Good morning. NVDA's seven-day drip is a de-grossing, not a thesis break — but the WSJ front page labeling Nvidia "a banker to the AI boom" means the market now prices the credit layer ahead of the print. The bar sits low: beat won't cut it. Guide has to answer margin under memory inflation and the accounting on that $105BN OpenAI backstop. US futures green, European semis bouncing after yesterday's rout, oil down another ~3% after Bessent's sanctions landed with a China carve-out (the tell). BTC ABOVE $80K, DXY at 99 — debasement money and AI money have decoupled cleanly. Hot Chips delivered two real signals: AMD's Helios rack lands in the ballpark of Vera Rubin on system watts — first credible parity datapoint — and INTC's EMIB win at SK Hynix is a direct steal from TSMC's CoWoS shortage. Demand-side splits: India's 9,000 Vera Rubin order adds a non-US anchor, but KYEC confirms Rubin volume slipped to Q4. CDS AT RECORD WIDES at NVDA and AVGO, hedging costs doubled — the first question is now whether the $100BN SPV made sense, not how many GPUs ship. Add Microsoft's AI revenue concentrated in fellow techs and today's bull/bear line snaps into focus: genuine demand vs. money that never leaves the system. NVDA's guide won't settle it; the credit market will. Warsh's Jackson Hole debut Friday and PCE this week both hit the long-end — the actual pulse of the circle-financing trade. We'll hit up NVDA and AMD first, then get to the AI credit complex and packaging names.
Thursday's print is a formality. The real signal is in the guide, the Google commentary, and whether management raises the Inphi target. This is still the cleanest way to play custom AI silicon — and the Street just moved in front of it.
FQ2 (July) is a locked deal. Street at $2.714B/$0.93, Benchmark within $5M at $2.709B/$0.93. Nobody's getting paid on the revenue beat — it's the optics after. Rosenblatt flags OPTICAL INTERCONNECTS +25% SEQUENTIAL, and custom-silicon ramps (Microsoft Maia, AWS Trainium, expanded Google) build through FY28/FY29. Susquehanna's read-throughs from Coherent/Lumentum remain robust. FQ3 guide (October quarter) is the battleground: $3.03B/$1.09 consensus, and any upside there gets rewarded. Don't expect a revised multi-year custom framework Thursday — Rosenblatt thinks management saves that for the early October investor day.
Street just shoved bogeys to a $250-310 CLUSTER. Rosenblatt to $300 (from $240) on ~$10.50 FY29 EPS power, 29x. Susquehanna to $265 (from $230), UBS holding the high at $310, Benchmark at $275, Oppenheimer/BMO at $250. One collective thesis: the Google expanded commercial agreement — ~$120B of custom product business through 2033 — is the re-rating event. Barclays frames the warrant math: up to $18.5B of annual incremental revenue if fully exercised.
The structural argument is even better. Chip development costs now run $500M+ PER DESIGN vs $50M a decade ago. That kills the supplier base. Marvell's IP-reuse model means every incremental socket gets cheaper to develop. Scale compounds.
Bull: Google TPU ecosystem is a platform relationship, not a socket win — storage controllers, NICs, AI inference accelerators. Custom silicon doubles in FY28 to ~$4B and hits ~$10B in FY29. Inphi optical franchise compounds with it, and Susquehanna expects a FY27 Inphi target raise. AWS raised capex to $220B (memory-cost driven, but still) and Trainium engagement remains "robust."
Bear: 215% return over the past year means the narrative is largely owned. The Google warrant revenue is an option, not a baseline — it won't convert linearly. And $10.50 FY29 EPS at 29x leaves no room for slips in Microsoft Maia or AWS Trainium ramps. The setup is good; the entry is not.
Benchmark's Cody Acree frames Thursday best:
"With headline numbers tightly grouped, Data Center mix, the October guidance and management's treatment of the new Google relationship should matter more than a small July variance."
That's the whole ballgame. Raise the Inphi target and monetize Google beyond the warrant language — $300+ bulls look right. Stay conservative and defer to the early October investor day — this thing waits for the next catalyst.
FOUR PT RAISES INTO TOMORROW'S PRINT. That's the setup here. UBS to $80 (from $62), Oppenheimer to $80, KeyBanc to $75 (from $65), Rosenblatt to $75 (from $60). The new cluster is $75-80, up from a prior $55-65 range. Piper Sandler is the lone dissenter — cut to $60 on supply concerns but still Overweight. When the entire Street raises into an event, the event better deliver. Fortunately, the setup is clean.
The consensus thesis is one story, not four: VMware migration is the gift that keeps giving, and it has a hard date attached — vSphere 8 perpetual support ENDS OCTOBER 2027. That's a real deadline that forces CIO decision-making. UBS's channel checks back it up with the single best datapoint in any of these notes: one partner's Nutanix pipeline +40-50% YoY.
The second leg: the headwinds that capped growth are clearing. KeyBanc sees license-delivery timing and OEM shipment delays as resolved, which gets revenue reacceleration from ~12% in FY26 to ~14% in FY27. That's the whole ballgame — a growth stock that re-accelerates compresses the multiple debate. Q3 already showed the underlying strength: 700+ new logos, NRR at 106%, 87% gross margins, TCV bookings accelerating above 20%.
KeyBanc's framing is the bull case in one line:
"We think this should support valuation multiple expansion back to historical norms, and accordingly raise our PT."
That's the trade — multiple expansion, not fundamental acceleration off a model rewrite. Rosenblatt kept FY27/28 estimates unchanged and raised PT anyway. Tells you exactly how much of this move is multiple drift.
Not Q4. UBS says a 1-2% beat vs the high end of the $725-745M guide is achievable — that's ~16% growth. The stock doesn't move on that. It moves on the INITIAL FY27 GROWTH GUIDE. UBS pegs ~13% total revenue growth as the reasonable bogey, in line with Street. Oppenheimer wants FY27 above $3.20B. Anything south of 13% and this de-rates fast from 70x. Also listen for hardware commentary — that's the one variable the bulls are hand-waving, and the bears (Piper) are hanging their hats on it. If supply is stabilizing, the short thesis loses its last leg.
