Good morning.
PLTR GAPPING +19% — THE CLEANEST GROWTH PRINT OF THE QUARTER. US Commercial +149%, revenue $1.94B, guide raised $496M. ALAB +10% post-market: EPS 80c vs 69e, clean beat on every line, PCIe switch thesis with AWS deeper in the story. AMD beat-and-raised too, but capex tripling to $808M triggers the ROI anxiety reflex at 200x forward. THE TAPE WANTS MONETIZERS, NOT BUILDERS.
The China optical transceiver ban draft is the day's overhang — new Chinese modules blocked reshuffles the 1.6T/3.2T supplier list from 2027, but AAOI/COHR/LITE lack the capacity. Bottleneck migrates to InP lasers, not modules. AXTI bearish if Beijing retaliates on InP exports.
Power supercycle confirmed but timelines stretch: POWL booked $934M at 3.0x book-to-bill, only 54% converts in 12 months. CAT +8%, AMRC +30%, GEV orders into 2033. Google-Anthropic's $200B arrangement ($44B of lease guarantees) reframes funding — hyperscalers build off-balance-sheet, CoreWeave adds 360MW in Indonesia.
Asia trades headlines, not macro: China optics risk plus the Indonesia expansion. No rate shock overnight.
We'll hit up AMD, PLTR, and ALAB first, then get to the optics complex — LITE, COHR, AAOI, AXTI — before we sweep the power group (POWL, CAT, AMRC, GEV).
SNAP DELIVERED THE CLEANEST BEAT IN SOCIAL THIS CYCLE — REV +19%, EBITDA +30% VS CONSENSUS — AND THE STREET BARELY BUDGED. PTs cluster $5.25-$8. BMO sits alone at $15. That disconnect IS the trade.
Q2 was a sweep: revenue $1.60B (+19% YoY, +4% vs cons, above the upper end of guidance), adjusted EBITDA $250M (+30% vs cons), gross margin 58.2%, FCF $121M. EIGHTH STRAIGHT POSITIVE FCF QUARTER. April restructuring showed up where it matters — total adjusted costs up just 4% YoY.
The rate of change is the story, not the level:
Q3 guide: revenue midpoint $1.72B (+1% vs cons), EBITDA $325M (in line to -1%). Solid, but not the beat-by-a-mile the quarter implied. World Cup spending normalizes. AI/ML infrastructure costs ramp. The "good quarter, softer setup" pattern.
BULL: The inflections are real and multiple. Ad revenue re-accelerated 6 points sequentially — NA large-advertiser spend is back in the exact channel that broke first in the 2024-25 downturn. The non-ad business at +85% and now ~20% of the mix structurally reduces ad cyclicality. Management redefined the playbook: FCF PER SHARE is now the primary financial objective, a multi-year dilution management program targets a STABLE share count in 2027, and they expect sustained net income beginning 2027. That's a company that stopped story-telling and started compounding. At 6.5x 2027 EBITDA, BMO says the r/r is simply too good to pass up.
BEAR: The guide is the tell. +1% revenue vs consensus into H2 despite the World Cup — that's deceleration, not acceleration. Event spend inflated the beat; that volume doesn't repeat. NA DAU down ~7% YoY with flat QoQ isn't a turn, it's a pause. Infrastructure costs ramp exactly when margin expansion needs to show. And Specs is a distraction — high price point, Meta funding the category leader, hardware has never been Snap's friend. Rosenblatt LOWERED its PT to $6 despite the beat. Skeptics read the quarter as good and the setup as worse.
"We will maintain our Neutral rating until we gain confidence on the ad revenue trajectory in the second half of 2026 and fiscal year 2027." — Cantor Fitzgerald
BMO "questioned Snap's ability to win in smart glasses over the long term given high price point and ramping competition from well-funded competitors." — BMO Capital
NEW: The FCF/share mandate and dilution management program. Management explicitly naming sustained net income for 2027. Ad rev growth doubling from 3% to 9%. Other revenue crossing 20% of the mix. Sept 16 Specs launch event date. Q3 guide details.
ALREADY KNOWN: NA DAU stabilization trend. Snapchat+ subscription traction. AI-driven Smart Campaigns pulling SMB spend. Cost discipline from the April restructuring.
