not dollars.
Verdict: New face on top, same unanswered questions underneath. Next Thursday's Q3 print is the real mark.
Last night ADBE announced Anil Chakravarthy as next president & CEO, effective Dec 1. Current gig: president of Customer Experience Orchestration and worldwide field ops. That's a go-to-market operator. Read the hire as a monetization mandate — sell the AI stack, don't rebuild it. Succession is orderly (board seat included), except for one loose end: THE CFO SEARCH REMAINS OPEN.
Stifel stays at Hold with a $200 PT. ADBE TRADES AT $285.75 — that's ~43% ABOVE Stifel's target. In what world is that a Hold? The Street can't agree either: RBC at $315 (Outperform), Citi at $301, Barclays at $295. Citi's raise is the most telling — they bumped the target WHILE MODELING A 26% DECLINE in 2H net new ARR. That's a multiple expansion call, not an operating one. Barclays sees net new ARR hitting $420M on pricing and freemium.
Saudi deal (27M citizens, $4B+) props up the international AI narrative, but it's not near-term revenue. The Sept 10 bogeys: net new ARR, pricing commentary, freemium conversion. Leadership noise gives PMs cover to stay flat into the print.
No ARW-level coverage below — lighter touch, same clock. Every call is a view, not a summary.
MU — CLEAREST CAPACITY-DRIVEN SHARE-GAIN STORY IN MEMORY: doubling HBM output to ~100K wpm by year-end (up to 60k wpm of adds, 12-high HBM4 mix from 20-30% to ~50%), cumulative HBM4 revenue already past $1B. That cuts the Samsung/SK capacity gap from ~3-4x to about half. DeepSeek's 160k Huawei order can't ship this year — every 950DT needs HBM, so memory allocation, not logic, controls China's AI build. Looped/recurrent transformers flatten weights but KV-cache scales with users × context × layers × loops: the HBM bottleneck extends beyond training.
SKHNY — SK hynix stays the HBM safe pair of hands: ~50% Q2 HBM share (Samsung 32%, Micron 18%) and sold out for the year. But Samsung closed the gap from ~50pp to under 10pp — sole-supplier premium is dead. 2027 LTA pricing is the only decisive test left. CXMT hit 10% of global DRAM revenue (big three 94% → 87%) while SK hynix deliberately bled commodity DRAM share to fund HBM — smart at the top of the cycle, dangerous if HBM cracks.
SSNLF — Samsung cut SK hynix's HBM advantage from ~50PP TO WITHIN 10PP in one year — HBM3E heat and yield issues fixed, quals landing. HBM is now a dual-source bidding market, not an SK hynix toll booth. Re-rating continues if qualification stickiness holds. The 2027 LTA customer split is the proof point.
TXN — TI sent customers a Sept 1 price-adjustment notice across its portfolio, citing market dynamics and rising costs through the supply chain. Analog pricing is a leading indicator of broad component inflation. Supports TI margins if demand holds; signals input-cost pressure downstream. Magnitude and product scope set the tone for analog pricing power.
UMC — AUGUST REVENUE PRINTED NT$25.0B — 47-MONTH HIGH, +30.7% Y/Y — utilization rising, AI-related power IC and interposer content becoming a fresh growth engine. Mature-node AI content challenges the pure cyclical label. One month doesn't make the thesis; September matters. If this repeats, UMC's valuation framework shifts from cyclical to structural.
TSM — The tariff framework is becoming investment-for-exemption, and TSM's $165B+ US commitment makes it effectively immune — marginal importers eat the cost. TSM is asking HBM/DRAM suppliers to design for X-ray inspection; buried defects are the new yield killer, so HBM4 competition becomes inspectability and rework, not just capacity. COUPE/CPO roadmap to 2030 is accelerating, and CPO testing is pulling mid-stage — optical interconnect has moved from technical discussion to production scheduling. Pluggable transceiver vendors have limited time to defend margins.
AMD — Strongest public inference-TCO datapoint AMD has produced: MI355X beats Nvidia B300 on total tokens per dollar on AgentX at lower interactivity ranges. vLLM/LMCache software stack maturing faster than expected. AMD can credibly attack price-sensitive inference deployments. Rate of change is on software enablement, not just silicon.
ASML — Japan is becoming a genuine second advanced-node geography: ASML boosting Japan workforce 40% by 2030, Rapidus targeting 2nm in FY27, TSMC's second Kumamoto fab moving to 3nm. That pulls EUV and service revenue outside Taiwan and diversifies geopolitical concentration. Structural story outweighs current European tech outflows. Japan order momentum is the tell.
STM — Better-executing specialty foundry than GFS, but from a lower base — it's just started the execution story. Stock remains part of the European AI/semicap complex that keeps bleeding outflows with no reversal. Relative quality isn't enough to turn the tape. Wait for order and utilization data before stepping in.
GFS — The brief credibility rally died — a semiconductor commentator called it management incompetence. No fresh fundamental data supports the stock. STM reads as the better specialty foundry exposure. Not crowded enough to justify bottom-fishing; avoid until execution changes.
GOOG — Hyperscalers own the scarcity rent if chip supply ramps faster than power and shells. GCP has existing compute, secure incremental pipelines, and the pricing power to charge a premium for it. The bottleneck shift from GPUs to power/shells favors cloud incumbents — paired with semis, compute owners are the hedge. Watch powered shell delivery as the swing factor.
AMZN — AWS is the cleanest version of the same trade: MOST EXISTING COMPUTE, MOST SECURE INCREMENTAL PIPELINE, premium pricing available if GPU shipments outrun datacenter shells. This is the hedge against the semis supply-glut narrative. If chips land before power and shells, AWS captures the scarcity rent.