Verdict: NVDA heads into earnings with two fresh confirmations — KeyBanc's R200 channel work derisks the near-term print, RJ's CPU modeling extends the story past the GPU cycle. Both land on the same trade: pay 15x for a compounder growing multiples of the market.
The print is right around the corner. Pre-earnings chatter is uniformly constructive — Rosenblatt's the most explicit, calling for an above-consensus print that acts as a positive catalyst. KeyBanc's channel checks support that:
R200 mass production began in July → modest incremental revenue in FQ2. Blackwell flat QoQ in FQ3. R200 shipments +~250K UNITS QoQ in FQ3. Full-year R200 estimate raised to ~1.9M units on loosening HBM4 supply from SK Hynix.
That's a baton pass, not a digestion quarter. The market waited for Rubin to be a step-function, not a replacement — KeyBanc says it is.
RJ went one level deeper. Extended the model through FY2029/CY2028 and isolated the quiet driver: CPUs. ~3% of revenue today, ~5% by CY2028. Doesn't sound big — but RJ's analysis has NVDA becoming the WORLD LEADER in CPU revenue within several years.
"Nvidia could become the world leader in CPU revenue within several years, with CPU growth representing the fastest-growing element in its model."
The market still prices NVDA as a GPU company. RJ is modeling it as a compute platform — Grace rides the same rack as Rubin, so every GPU win drags CPU attach along. Agentic use cases are the pull-through.
NVDA trades at <15x CY2027 GAAP P/E. The S&P 500 sits at 18.6x. Five-year median NTM P/E for NVDA: 35x.
KeyBanc's framing is even starker: 15x FY2028 EPS, PEG ratio of 0.29. The stock trades at a fraction of its own growth rate.
Both firms see the earnings delivering. The disagreement is the multiple. RJ's $352 PT implies 22x CY2028. KeyBanc's $330 implies roughly 15x FY2028. Same conviction, two different re-rate assumptions.
Bull: Rubin is a capacity unlock — HBM4 supply is the gating factor and it's loosening. CPU adds a second growth engine the models barely credit. At a 0.29 PEG, you're paying for the GPU cycle and getting platform optionality free. Side note: NVDA reportedly in talks to invest in Perplexity at a $30B+ valuation — Jensen placing bets at the application layer, which is where agentic demand ultimately gets created.
Bear: Wolfe's the lone cautious voice — rising long-term yields pressure AI capex, and NVDA is a long-duration asset in that frame. Plus the stock has delivered ZERO alpha this year: +12% YTD, matching the S&P 500. If the multiple stays pinned at 15x, even a beat-and-raise just marks time. You need the re-rate to make money from here.
Morgan Stanley finally came off the sidelines: upgrade to Overweight, PT to $65 from $58. UBS already at Buy/$65. The street is converging on one thesis — observability demand is inflecting and Dynatrace's renewal wall does the heavy lifting into FY27.
MS says this is the healthiest observability market since 2022. Public cloud, software dev initiatives, early enterprise AI dollars. That's the macro. The micro is the renewal cohort: Platform Subscription renewals 50% LARGER than the FY26 cohort, 70% weighted to 2H FY27, with underlying usage tracking above 20%. That's growth already installed and consuming.
UBS frames the same setup from the pipeline side, fresh off the IR call:
"We didn't drain the pipeline, we have a lot of large meaningful deals in the pipeline."
UBS sees three-year DPS renewals skewing toward larger ACV platform consolidation deals. That's the consolidation trade — ripping out point tools, going platform. MS makes the same call: Dynatrace is one of the few platform consolidators with the breadth to win.
MS model: net-new ARR growth re-accelerates to 20%+ in FY27, compounding into 20%+ total ARR growth FY28-29. UBS is more conservative on the organic constant-currency bogey — high-teens by end of FY27 — but still values the stock at ~20x CY27 estimated FCF. For an 82% gross margin business re-accelerating, that's not a demanding multiple. Stock sits at $49.53, PTs clustered $60-65. That's 20-30% upside.
Bull: demand inflecting, renewal cohort massive and usage-backed, Arize ($915M, ~22x ARR) adds AI workload relevance without wrecking the model. The GTM overhaul from two years ago finally showing up in sharper execution.
Bear: the 41% organic net-new ARR print flatters. Strip out the eight-figure new logo pulled forward from Q2 and it's no higher than 22% — solid, but not the acceleration the headline screams. And $915M for Arize is a real bet that agent observability is a category owners pay for, not a feature.
Q1 was clean: EPS $0.48 vs $0.44 est, rev $554.5M vs $549.3M, full-year outlook raised. $275M buyback in the quarter. DA Davidson reiterated Buy post-Arize; Scotiabank to $61; Canaccord to $60. Twelve analysts just revised estimates up. The setup into FY27 is about as good as it gets — the question is whether the 2H-weighted renewal story gets crowded before the prints actually land.
Verdict: The tape is priced for perfection, and we get the proof in a week. PANW up 90% YTD, 8.5% MTD, trading at ~306x trailing. That's not a valuation — that's a dare. (And heads up: BofA says September 4, Benchmark says September 1. Two good desks, two dates. Figure that out before you size the overnight.)
Six firms in the window, and the PT cluster is tight: $390-425 (Cantor at $425, BofA at $420, Stifel and Citizens at $415, Benchmark up to $400 from $340, UBS at $390 Neutral). Nobody's short. At 306x, nobody needs to be — the multiple does the hedging for you.
Platformization is the whole thesis. Every desk lands on the same three bogeys: NGS ARR growth, RPO, and large deal momentum. BofA models NGS ARR +59.9% YoY, revenue +32.1%, RPO +32.6% — all in line with consensus. So the bar is simple: beat and raise, or get clipped.
Benchmark's channel checks show healthy demand across the AI security stack — network security for AI infrastructure, Prisma Cloud, SASE, AIRS, Cortex XSIAM. Cantor's partners are selling ahead of expectations, and the Frontier AI Critical Defense Program gives the AI narrative a product hook, not just a slide deck. Bulls argue this is the first real monetization of the agentic security wave, and PANW owns the distribution.