The FCF/share shift is the most underappreciated line in the print. Snap linking valuation to cash generation per share — WITH a share count stabilization program — changes the equity narrative from "will they ever make money" to "how fast can they compound FCF." At $5-6, that's a regime change, not a quarter.
SNAP IS A NARROW LEADING INDICATOR FOR DIGITAL AD SPEND. If NA large-advertiser spend returns here, it returns across the group. Watch the read-through to CTV/premium video (World Cup premium volume = event dollars back in the market) and to the AI-ad-platform cohort (Smart Campaigns driving SMB efficiency = positive for META/PINS). The +85% subscription growth validates the ads-plus-subscription hybrid model across social. And Specs puts Snap squarely in the Meta Ray-Ban smart glasses narrative — Sept 16 is a category event, not just a single-name catalyst.
Bottom line: best quarter in two years, a guide that says "ok, not amazing," and a stock that prices in none of the improvement. The bull case is real; the burden of proof shifts to Q3. Watch NA DAU and ad growth post-World Cup. Hold those and this is a $10 stock with a $5 handle. Slip, and BMO's $15 stays a mirage.
Blowout quarter, split Street. Nobody disputes the fundamentals — the fight is the multiple. THE LOWEST PT ON THE STREET IS CANTOR'S $156, STILL 24% ABOVE THE $125.65 TAPE. Bulls cluster at $200-255. Watch the r/r, not the level.
REVENUE +93%, accelerating 8 points Q/Q, on a $1.94B print that beat consensus by 8%. Adjusted EPS $0.41 vs $0.34 est. Truist notes consensus was already above the high end of guidance — and they still blew through it. The engine: U.S. COMMERCIAL +149% (vs +133% last quarter, above the 135-140% guide). Piper flags the $169M sequential add as the largest quarterly increase of any segment in company history. Bookings are the real tell: U.S. COMMERCIAL BOOKINGS +153% vs +45% the prior quarter. That's a step-change, not a blip.
Quality metrics hold up: NDR 157%, GM 84%, Rule of 155 (up from 127), 220 deals over $1M, 98 over $5M, 73 over $10M. FY26 REVENUE GUIDE RAISED TO 82% FROM 71% — call it >$8B. The one soft spot: international +33%. This is a U.S.-led story.
Two camps, same facts. The bulls: BofA $255, Piper $230, Truist $223, UBS $220, DA Davidson $200, William Blair Outperform (no number). The skeptics: Cantor $156 Neutral, Benchmark Hold. Collectively they call it the same quarter — the split is what it's worth.
That split lives in the multiple math: UBS pays 44x 2027 FCF to get to $220. Cantor needs 54x 2027 FCF just to reach $156. The bulls aren't just paying more — they're modeling more.
Bull — This is the first hard proof that AI spend expands within customers, not just pulls forward. Deceleration? No: U.S. commercial accelerated 133% → 149%, bookings went 45% → 153%, NDR 157%, and Piper sees Commercial overtaking Government as soon as Q3. PEG at 0.46 neutralizes the multiple if growth holds. UBS says there's no evidence competition is denting share. Truist's kicker: the forward-deployed engineer model bears mocked for years is now the blueprint, not the bug.
Bear — The quarter was great. That's the problem. AT 174X EARNINGS (30x 2027 revenue per Cantor), the market already pays for perfection. Cantor flags the OpEx ramp in 3Q and 2H26 — partnerships and GTM spend — as incremental execution risk hitting right after peak acceleration. Benchmark's line: the performance is in the price. And the one gap in the story — international at +33% — says this is a U.S./sovereign AI story, not a global platform story yet.
"Palantir remains its highest-conviction idea following another beat-and-raise quarter." — Truist
"Palantir's high-touch deployment model using forward-deployed engineers, which previously drew criticism, increasingly appears to be the blueprint for enterprise AI adoption." — Truist
"This introduces incremental execution risk against a backdrop of significant accelerating performance." — Cantor
"The market has already priced in the company's performance with its current valuation." — Benchmark
The AI software trade just got its hardest data point. U.S. commercial +149%, bookings +153%, NDR 157% — that validates the whole complex (SNOW, NOW, CRWD) and the sovereign AI theme Truist is flagging. But it cuts both ways: when a stock delivering flawlessly at 30x forward revenue still draws Holds, the bar is high. Watch the relative value trade — if PLTR deserves 30x 2027 revenue, names growing half as fast at 15x look either very cheap, or the whole complex is pricing perfection. Your book, your risk.