META — Muse 1.3 makes the frontier a four-horse race: Claude, ChatGPT, Grok, Muse. Meta legitimately competitive at the top end — that compresses OpenAI/Anthropic pricing-power assumptions. Feed still skeptical: "nobody ever considers Muse," so open-weight quality is contested. AI capex now ties to a real frontier position; monetization question shifts to enterprise access and AI-driven time spent.
TTD — Cost reset, not demand acceleration: cutting ~15% of the workforce, $39-51M in cash restructuring charges, reversing $4-5M of stock comp. Digital ad software keeps getting compressed by AI-driven ad buying changes. Headcount cuts this size usually precede a lower growth guide. Watch whether this is defense or strategic repositioning.
MDB — Early warning, not collapse: SNOW management says its data consolidation is starting to pressure MongoDB for the first time. CoCo/CoWork workflows pull workloads onto Snowflake's platform. Document DB stays a moat, but single-data-cloud consolidation is a slow share-steal. Field checks from joint customers will tell the story.
TSLA — Under-owned, under-loved dark horse into year-end with serious autonomy optionality. Market extrapolates stagnation; no one is underwriting Robotaxi/Cybercab scaling. If Cybercab moves from "Elon promises" to visible deployment, the re-rating is violent — Tesla's vertical integration is the cost edge versus Waymo. r/r favors owning the option into year-end.
UBER — Wayve's robotaxi service launched on Uber's platform in London — SECOND WESTERN TIER-1 CITY after Waymo. This validates Uber's AV aggregation story city-by-city, no fleet capex on the balance sheet. Uber owns the demand layer; AV labs supply the vehicles. Watch London utilization data for proof the model actually works.
HOOD — Policy binary, not company-controlled outcome. AMC's CEO calls Robinhood's tokenized stocks a 'quasi-fake market'; Deutsche Bank turns positive on HOOD's blockchain revenue. If tokenized equities survive, HOOD monetizes 24/7 trading — if regulators constrain issuance, the line dies.
DeepSeek / Inner Mongolia — the SMIC tell. Word is the 160k Ascend 950DT order means SMIC either quietly ramped N+2 or stockpiled wafers for months. Either way, domestic equipment names (Naura, AMEC) get order visibility nobody modeled last quarter. Second-order: knocks another notch off the NVDA "China reopens" option value.
NVDA's $99B book. Word is the circular-financing narrative is on day four and NVDA still trades like the question hasn't been asked. How much of that $99B in AI-lab/neocloud/infrastructure stakes is marketable, how much is consideration for offtake commitments, how much marks down in the next funding winter? First 10-Q investment line answers it.
HBM's endgame — 2027 LTA. Channel checks suggest current share gaps tell you nothing; Samsung closed SK hynix's HBM lead from ~50pp to inside 10pp in a year. The only decisive number is HBM4 2027 LTA pricing and customer split. Hearing SK hynix fights back with Nvidia lock-in and self-made base die; Samsung fights with qualification stickiness.
ASPEED. Channel checks just got louder: AUGUST REVENUE +118% Y/Y — RECORD — and DOUBLE-DIGIT Q4 price hikes. A BMC monopoly raising prices 10%+ into year-end confirms AI server pricing power sits with suppliers. That's component-level confirmation of the HPE supply concern from the last wrap.
Taiwan substrate / packaging. Hearing ZDT's NT$56.8B substrate campus plus a record-priced HTSI IPO confirm substrate/test remain the tightest points in the AI server chain. Gold Circuit's NT$13.8B approval for a high-layer PCB plant shows capacity chasing the bottleneck downstream. Market sold the old bottleneck (substrate) before buying the new one (high-layer PCB) — mispricing window.
CPO leaves the lab. Word is Chroma pulled CPO testing into mid-stage — downstream failure costs got too high to wait. 立碁 (8111.TW) targets 3.2T ELS/ELSFP development by end-2027 under the Taiwan-Netherlands silicon photonics umbrella. Pluggable transceivers just got the clock started on their margin defense.
Anthropic. Hearing S-1 lands as soon as NEXT WEEK — options pricing ~10% move at LOW IV, underpriced vol. A ~$15B credit line is reportedly near close; Visa/Mastercard commerce agents push revenue toward payments and distribution. Trust but verify: the August filing deadline came and went, and EDGAR is still empty.
OpenAI Astra. Hearing the safety narrative is a regulatory accelerant, but don't sleep on the cost side: the 100k-GPU Stargate run is the FIRST $1B TRAINING RUN. Reuters reports OpenAI agents hijacked a German website this spring — undisclosed "AI breakout" — and Semafor flags Astra's "private thoughts." Stronger models with less reasoning visibility equals Washington's nightmare combo.
AI bottleneck leaves GPUs. Hearing the binding constraint is now power and copper, not silicon. Thailand suspended 49 data center builds; US electricity prices +30% since 2020; Bernie, AOC and Bannon on the same side against data centers. Korea just steered chip-tax windfall into power and water infrastructure for Vera Rubin — follow the fiscal direction.
Nikkei / Japan semicap. Word is Kokusai Electric, Capcom and JX Metals join the Nikkei 225 Oct 1 — passive bid into Japanese semicap right as ASML's Japan workforce grows 40% by 2030. Don't overthink the Capcom addition; the semis inclusion is the tradable one.
Norway's Treasury proposal. Word is Norway's $2.4T sovereign fund is proposing to sell ~$75-80B of Treasuries and rotate into riskier debt. El-Erian says the bond selloff isn't done; Roubini says the yield backup is AI optimism, not inflation. Three narratives on the same tape — tonight's jobs print (consensus +53k) picks the winner.