"Investors now require continued evidence that AI enthusiasm is converting into measurable ARR growth, accelerating platformization expansion, and a compelling fiscal 2027 outlook to justify the premium valuation."
That's Benchmark's Yi Fu Lee — with a Buy rating. Read that again. A bull is telling you the multiple only works if FY27 guide is a cannonball, not a wading pool.
Bull: Platformization compresses the sales cycle, NRR stays strong, and AI security opens a TAM that isn't in the model yet. Momentum begets momentum at this point — the 90% YTD move is a feature, not a bug.
Bear: One NGS ARR miss or a tepid FY27 guide sends this 15% lower overnight. Nobody's modeling a hardware haircut — memory, storage, and supply chain constraints are real, and BofA flags them explicitly. UBS at Neutral is the lone voice saying the CY28 math doesn't support chasing.
Three bogeys, in order of importance:
1. NGS ARR — ~60% growth is the bogey. In line is meh. Beat is green. Miss is a hole in the book. 2. RPO — 32.6% modeled. Tells you if platformization deals are actually landing or just being announced. 3. Hardware revenue — the hidden risk. Supply chain constraints on memory/storage could clip the July quarter, and nobody's positioned for that.
July is seasonally soft. Management has a "disciplined track record" of finishing strong, per Benchmark. They'll need it — the stock has already priced the beat, and the entire question is whether FY27 guidance gives the algo-buyers a reason to stay.
Bottom line: I'd rather own this into the print than chase after it. The r/r is tighter than the PT parade suggests. High reward, but at 306x, "good but not good enough" is a 10% drawdown waiting to happen.
All buys, no sells. ServiceTitan prints F2Q27 on Sept 3 and the street's already celebrating — 5 FIRMS, PTs CLUSTERED $103-$125, ZERO NEGATIVE REVISIONS off a ~48% six-month tape. BMO holds Outperform/$103 (no estimate changes — reads as "we already have it right"), KeyBanc holds Overweight/$120 and calls TTAN one of its highest conviction ideas.
The collective thesis is remarkably consistent: Max is the story. Piper frames it as the premium AI add-on with potential to DOUBLE subscription revenue — everyone else just calls it the subscription growth engine, alongside Pro, commercial, and enterprise residential. BMO's angle is execution on the incremental growth initiatives, specifically Max adoption and commercial momentum. Stifel expects a top-line beat of ~5pp and a LOW-20s GROWTH GUIDE for FY27.
KeyBanc added a seasonal wrinkle worth trading: smokier summer = more service calls = usage tailwind. They pencil ~$4M into the beat. (Love a good weather-driven SaaS factor.)
Bull case: 24% LTM revenue growth, 71% GROSS MARGIN, take-rate expansion still early innings. Max is a land-and-expand machine — subscription strength has driven beats "similar in magnitude to the past several quarters" with no sign of fatigue. Commercial is the next leg. At $94.94 against a $103-$125 target cluster, the r/r into a catalyst is skewed up.
Bear case: Insiders are selling into strength — Iconiq Strategic Partners dumped ~$6M, director Griffith also trimmed — not a red flag, but not a warm fuzzy pre-print. And after a 48% run, the bar is high. In-line beat + low-20s guide could trigger "buy the rumor, sell the news" even if the quarter is clean.
KeyBanc's note title does the work:
"Where There's Smoke, There's Buyers; We approach ServiceTitan's F2Q27 results positively, seeing potential for revenue upside to be similar in magnitude to the past several quarters, again driven by subscription strength from Max, Pro, commercial, and enterprise residential."
Buy the dip into the print. One-sided bullish street, proven beat cadence, a weather tailwind, and a guide that's already telegraphed. The risk is post-print profit-taking, but the PT cluster says there's still meat on the bone. Insider sales are small-ish — not sure we can read too much into them — but worth flagging for PMs who care about optics. Net: TTAN is the kind of setup you want into a print — consensus is right, but nobody's positioned for a miss, and the upside skew is real.
Raymond James just became the loudest bull in the room. Upgrade to Strong Buy from Outperform, PT to $641 from $565 — and the thesis is bigger than GPUs. They extended their AI Factory framework to server CPUs, calling a 44% five-year CAGR to ~$201B in CY2030 revenue for AMD.
The logic: accelerators do the dense math, but CPUs manage those accelerators and run the retrieval, database, app, security, sandbox, and tool workloads around every model call. That's a huge TAM expansion for EPYC, not just MI-series. Agentic workloads = more CPU per inference, full stop.
The honest caveat: RJ explicitly says workload growth does NOT translate one-for-one into processor shipments. Higher utilization, software efficiency, custom silicon, and offload absorb part of the demand. So this isn't a straight-line volume story — it's a value-per-workload story with a nonlinearity discount baked in.
Worth noting where RJ lands: their $201B forecast sits between NVIDIA's $200B long-term framework and AMD's own $220B estimate. They're not the craziest bull — they're the most constructive on AMD specifically:
"AMD offers the strongest combination of direct earnings leverage, datacenter positioning and market-share gains among the companies it covers in the sector."
Already on board post-Q2. Data Center revenue +107% Y/Y, +16% Q/Q — that's the number carrying everything. DA Davidson at $550, BMO initiated Outperform, TD Cowen Buy. Benchmark held Buy through the 8-9% post-earnings flush, blaming expectation fatigue after Advancing AI and Intel's strong print rather than a broken thesis.
The one dissenting note: Phillip Securities trimmed FY26 profit forecast by 8% on client and gaming downturns. Fine — nobody's paying for consumer AMD right now. The datacenter story is the whole story, and RJ's PT implies the Street is only starting to reprice the CPU-as-AI-traffic-cop narrative.
Bottom line: $641 is the high-water mark. If RJ's 44% CAGR holds, AMD is a $1T market cap story in the making. Long way from the $550 cluster — but that's what conviction looks like.
Truist says pricing power wins. And it's slapping a $165 target on it. Up from $155, Buy maintained, while the stock trades at $83.62 — down 17% on the week. That's a PT roughly 2x spot. Bold either way.