Two PT cuts, one reiteration — and the stock's still up 48% YTD. That's the whole setup in one sentence. ON beat, raised, and guided GM to 40-42%, yet Stifel slashed to $90 from $107 (Hold) and Needham trimmed to $116 from $130 (Buy). Cantor holds Neutral at $110. Same quarter, three different frames — the disagreement is about what multiple the AI story deserves, not whether the recovery is real.
Clean print. Q2 revenue $1.60B, +6% QoQ, +9% YoY — above the guide midpoint and 1.2% over Stifel's number. EPS $0.74 vs $0.71 consensus. GM 39.3%, FOURTH STRAIGHT QUARTER OF MARGIN EXPANSION. Q3 guide: $1.70B (+6% QoQ) with GM guided to 40-42%. That's the trajectory the bulls want to see.
AI is the engine. Management now expects data center revenue to MORE THAN DOUBLE in 2026 — an upgrade from prior "double" language. Needham says lead times are stretching INTO 2027 AND 2028, and ON is pushing through a SECOND ROUND of price increases to offset input costs. Utilization jumped from 77% to 83%. That's real operating momentum.
Stifel's the skeptic with receipts. CY26 revenue growth lands around 9.3% — BELOW the analog industry's 11-15% (per WSTS/SIA) and at the low end of Stifel's coverage. Their $90 PT works out to 19.8x CY27 earnings. The Hold is conditional: they want to see gross margin improvement sustained before paying up. Fair. The 40-42% Q3 guide is a step, not a trend.
Needham's the bull with the same facts, different frame. New PT at 20x 2028 EPS of $5.80 — earnings compounding ~4x faster than revenue. Pricing power, utilization recovery, GM expansion every quarter through 2026. That's the operating leverage story, and it's coherent.
"Demand signals continue to improve, lead times are extending, and visibility is extending into 2027 and 2028."
Cantor's the agnostic: $110 PT on $6 CY28 EPS, but openly questioning the size and growth trajectory of the data center business and flagging auto/industrial dynamics vs peers. Their line — "better upside opportunities elsewhere" — is the shrug of a firm that isn't fighting the tape but won't chase it either. The tell: EVERYONE is pointing to the September 16 analyst day in New York as the moment the data center TAM question gets answered.
The fundamentals improved. The stock's 48% YTD run means the easy money's made. PTs cluster in the $90-116 zone — a 12-45% spread from the current $80 — and that dispersion is the Street's honest confusion about margin durability. The sector-wide multiple compression (Needham cited it explicitly) isn't ON's fault, but at ~57x earnings, the stock prices in flawless execution. The 9.3% grower narrative doesn't justify that multiple. The AI doubler does. Which one wins? That's the September 16 question. Until then, the tape's the judge — and next week's tape starts with three Wall Street PT cuts hitting screens this morning.
Verdict: AEIS just ripped the tape on a massive beat-and-raise — EPS ~25% OVER consensus, guide up across every segment, and a fresh Buy upgrade at $410. Three calls in this morning: Seaport to Buy ($410 PT), Cantor reiterating Overweight ($450 PT — highest on the Street), TD Cowen holding at Hold with a grudging PT bump to $370. The bull case is loud. The question is whether the 111% one-year run has already paid for it.
Q2 FY26: adjusted EPS $2.74 on $574M revenue — +30% Y/Y, +12% Q/Q. RECORD REVENUE. RECORD GROSS MARGIN. RECORD OPERATING INCOME. Semis and data center both cleared sell-side projections.
But the guide is the real story. Management pulled FY26 revenue growth up to LOW-TO-MID 30% from low-to-mid 20%. Data center now guided >50% growth (was >35%). Semi H2 revenue guided to NEARLY 50% Y/Y. TD Cowen flags the guide implies EPS of ~$11+ vs the $9.42 Street consensus before the print — a ~17% gap. Seaport raised H2 estimates by ~35% and 2027 by ~30%. Nine analysts have revised numbers up post-print.
(The margin story matters too — gross margins expanding on mix/volume, Thailand quals accelerating on customer pull. This isn't just revenue leverage; it's operating quality.)
Bull: Power delivery is the choke-point trade in AI data centers, and AEIS is winning sockets. Cantor's stretch $20 EPS goal for 2028 intact with upside bias. At ~$341 with PEG of 0.4, the growth-adjusted multiple looks cheap — and the 2027 guide of >20% growth gives the setup a second act. Operational execution on concurrent throughput and capacity expansion has been clean — no supply chain excuses, just delivery.