The thesis: CoreWeave pushed prices up 25% across the lineup in July, just ahead of Nvidia's 17% GPU price hike. Net-net Truist models a 9-percentage-point margin uplift on longer-duration deals, with contribution margins on future contracts jumping from ~24% to ~33%. Only Blackwell and Vera Rubin are caught in the price increase — shorter-duration deals and re-contracting older-gen GPUs are extra upside.
"Truist estimates the pricing dynamic will result in a net 9-percentage-point margin uplift on longer-duration deals."
The offset is capex: Truist figures ~$8B extra per 1GW of capacity from the GPU hikes. That's real money against a FY26 capex guide of $35-39B. And not everyone's buying the bull case — Bernstein stays Underperform at $74. Truist's answer: the 25% pricing action more than covers the 17% cost inflation, and with revenue +115% LTM and FY26 guidance raised, the topline gives the pricing power room to work.
Needham bumps SK Hynix to $220 from $200 (Buy) after the board approved a ₩40T BUYBACK — ~24M shares, 3.3% of the float. That's management calling the stock cheap at 7.4x P/E, and Needham is listening.
The bigger tell is the shareholder return framework: SK Hynix moved its 2025-2027 cumulative payout target from "within 50%" of FCF to "OVER 50%." Needham now models ~₩500T cumulative FCF and ~₩250T of buybacks plus dividends by end-2027 — a staggering amount of capital heading back to holders. The new PT is just ~6x CY2028 non-GAAP EPS, specifically adjusted for the lower share count.
"The decision reflects management's view that intrinsic value, driven by business competitiveness and cash generation, is not fully reflected in the current share price."
The buyback doesn't fix the bear case — memory pricing is softening and inventories are building (Goldman flags both). But with Rosenblatt at $320, Barclays at $300, and Needham now at $220, the entire sell-side coverage is Buy-rated. At 7.4x with 3.3% of shares coming off the table, the buyback creates a valuation floor while HBM leadership supplies the upside narrative. That's the setup.
STIFEL REITERATES BUY AT $350, AND THE MESSAGE IS SIMPLE: THE MULTIPLE RESET, THE THESIS DIDN'T. Stock's down 18.5% in a week to $230 — brutal tape, but that's the entry point Stifel wants you to pay attention to. They see July quarter EPS beat ($1.20 vs $1.17 consensus) and expect October guidance above the $518M consensus. (Funny nuance: their own revenue estimate is $470M, below the $472M consensus — they're not even modeling the beat they're calling for.)
The engine is unchanged: AEC ramps across multiple hyperscalers, the $600M+ optical ramp, and ScaleUP as upside optionality in 1H FY27. 68% gross margins with >200% revenue growth — this is a quality compounder that just got cheaper. Valuation reset while estimates went up; that's the r/r PMs should care about.
"Credo's valuation has reset while estimates have risen, creating an improved entry point ahead of potential upside to its greater than 80% fiscal 2027 growth outlook."
Street's on the same page — Evercore $325, TD Cowen $260, Rosenblatt $215 all pointing at the same optical/AEC expansion. The $350 PT (42.9x CY27) is defensible, but the guide matters more than the target. This is a print-driven stock into earnings.
BABA JUST RAISED $10.3B (HK$80B) FOR AI — primary placement of 710M shares at HK$112.70, a 3.6% discount to the ADR close. Dilution is real: 3.7% of pre-deal share count. But BofA keeps Buy and $172 PT (45% upside) and frames it as growth financing, not desperation. Net cash goes from $31B to >$41B pro forma before deployment.
Bear case is straightforward: aggressive AI infra capex means rising depreciation, margin pressure, and possibly more external funding down the pipe. Baird already cut PT to $160 on exactly that. But BofA leans on strong cloud guidance, improving AI investment economics, and sequential FCF recovery. At $118.47 and 29.5x earnings, this is a stock that needed to fund the AI story properly — now it has.
"The firm views the transaction as a mix of growth financing, funding diversification and pre-emptive balance-sheet strengthening."
Benchmark ($220) and Mizuho ($195) are the high-side outliers. BofA at $172 is the cautious bull — constructive, but not reaching. No one's walking away from the cloud/AI thesis; the debate is purely about how much the buildout costs before it pays off.
Canaccord steps up: PT to $225 from $200, Buy maintained, and the thesis is cleaner than the Silver Lake chatter suggests. Growth stabilizes in the low teens, margins keep grinding higher, and that combo gets you to mid-to-high-teens FCF-per-share growth. The math: ~14x EV/FCF on FY28/CY27 estimates plus net cash — roughly 4.5x forward revenue. Same multiple as before, higher estimates. That's the whole ballgame.
Recent print supports it: REVENUE +13.3%, GROSS MARGIN 75.8%. Pair durable bookings momentum with operating leverage and the standalone story works — no transaction required.
"The standalone stock deserves a higher valuation even if a transaction never materializes."
Silver Lake interest changes the near-term setup, but Canaccord's case doesn't lean on it. The rest of the Street is mostly in the same camp: Oppenheimer and Cantor at $220, Guggenheim at $275. TD Cowen sits at $220 but holds (mixed checks). Barclays at $200, Overweight, expecting muted top-line beats with margin upside as the offset. Consensus read-through: demand stable, AI products gaining traction, margins the real upside lever.
Stock's around $199. That's ~13% upside on standalone fundamentals. Decent r/r, not a home run — but the margin trajectory is the thing to watch into the print.
RJ's the first clean SOTP on the new Arm — and the fabless pivot is the whole ballgame. Headline says PT to $641 from $565, but the body math lands at $272 (78x on royalties/licensing EPS, 35x on fabless) — ~14% upside from here. The structure matters more than the bogey: value the IP annuity separate from the merchant silicon business, because Arm is no longer just an IP licensor. (243x trailing P/E with 97.5% gross margin — the premium IS the debate.)
Own CPU ships late FY2027. SERVER ROYALTY IS <20% OF TTM ROYALTY REVENUE TODAY, GOING >1/3 BY FY2028. That's the rate-of-change the Street keeps trying to price.
The gap that matters: RJ models $1.4B fabless CPU sales in FY2028 vs mgmt's ~$1B guide and the $2B in demand mgmt itself disclosed on the last call. Either mgmt is sandbagging or TSMC capacity is the true ceiling. Bull case: CY2028 CPU revenue at $2.7B, royalty at $5.4B.
Less than 20% of trailing twelve-month royalty revenue comes from servers, but this is expected to exceed a third of revenue in fiscal year 2028.
Everyone else clusters $250-326, all Buy/Outperform/Overweight. Rosenblatt sits low ($250) on margin/expense pressure; BofA ($260) and Wells ($280) are pure multiple and valuation calls. Guggenheim ($255) leans on AI strength as the smartphone offset. Evercore's the bull at $326, leaning on the 12% EPS beat in June. And 25 analysts just revised estimates up — the direction of travel is clear, nobody's just willing to pay up before the fabless silicon actually ships.
Truist hosted CFO Grabscheid and VP IR Schreiner last week and walked away MORE constructive on the durability of the story. Reiterates Buy, $110 PT — roughly 23% above the $89.38 print. The near-term bogeys: coding agent adoption and monetizing binaries as AI-generated code reaches production. That's the AI volume story, and it's real.
The security narrative is the quiet compounding engine — management pushed it for the near AND long term. But the real optionality is ModelOps. Nascent today, Truist sees it becoming a material contributor long-term, with an actual OpenAI angle. That's the kind of call option the market isn't fully paying for yet.
THE MOMENTUM IS CONFIRMED. Q2 printed $0.27 vs $0.24 est on $163.8M revenue. Cloud +53% YoY to $87.5M. NDR 121%. Needham caught up in a hurry — PT to $115 from $80. Morgan Stanley nudged to $100 (Equalweight, grudging). THIRTEEN ANALYSTS HAVE REVISED UP since the print. At 78% gross margins, this is a compounder wearing a story-stock costume.
"Truist believes ModelOps is nascent but could become a material contributor in the long term, including an opportunity with OpenAI."
Multiple paths to sustained growth past FY26. Rare for a $9B software name at this stage — and the reason Truist left the meetings incrementally more positive.
DA Davidson tweaks PT to $155 from $145 — but keeps Neutral, and with the stock at $157.36, the market's already trading through that target. This is a "right story, wrong price" call.
The firm's conviction is up on Emerson's spot in the power supercycle: data center demand, LNG capex, nearshoring. They model FY28 EPS at $8.20 and apply 18.9x (just under the 5-yr median 19.4x) to get there. Not exactly bold multiple expansion. But they also note the perfect Piotroski Score (9) and 53% gross margin — quality is real.
"DA Davidson's model suggests achievement of Emerson's 2025 analyst day targets are well within reach."
Offsets: project delays, trade/conflict headwinds. Q3 beat with adjusted EPS $1.71 on $4.87B rev, FY guide raised — yet shares faded after hours (classic). Citi's take on a short-cycle recovery with data center spillover into broader industrial is the real swing factor. Just not enough cheapness here to chase at current levels.
Scotiabank ups PT to $114 (from $95) into Thursday's FQ2 print — and they're not alone. The Street's clustering at $100–$120 (Guggenheim $110, Goldman okay at $106, Loop initiates buy at $100, Cantor and Oppenheimer both at $120) on one collective thesis: cyber resilience is an AI preparedness budget line item, and Rubrik's security-first architecture is the wedge. EIGHTEEN UPWARD ESTIMATE REVISIONS into the quarter. Market expects a beat and a raise.
"Rubrik’s positioning among backup and recovery vendors to benefit from AI preparedness spending due to its security-first architecture."
The bull case writes itself — backup is becoming a security control, and Identity Resilience keeps surfacing in fieldwork as a priority. The bear case is sharper than the headline suggests: former CRO Brian McCarthy is raiding the building for Cursor, and the count is ~30 senior leaders and quota-carriers out the door. That's go-to-market disruption, not demand destruction — but it lands at the exact wrong time for a company trying to guide up with a new sales org. Hardware pricing and lead times stay elevated too.
Stock's +50% over three months. A chunk of this move is already in the tape. But PTs keep climbing and the revision flow says the fundamental narrative is running ahead, not catching up. PMs fading this print are betting on sales execution risk, not product demand. The guide Thursday tells you who's right.
Oppenheimer starts Kodiak at Outperform with a $9 PT — 137% upside from $3.79, and the stock is sitting ON the 52-week low ($3.76) after a 67% drawdown over the past year. The call is a direct challenge to the market: this isn't a pre-revenue AV story, it's a Physical AI platform already running customer-owned driverless trucks in commercial revenue service.
The thesis hinges on the asset-light, recurring Driver-as-a-Service model — Kodiak provides the autonomy stack, customers own the trucks. Multi-domain expansion across long-haul trucking, industrial trucking, and defense creates a data flywheel (synergistic learning, as Oppenheimer frames it), with the long-haul highway safety case targeted for late 2026.
PT math: 8.0x 2030 revenue estimate of $550M, discounted back two years at 15%. That implies a massive leap from the current run rate — Q2 revenue came in at $3.5M, +91% QoQ on driverless fleet expansion, but that's still roughly $14M annualized. $550M by 2030 is a ~40x jump.
Bear case writes itself: 67% decline, narrowed FCF guidance, heavy losses, and a long-haul catalyst that's still a year-plus away. Oppenheimer's rejoinder is that the market is paying zero for the defense franchise and the 2026 safety-case catalyst.
"Kodiak is demonstrating synergistic learning across its multiple operational domains."
Not sure this is more than a call option on long-haul autonomy clearing the safety bar — but at $3.79 with a PT 2.4x higher, the r/r skews favorable. The stock's already priced for failure; Oppenheimer is just asking you to price in the execution.
"50% of checks positive in the second quarter of fiscal 2027 versus 22% in the first quarter of fiscal 2027."
Earnings in two days. Stock's UP 38% SINCE FQ1 at 44x. Expectations aren't the problem — the upside bar is.
Everyone's looking for a beat-and-raise: $905M revenue, $742M billings consensus. TD Cowen thinks the 17% R&D sub growth estimate is conservative vs a 20% trailing three-quarter average. Stifel sees 15-16% sub growth with a 2H guide bump. Buy ratings everywhere — PTs from $225 (Oppenheimer) to $276 (Guggenheim) — but TD Cowen's Buy PT sits at $235, BELOW the $248 tape. Bullish analysts with marks already overtaken. Crowded trade into the print.