Bear: Three analysts, three PTs, one Hold. TD Cowen sits at $370 — essentially spot to the tape — and won't go to Buy at this multiple. The stock is up 42% YTD, up 111% in a year. The guide raise is already in the price. At $341, you're paying for 2027 execution, not the Q2 print.
Cantor's note captures the magnitude of the demand shift — this is a re-rating event, not a quarterly beat:
"Data center revenues are tracking better than expected in the second half of the year, supported by market strength, new program wins, and successful supply chain navigation by customers... Management increased its guidance for 2026 data center revenues to grow more than 50%, compared to a prior forecast of growth in the mid-30% range."
The r/r here is clean: bulls point to the $450 PT and a $20 2028 EPS stretch case. Bears point to a Hold-rated PT $30 above the tape. The middle ground — and probably the right one — is that the fundamental acceleration is real, but the multiple expansion is largely done. Watch the semi guide (nearly 50% H2 growth) as the next marginal catalyst. If that holds, the Street's revised estimates still lag reality.
The market still prices AMRC like a slow-growth ESCO. The backlog says data center power play. Cantor says $44, stock says $22.73. One of them is wrong.
Cantor re-ups Overweight with a $44 PT after the print. Oppenheimer parks at Perform. That divergence IS the debate — and it's a good one.
Q2: revenue $515.5M vs $465M consensus (+9% YoY), EPS $0.20 in line, adjusted EBITDA $62.8M vs $61M. Solid, not spectacular. The headline is the book of business, not the quarter.
RECORD $1.8B IN NEW AWARDS — INCLUDING $1.2B IN BEHIND-THE-METER DATA CENTER POWER PROJECTS, WITH CONTRACTS FROM HYPERSCALERS AND NEOCLOUDS. Total backlog $6.7B, +32% YoY, +28% QoQ. That's a serious rate of change for a name this size.
The data center breakdown matters:
Bull (Cantor, Overweight, $44): The pivot from efficiency contractor to behind-the-meter power provider for AI data centers is the re-rating catalyst. $1.2B in a single quarter validates it. $6.7B backlog derisks it. At $22.73, you're paying roughly 15-18x forward non-GAAP EPS ($1.25-1.57 per Oppenheimer's estimates) for 32% backlog growth. The r/r skews hard to the upside.
Bear (Oppenheimer, Perform): Backlog isn't revenue, and Phase 1 doesn't print until 2028. This is a $1.2B market cap company with a 39x trailing P/E, flagged cash burn, and debt on the balance sheet. Oppenheimer nudged estimates in line with guidance — FY26 $2,093M rev / $258M EBITDA / $1.25 EPS — but won't step in front of the story until conversion shows up in cash flow.
"Management indicated that the value of the current awarded data-center scope could increase to as much as $2 billion as project phases and specifications are finalized."
That sentence carries the bull thesis. The $1.5B awarded DC book scales to $2B, the 5+ pipeline opportunities keep feeding the funnel, and the hyperscaler demand backdrop hasn't cracked. The question isn't demand — it's the balance sheet.
If AMRC self-funds the buildout, this re-rates hard. If they need equity, $22.73 might be the floor, not the entry. Watch the cash flow trajectory and the debt math. That's the tell.
THE SIGNAL IS IN THE GUIDANCE, NOT THE PRINT. Q2 came in a hair above — $0.20 vs $0.19 est, $204M vs $202M — but that's noise. The real number is the FY26 CLOUD GROWTH GUIDE CUT TO 18% FROM 25%. That's a 700bps downgrade in one quarter. DA Davidson cuts PT to $12 (from $15), BMO nudges $14→$15 on e-invoicing hope. Stock sits ~$11.66, down 60% off the year-ago mark. Nobody's pounding the table.
Two firms, same read: customers are re-evaluating IT priorities and cloud migrations are pushing right. BMO calls it "elongated deal cycles and uneven cloud migration." DA Davidson calls it "slower than expected cloud conversions." The core tax business is stabilizing, but the cloud engine — the growth story that was supposed to re-rate this name — is sputtering.
Management reduced its fiscal 2026 cloud growth guidance to 18% from 25%. The company cited customers re-evaluating IT priorities and cloud migrations moving forward.