Following Eli Lilly's CRM commitment, TD Cowen expects focus to shift toward Veeva's artificial intelligence roadmap.
That's the multiple-expansion story: Copli acquisition (now Falcon MLR, 70% manual labor reduction target), Vault AI, Falcon AI. Labor augmentation, not replacement — smart framing. But at these levels, the 2H guide carries the weight. Beat + hold and the stock gives back some of that 38% fast. Beat + raise and the $276 crowd looks prescient.
BMO's not budging — Market Perform, $200 PT, and that's a BIG spread vs. Citizens' $280 (nearly 50% apart tells you the Street has no idea what this thing is worth until the churn cycle breaks). The bear case in one line: AFFO/share gets crushed by elevated churn and debt refinancing THROUGH 2027. The bull case: that's priced in, and 2028 is the reacceleration year as spectrum-auction leasing and buybacks kick in.
The stock at $183.28 vs. the $200 PT leaves some meat on the bone, but this is a "wait for the turn" setup, not a "catch the knife" one. Q2 was a beat ($1.87 vs. $1.84 est., rev $715M vs. $706M) and they raised guidance + restarted buybacks — but that's table-stakes support, not a catalyst.
"Management sees a clearer path to reacceleration in 2028 and beyond as headwinds subside."
Edge compute is the optionality everyone wants to talk about — BMO says show-me, and they're right. Financing costs and churn are the bogeys. Watch the buyback pace as a tell: aggressive repurchases at these levels = management thinks the floor is real.
BofA keeps Buy / $19 on XPEV after Dogotix closed a ~$900M round — and that's the only story that matters today. The subsidiary raised $600M from external investors (IDG, Alibaba, Tencent, Gaorong), $200M from XPeng itself, $100M from management. POST-MONEY: $6.3B. XPeng keeps ~73.8%.
"The fund raising sets a new record for a single private equity financing round in China's embodied AI sector."
That's the bull case: external validation at a serious mark. XPeng's stake is worth ~$4.6B at that print — the market is giving the entire EV business almost no credit while the stock sits at $11.15, a hair off its 52-week low. BofA sees 70% upside to $19.
The bear case is the base business. Q2 was a double miss: revenue RMB 19.7B (+8% YoY, +51.5% QoQ) came in light, and ADJ EPS OF -$1.29 cratered vs. the -$0.29 bogey. Vehicle margin still soft — growth is coming from services and VW, not car sales. Macquarie and Bernstein both trimmed PTs to $18 (Outperform / Market Perform); Tiger went to $15 (Hold).
Not sure we can read too much into BofA holding its PT — that's a reassertion after a headline event, not fresh analysis. The actionable signal is the round itself. Blue-chip money at a $5B pre-money makes XPEV a call on Dogotix, not on EV margins. Until the auto trajectory stabilizes, that's a lottery ticket with a $19 target.
Verdict: Canaccord goes against the grain — RAISES to $175 while the rest of the street cuts. The stock's ALREADY +32.5% THIS WEEK, so some of this is chasing. But the argument isn't really about Bitcoin price — it's that the treasury model survived the stress test.
Joe Vafi's framing: the past few months tested Strategy's model, and both common and preferred (STRC trades back near par) came out battle-tested. The balance sheet now has real dry powder — current ratio 5.39x, plus a NEW $1.59B CASH POOL earmarked for BTC purchases and treasury ops. That's the acquisition engine restarting, and Canaccord sees mNAV retracing to ~1.2x from ~1.01x. Do that math and you get ~20% re-rating BEFORE any BTC move.
"The balance sheet has been strengthened with dry powder available for more than servicing dividends, reflected in a robust current ratio of 5.39."
Macro tailwind supports the timing: Canaccord reads the Treasury as focused on flattening the long end — that liquidity impulse is lifting crypto spot. Their PT bakes in ~20% BTC appreciation on a forward twelve-month basis.
The rest of the street isn't as constructive on the underlying. Cantor cut to $186 (BTC at $98K), B.Riley to $155 (BTC at $65K), Benchmark slashed to $435, Wainwright holds $325. EVERYONE STILL AT BUY. The real signal: the PT range is SILO-WIDE ($155–$435), which tells you this is a pure BTC beta trade with different input assumptions, not a fight over the business. Down 64% on the year, the common's been through the wringer. The bull case is simply that the worst of the deleveraging is over and the buy button comes back online.
Baird just called MRCY an "aggressive buyer" at $91.36 — and they're right. The 19% post-print selloff is a gift if you trust the bookings tape.
Here's what spooked the market: Q4 revenue of $289.8M beat (+6.1% organic) but adjusted EPS of $0.37 missed the $0.38 bogey. That's the whole bear case — margin pressure. Fine. Now the bull case: BOOKINGS +93.1% Y/Y, BACKLOG AT $1.945B, A RECORD, UP 38.4%. Multi-year visibility doesn't get much better in defense tech.
Three firms all landed on the same side of the boat. Baird reiterated Outperform/$130 and slapped a Bullish Fresh Pick on it (expires 12/31/26). Piper Sandler raised to $131. RBC moved to $120 from $105. That's a $120-131 PT cluster — no one's blinking.
"The recent selloff following the fourth-quarter 2026 results does not reflect the company's strong execution, multi-year visibility, and margin inflection story that continues through the late-decade period."
The kicker: the guide doesn't even include CPA platform follow-on awards, the seven-year missile framework agreement exposure, or the PLTR agreement. All upside. 4 analysts revised estimates upward since the print — that's the rate-of-change tell.
Steelmanning the bear: EPS miss is an EPS miss, and defense primes hate margins surprises. But at 19% down on a beat-and-raise visibility story, the r/r is clearly skewed. This is a position to own, not trade around.
The take: KeyBanc spent two days with management and didn't blink — Overweight reiterated, $65 PT against a $35.65 tape. That's 82% implied upside. The entire debate reduces to one question: is the bookings wobble timing or share loss? KeyBanc's answer is loud: timing.
KeyBanc hosted CFO Baird Radford, President/COO Nnamdi Njoku, and IR David Unger across Boston and NY. Management's pitch — the widened bookings guide is deal slippage, not demand destruction. Mid-to-large competitive RFPs are sitting in active evaluation right now. That's the leading indicator to care about.
"The widened bookings guidance reflects deal timing rather than softening demand or competitive losses."
The market sold the Q2 print anyway — EPS $0.94 vs $0.42 expected, revenue $312M vs $305M — and shares still faded premarket. Fair pushback: there's a one-time tariff refund in that beat and the near-term guide went the wrong way. But KeyBanc's structural counters have teeth: competitor leases are expiring, which historically unlocks switch activity, and Omnisphere is resonating ahead of the full go-to-market push.
At $1.62B market cap, this is a cheap options on the pipeline being real. KeyBanc says the pipeline is the largest in years and still growing. If that's accurate, $65 is arguably conservative. If it isn't — if the guidance widening is actually competitive share loss — you're catching a falling knife. The RFP conversion data over the next two quarters resolves it either way.
Rosenblatt staying long into tomorrow's FQ3 print — Buy, $575 PT vs $394.51 spot. That's a 46% implied upside on a name sitting 36% off its $615.79 high. The bet: Ansys is the compounding engine now, and the market has over-discounted integration risk.
Revenue bogey $2,435M, +40% YoY. Ansys alone was ~29% of revenue last quarter. Gross margin 82.6% gives cover on the ~40% op margin estimate. Nine analysts revised FY numbers up into the print. ARC processor sale to GlobalFoundries closed in June — portfolio cleaner, focus shifts to interface/foundation IP and clawing back China traction.
The real marker is September 30 Investor Day. Medium-term targets, Agentic AI strategy, next phase of Ansys integration. Rosenblatt's F2028 model: 12% revenue growth, ~1% margin expansion. That's the bull case in a nutshell — modest acceleration, disciplined execution, no heroic assumptions.
Benchmark ($570) and KeyBanc ($600) echo the same side. S&P flipped the outlook to positive after the $4.3B debt paydown. Nobody's short this into the print — but the 36% drawdown tells you the Street needs to see IP segment stabilization and Ansys synergy timing before paying up again.
AVGO — CDS +5BPS TO AN ALL-TIME HIGH and the market is finally splitting the semiconductor franchise from the financing operation. Credit repriced before equity; that's the tell. Anthropic's internal silicon team shifts negotiating leverage permanently — not an order cut, but a permanent discount on AVGO's pricing power going forward. The junior tranche print is the real verdict. Harder than any guidance.
INTC — SK hynix adopting EMIB for next-gen HBM, joining Apple and MediaTek as external packaging wins. The rate of change is accelerating on packaging while the merchant foundry story still lacks the one missing ingredient: an external 18A-P customer. Hot Chips showed a credible 18A-P roadmap (Diamond Rapids, Crescent Island, Wildcat Lake) but own-Xeon-on-own-process isn't the same as winning merchant business. Narrative arb until a tapeout confirms.
ASML — The MATCH Act DUV servicing ban is the real policy tail, and the cheap kill shot is installed-base service, not new machines. China is ~1/3 of revenue and the high-margin service book gets hit too. Legislative proposal, medium confidence — but the direction matches Washington's escalation pattern. Don't price the rumor; reprice at landing. Asymmetric downside if it moves.
TSM — CoWoS tightness just cost TSMC a marquee packaging slot to Intel EMIB. Slow-burn competitive vulnerability, not an order cancellation — but second-source openings compound if tightness persists. Hot Chips still showed the reference flow (SOIC, CoWoS-L) as the technical leader. Bottleneck is capacity, not capability, and that bottleneck is feeding competitors. Rapidus state money (¥2.4T R&D aid, ¥300B more, 80% guaranteed loans) keeps 2027 pricing honest.
MU — The structural-shortage repricing thesis is accelerating, not the cyclical one. Micron's own data shows the memory wall widening: AI compute triples every 2 years, HBM bandwidth less than 2x. BUT the Rubin Ultra HBM content cut to 192GB from the previewed 1TB is the first real crack in the bit-growth curve — bulls need to re-underwrite. Buyback calls from memory bulls = management confidence signal. NVDA prints 8/26, MU 9/23. Everything else is chop.
SNDK — Same shortage thesis as MU but the pure-play NAND angle. The AI specific is HBF and advanced NAND from AI servers, not commodity. Buyback talk is louder here — a floor under the equity while the narrative plays. Bear offset: Samsung Xi'an V9 plus SK hynix Dalian adds ~70–80k wafers/month of advanced NAND into 2027. Supply response is the overhang.
SKHNY — EMIB adoption is a rare second-source move in HBM — secures packaging capacity without waiting on CoWoS. Union rejection by 25 votes (50.08% against) is noise, but in a full-production market with lead times into next year, labor tail risk amplifies. Miyagi fab evaluation = greenfield Japan memory capacity, a geopolitical capex response. Micron's HBM4 spec (2800 GB/s, 2x I/O) now sets a public bar — the roadmap race is on.
SSNLF — Volume-producing an NVIDIA inference chip is a rare external anchor for Samsung foundry and confirms the mix shift from training to serving economics. DRAM capacity discipline is the friendlier read: Omdia sees +6% in 2026 and +1.3% in 2027 — supports a longer upcycle. Xi'an V9 conversion adds NAND supply into 2027. The US pressuring Korea on US memory fabs is a capex tail risk building quietly.
MSFT — Bloomberg's AI revenue concentration datapoint is the single most important demand-side item this window: if the biggest buyers of AI are other AI companies, the circle-financing argument just moved from balance sheets to income statements. First large-scale Vera Rubin deployment validates infra, not end-market breadth. Demand loop is increasingly self-referential.
GOOGL — Jeff Pu modeling TPU shipments surpassing Nvidia by 2028 is the biggest accelerator share-transfer datapoint if credible. Google controls both silicon and the largest consumer AI distribution — structural cost hedge. Ads still printing pricing power (GOOG/META up >20% since 2021) but Amazon is eating the marginal ad dollar. Slow share shift, not an inflection.