Management's counter: Q4 re-acceleration on ramping e-invoicing revenue. Fine. But BMO flags that outlook as risky given the same back-office digital transformation headwinds. If the macro is pushing cloud right, why would e-invoicing be immune?
UBS reiterates Buy with $132 PT into Thursday’s print — and the setup actually looks good. The firm’s checks show AI infrastructure demand ACCELERATED over the past three months across hyperscalers and neoclouds, which supports the $9.6B networking revenue forecast (+26% y/y, accelerating from 25% despite a four-point tougher comp).
The order momentum is the real story here. Last quarter's 35% product order growth was the strongest in over a decade — and that includes 19% growth excluding hyperscalers, which tells you this isn't just AI capex concentration. UBS sees scale-across pluggables and systems upside following $1B+ in Acacia orders last quarter. They're modeling a modest decel to 29% order growth in July, but that feels conservative given the demand commentary.
"Industry checks and commentary from hyperscalers and neoclouds indicate demand for AI infrastructure accelerated over the past three months."
The margin math is the only real capper here. Component costs keep gross margin pinned around 66%, so UBS lands at $1.19 EPS vs consensus $1.17 — not a blowout, but enough to matter. At $119.55 with a 40x P/E, the premium is already in the tape. The question isn't whether AI networking is real — it's whether Cisco can keep taking share from Arista in the scale-out layer. So far, the Acacia pluggables attach is doing exactly that.
The take: This is a "beat and raise is probable, guidance is the swing factor" setup. Stock's up 73% in a year, so a lot of good news is priced in — but if orders accelerate again, the 40x multiple starts looking justified rather than rich.
DEMAND IS OUTRUNNING THE BUILD. ASML can't make EUV machines fast enough, and the market is finally paying for that constraint. Stock +136% Y/Y, +54% YTD, and the quarterly print just poured fuel on it: Q2 EPS $8.69 vs $7.95 bogey, revenue $10.67B vs $10.29B. FY26 guide raised to €43-45B — that's 35% GROWTH.
The order book is the story, not the quarter. 2027 Low NA EUV orders are nearly fully covered. A "significant" 2028 book is already placed. Management isn't waiting for more POAs — they're committing to a 30% capacity increase for 2027 and already evaluating a second expansion for 2028. That's not a cycle bet, that's customers' multiyear agreements forcing their hand.
The broker herd is following. Freedom Broker lifted PT to $2,100 from $1,650. RBC to $2,100. Deutsche Bank to EUR2,150. All Buy/Outperform. The collective thesis: visibility extends years, margins are inflecting (Deutsche sees 56% GM in 2H26), and Intel's High NA insertion on 18A turns the next platform into a capacity option, not a science project.
"The order book behind the guidance is more durable than the visible results."
Not all roses. Management says they're capturing a bigger share of imaging/overlay value, but long lead times delay the financial benefit. And InvestingPro flags the equity as overvalued against fair value. Fine. But name another way to own the only shop that can print these machines. The r/r still leans long.
PTs stacking up ahead of the Aug 6 print: Canaccord to $295, MS to $300, BofA at $305 (top pick), Cantor at $327, Benchmark at $330. Stock at $273.60, a hair off the $278.76 52-week high, 101% YTD. Options market pricing a 13% move. This is no longer a question of whether DDOG beats — it's whether the re-acceleration is durable enough to justify the multiple.
Canaccord frames the bar well: high-quality print = non-AI growth holding mid-20%+, AI revenue broadening beyond the top account, NRR staying low-120% or improving, and a Q3 guide near 30%+ that forces a meaningful FY26 raise. They see 30% FY26 growth as reasonable by year-end. The stock trades ~22x FY26 revenue midpoint and 18x EV/rev FY27 — rich, but 80% gross margins and 38 upward revisions give the bulls cover.
"The market should focus less on the absolute second-quarter beat than whether the breadth, forward guidance, and margin profile prove the acceleration is durable."
Bull: AI-native observability is the wedge (Cantor's point), and DDOG is the highest-quality pure-play for enterprise data/AI spend. Durability argument steel-manned: if Q3 guide lands at 30%+, FY26 estimates go up, not the multiple down.
Bear: The easy money is made. 101% YTD means a "solid but not spectacular" print — think Q3 guide at 27-28% — hits the stock hard even if Q2 beats. At 22x forward revenue, there is no valuation cushion. This is a high-beta momentum name trading on rate of change, and the rate of change has to keep accelerating.
Setup is binary, hence the 13% options move. Good print = $300 handles. Bad print = multiple compression in a stock nobody wants to catch. If you're not long into this, you're either early to a post-earnings dip or you're paying up for certainty that doesn't exist.
Needham starts SKHY at BUY, PT $200 vs $142.72 — roughly 6x CY28 EPS. That's the call: pay up for a normalized AI-memory year, not the current blowoff. LTM REVENUE +145%, GROSS MARGIN 70% — the inflection is real, and the Street clearly thinks the durability is improving.
The differentiator is structure. SKHY holds #1 in HBM and DRAM, #2 in NAND (2025, IDC). Five-year LTAs with ~10 leading customers as of Q2 2026 — that's the needle-mover. Long-dated agreements mean less spot-price beta, more predictable earnings, fewer of the boom-bust cycles that used to make memory a beta trade on DRAM spot. William Blair notes prices have TRIPLED in six months on supply restraint. Wolfe sees 2027 EPS at ~$32.19 (KRW 460,505). Rosenblatt and Stifel both initiated Buy, Morgan Stanley upgraded post-"leverage washout." The initiation wave is unanimous enough to feel crowded.
"SK Hynix is positioned as a full-stack memory enabler of artificial intelligence infrastructure."
LTAs lock in volumes, not necessarily peak prices. If HBM supply catches demand by 2027, the multi-year visibility cuts both ways. But at 6x 2028 EPS, Needham's paying a reasonable multiple on a normalized year, not peak multiples on peak earnings. That r/r is fine for a leader with pricing power and contracted revenue.
New spin, new call. Oppenheimer starts ADIG at Outperform, $30 PT. That's 9.4x FY27 EBITDA — a discount to WESCO and broadline distribution comps, justified by smaller scale and sub-7% margins.
The thesis is a margin story disguised as a growth story. Commercial retrofit, tech refresh cycles, connected-device adoption → 4-6% organic. The real kicker: $80M+ RUN-RATE SAVINGS BY END-2027. That's the bridge from 6.2% EBITDA margins (FY26 est) to 8%+ by 2030, with FCF of $159M/$183M crushing leverage from ~3x to sub-2x.
"The $30 price target implies 9.4 times fiscal 2027 EBITDA, a discount to WESCO and distribution peers due to ADI Global Distribution's smaller scale and lower margins."
One voice, day one. Distribution is 1 ADIG per 2 Resideo shares — float's heavy, no tape to validate the bid, no supply dynamics priced. The near-term bogey isn't the EBITDA ramp; it's the spin-overhang and day-one VWAP burn-off. R/r is fine for the patient 2-3 year hook, but let the first few prints tell you where the real bid sits before leaning in.
Piper opens CRWV at OW, $151 PT — that's ~75% above the $85.76 print (stock already +27% on the week). The street keeps pricing this like a leveraged lease roll-up; Piper says the MSA structure + financing approach kills most of the financing risk the market still flags.
The differentiation angle is engineering-led: lower cost per token, multi-cloud optionality, 130% LTM revenue growth. Zero profitability expected this year — and nobody cares if the megawatt math holds.
"CoreWeave's master service agreements and financing approach removes some of the financing risk that remains misunderstood by the market."
BofA at $140, Wolfe at $150, Rosenblatt at $250 (that one's got Meta entering the cloud business baked in — take with salt). The cluster says one thing: CRWV has become the liquid expression of the AI infrastructure buildout, and the tape agrees.
Meanwhile the Nvidia/OpenAI chatter — ~$250B in financing guarantees for a 10GW Ohio project, $500B+ total cost — keeps feeding the narrative. Piper's Megawatt Model reads conservative over the medium term. That's the bull case in one line: capital is the constraint for everyone else, not demand for CoreWeave.
Verdict: the CoreWeave deal is the whole story, and the Street is finally pricing it. Needham nearly doubled its target ($23 from $14, Buy) with Citizens right behind ($21 from $16). The tape already front-ran them — BLZE +29% on the week, +271% in six months — but the PT hikes confirm this is a re-rating, not a dead-cat bounce.