META — Hatch agent platform with a $199.99/month premium tier across DoorDash, Etsy, Reddit, Yelp, Outlook is the clearest consumer AI monetization path in the market. Product cadence accelerating. The $100B capex / $12–15B opex cut tension is real shareholder frustration, but it keeps 2027 spend ROI-gated — better discipline read, slower growth read.
AMZN — Retail media is the relative share gainer: GOOG/META CPMs up >20% since 2021, Amazon rose less and is harder to avoid. Secular ad-share shift, not a quarter event. Brands have fewer places to go, and Amazon owns the shelf.
ORCL — NVDA's 15%+ server price hike is a direct margin squeeze on leveraged AI cloud. Pricing power shifted upstream, and Oracle can't pass through GPU cost without breaking the backlog economics. The Taiwan read says NVDA picks 75% GM over share — that's a neocloud competitiveness problem as much as a cost problem.
QCOM — Xiaomi's Xring O3 (ARM CPU/GPU on TSMC 3nm) takes the in-house silicon trend into the flagship Android tier. Lost content per phone, and QCOM is more exposed to the flagship than MediaTek. Rate of change: Chinese OEM design wins are accelerating.
MDTKF — North America's first LTM field test with Ericsson/AT&T, 25% handover interruption cut — 5G Advanced credibility for the modem roadmap. But Xring O3 sits right on top of MediaTek's high-end target. Defend the flagship or lose it.
IBM — Dual-ISA Z mainframe with native Z and ARM plus HBM3E is genuinely differentiated silicon. Revives IBM's relevance and adds an HBM demand vector outside the GPU complex. Watch if it converts to system revenue.
LRCX — Samsung and SK hynix finalizing etch tool POs for China NAND — Xi'an V9 (~40–50k wafers/month) and Dalian (~30k). V9 at ~280 layers is etch-intensive. Tools start arriving next month; POs finalize around year-end. Order book turning from narrative to purchase orders.
AMAT — Broad deposition/etch exposure to V9-class NAND capex. Same China expansion event, same 2027 wafer starts. Equipment capex is inflecting up.
TEL — Etch/clean tool leverage to the same memory capex uptick. Purchase order finalization approaching year-end. Direct beneficiary of the China NAND build.
VRT — New-start DC complex round-tripped Q2 gains — the market now prices zero incremental greenfield growth. Organized-market grid headroom goes negative in 2027 (-6.6 GW) and -31.1 GW in 2028. Power remains the binding constraint, not demand. Rubin's 45°C liquid cooling raises content per rack. Entry point forming if backlog holds.
FIX — Direct greenfield-start play that gave back Q2. Same grid math: US power headroom negative in 2027. If starts slip on transmission approvals, downside; if backlog holds, structural entry. Power is the chokepoint.
STRL — Texas 765-kV approvals slipping into late fall on landowner opposition — interconnect timelines slide with them. Round-tripped with the complex. If approvals clear, asymmetric upside; if not, more downside. Grid is the bottleneck.
BE — +7% pre-market, +7% intraday in a weak AI capex tape — Pelosi disclosure up to $12M moved the name +11.83%. Sentiment event, but it exposes the political tailwind for behind-the-meter gas and onsite power. Counter-trend bid is real.
GEV — New medium-voltage UPS sits between AI facilities and the grid, buffering load swings before they hit turbines. First systems ship mid-2027. New TAM beyond turbines and transformers — power-equipment cycle extending on product expansion, not just capacity orders.
FLNC — UBS upgrade to Neutral from Sell, PT $12 from $9 — an admission that the FY26 guide cut troughs the near term. FY26 EBITDA still negative at $(12)M. Not a bullish call; balanced r/r after the damage.
LITE — Goldman's internal marquee AI tool is qualitative demand color for the full AI stack, optics included. Wall Street GAI workflows hitting the F500 means more datacenter traffic and more optical content. Not an order datapoint yet.
UMC — US bank raised the target on photonics: TFLN 8-inch at 20k wafers/month by 4Q27, 40k by 4Q28, PIC GM >70%, ASP ~$2.5k vs <$1k normal. Hyperlight showing possible Google progress, TeraHop's key customers are NVDA and GOOGL. Specialty foundry transformation inside a mature logic house.
XIACY — Xring O3 takes Xiaomi from assembler to silicon designer — ARM CPU/GPU on TSMC 3nm plus a companion AI PC. CXMT as key LPDDR6 supplier links it to China's memory localization push. The clearest evidence yet that Chinese OEMs won't stay dependent on merchant SoCs.
TCEHY — Strategic participation in XPeng's $900M+ humanoid robotics round. Embodied AI underwriting is now a Big Tech capital allocation theme. Not an earnings driver, but the direction is unmistakable.
IFNNY — C2i acquisition buys software-defined multiphase controllers and smart power stages — the missing control layer in AI server power delivery. Content-per-rack keeps rising and power semis consolidation continues. One of the least disputed sub-narratives in the AI hardware chain.
ALM — Tungsten chatter building, cited in a US-listed tungsten tracker. Critical metals scarcity feeds carbide, tooling, and semis supply chains. Monitor signal, not a confirmed order book.
KMT — Same tungsten chatter, US-listed carbide tooling exposure. Early and thematic; no confirmed supply-chain datapoint yet. Direction consistent with scarcity. Watch, don't chase.
BMM — Listed in the US tungsten tracker as chatter builds. Nothing confirmed. Monitor only.
ERIC — First North American LTM field test done, handover interruption down 25%. 5G Advanced moving from lab to field. Weak commercial-paving signal, not an order catalyst.
T — Same LTM datapoint — network quality positive, but not a material capex signal. 5G Advanced remains a roadmap story, not a spending story.
(MDB, IOT, ZM, DELL, BOX, CIEN, AMBA, SNOW, TYL, ASAN, FTNT, VST, WEAV, XYL, ROC, RDDT, ATEX, ROKU, YOU, TLN, MAX, XMTR, ALKT, TSLA, NTES, RSKD, ROK, ABNB — no fresh 3p prints in this window. Clean tapes.)