The quarter delivered: REVENUE $42.7M VS $39.9M CONSENSUS (+18% YOY), EPS $0.08 vs $0.02 expected. Management lifted FY revenue guidance by $10.5M (6.5% at midpoint). The $335M, five-year CoreWeave deal turns the "cheap cloud storage" pitch into "AI infrastructure with contractual minimums." That's why B2 Cloud is guided to >40% growth in 2027 — and why capex jumps to 55-65% of revenue. They're buying equipment to feed the backlog.
Backblaze raised its full-year revenue guidance by $10.5 million, or 6.5% at the midpoint, based on business execution, a price increase, second-quarter performance, and a previously announced $335 million five-year deal with CoreWeave.
Still not profitable on a GAAP basis (-$0.39 LTM), but the Street has them at $0.13 this year. The CoreWeave minimums hit mid-2027 — that's the visibility edge. The risk is the stock price has already banked a lot of it. +271% in six months is a crowded trade if the next guide disappoints.
We like the story. We'd rather buy a dip near the old $14 target than chase $15.59. But for momentum PMs, this is exactly the kind of narrative-driven name that keeps working until it doesn't.
Ladenburg cuts WEC to Neutral from Buy, PT $110.50. Stock trades $109.28 — basically AT the PT and within spitting distance of the $102.95 52-week low. Forget the fundamentals debate for a second: this downgrade is pure politics.
Bipartisan opposition in Wisconsin to data center development. BOTH gubernatorial candidates (D and R) back killing the sales/use tax exemption for data centers. The Dem Socialist candidate wants a one-year moratorium on new builds. That's the rate-base growth engine getting a regulatory wrench thrown in it.
"Both candidates advocate for the termination of the sales and use tax exemption for data centers."
Bull case intact on the surface: 22 straight years of dividend raises, 3.49% YIELD, and Q2 EPS of $0.91 BEAT the $0.83 consensus. But revenue missed ($2.06B vs $2.11B) and the platform flags WEC as OVERVALUED vs fair value. When the load-growth narrative hits political headwinds AND the multiple already embeds perfection, a defensive yield doesn't save you. Not at $109.
The headline is brutal: Q2 revenue MISSED, PTs slashed across three shops — but the guide went UP and the loss narrowed. That's the "ugly print, better trajectory" setup. The question is whether you trust the raise or the miss.
Revenue $382.5M (+21.6% YoY), MISSED HCW's $390.9M projection. Diagnostics $289.3M (+19.6%) — oncology volume +31% ripping, hereditary growth at 5% dragging. Data & Apps $93.2M (+28%), Insights +36%. Gross margin 64%, TTM revenue growth 50%.
The EPS beat is real but read the footnote: GAAP net income of $5.6M comes from $102.8M of unrealized gains on marketable equity securities, not operations. Non-GAAP loss of -$0.04/sh beat the -$0.08 consensus. Clean beat on ops, flattered print on the bottom line.
Management RAISED FY26 revenue guidance to $1.595-1.605B (from $1.590-1.600B), ~25% growth, with ~$65M adj. EBITDA. Not a barn-burner raise, but directionally correct after Ambry's first full quarter.
PT moves tell the story: HCW to $56 (from $64, cutting EV/Revenue from 7.0x to 6.0x), BTIG to $70, Stifel to $50 and Hold. The $50-$70 cluster on a $46 stock says "buy the dip but not aggressively." Consensus range $35-$100 — wide enough to drive a truck through.
Bull case: oncology volume acceleration + xT CDx approval reduces pricing risk, and the Personalis acquisition extends the sequencing story. Stifel flagged the CDx approval as the potential unlock for a future price increase — that's the margin catalyst nobody's modeling yet.
Bear case: revenue missed, hereditary is sluggish, and trimming multiples into a deceleration is cope. At 6x EV/Revenue on 21.6% growth, this isn't cheap. HCW's multiple cut is the tell.
Net: plenty of strategic arrows, but the market needs Diagnostics re-acceleration before this works. I'd rather be a scale-in at $46 than a chaser on the next gap up.
BMO's the loudest bullish voice this morning — PT to $46 from $42, Outperform, and the stock's already sitting at $41.15, basically pinned to the $41.94 52-week high. That's ~12% implied upside, but the r/r skew matters more than the headline: BMO sees multiple paths to FY26 upside, not just one clean narrative.
The Q2 print lands August 12, and BMO's getting a read-through from Visa/Mastercard cross-border commentary — both called out strong international e-commerce activity. That's the setup. The real debate is H2: Managed Markets v2.0 ramp potential, easing FX tailwinds, and moderating promo spend. BMO's framing is that guidance embeds conservative same-store sales assumptions, while tariff refunds and duty drawback sit as unmodeled upside.
The broader tape is turning constructive too — Truist upgraded to Buy, Piper and Benchmark remain long, and the Passport acquisition closed with a $500M buyback layered on top. Shareholder returns and a growth story. Rare combo at this valuation.
"The combination of conservative same-store sales assumptions in guidance and multiple upside opportunities including tariff refunds and duty drawback is attractive."
Not a slam dunk — the stock's already re-rated toward the high end. But with revenue growth at 28.6% LTM and multiple potential accelerators, this is a stock that can grind higher on execution rather than multiple expansion. Keep it on the watchlist into the print.
Needham to $88 from $72, Buy maintained. Stock at $80.23 — the new PT is barely 10% above the tape after a 335% YTD run. Momentum-confirming raise, and the numbers justify it.
Q2 revenue $294.8M (+15% QoQ, +24% YoY) slipped just under the $300M guidance midpoint. Cause: mass flow controller component shortages pushed shipments a week into Q3. Pure timing — the print crushed the $268.2M Street bogey. What matters: NON-GAAP GM HIT 14.1%, ABOVE THE 13-14% GUIDE, up from 12% LTM. EPS $0.34 vs $0.31 consensus.
Q3 guide: $330M midpoint (+12% sequential), GM ~15% on mix and higher internal content, EPS $0.40-0.50. CY26 revenue +30%+ YoY implied, Q4 +12% sequentially. Capacity: $2B installed, $3B in reach. Operating-leverage story still inflecting.
The disappointment flag: Q3 guidance only marginally topped consensus. Investors got a beat-and-raise quarter and wanted a blowout — the bogey was perfection after +298% in a year. But the margin cadence (12% LTM → 14.1% → ~15% guided) is compounding, and deposition/etch demand hasn't rolled over. Momentum may be late-innings; the P&L ramp is early.
Evercore cuts PT 21% to $95, keeps Outperform — this is a multiple cut, not a thesis change. Stock sits at $66.39, down 45% YTD, and the new target still implies 43% upside. Classic "good quarter, bad tape" setup: Q2 beat across the board, but the market is staring at the Partners drag and a reacceleration path that keeps slipping.
The headline number is ADJ FCF OF $61M VS STREET'S $36M EST — a 70% beat. EPS $1.59 vs $1.16 est, revenue $563M vs $556M. Bookings +12%, revenue +15%, driven by AI product demand. FY26 guidance reiterated: low-to-mid teens revenue growth, low-teens bookings, high-teens FCF margin. Q3 guide consistent with Street's 13.5% growth bogey. No fireworks, but no bogeys either.
The margin angle is the real signal here. Base1 (their proprietary LLM) lifts Base44 Non-GAAP gross margin to ~60% IN 2H, from NEAR-ZERO entering FY26. That's a step-change. But no Base44 ARR disclosed — "growth trajectory similar to prior disclosure" is management-speak for "we're not giving you the number." Watch that disclosure like a hawk.
Target multiple goes to 10x FY27 EV/FCF from wherever $120 implied. Evercore's verdict:
"Estimate revisions are modest ... cautious view on the reacceleration path given Partners drag."
So the bull case: 21% FCF yield, AI demand real, margins inflecting, profitability this year. The bear case: Partners is the albatross, and 10x on a decelerating top line isn't a value multiple — it's a value trap until reacceleration actually prints. Not buying the dip yet. Need proof, not promises.
Needham starts VSH at Buy, $45 PT — and the whole thesis is about the margin inflection, not the cyclical snapback. Stock already +146% YTD, so this is a "we see the second act" call, not a "catch the bottom" call. Book-to-bill ACCELERATED TO 1.34, backlog +42% YOY TO $1.59B, channel inventory still bleeding off — the demand picture is genuinely tight.
The bigger story is structural. Newport is at gross profit breakeven, pricing is firming, and management guides to 45%+ contribution margins on new wins. Against a historically horrid ~20% gross margin profile, that's a multi-year re-rating if it delivers. Raymond James is on the same page with Outperform/$40, both leaning on the "Vishay 3.0" repositioning into larger, growth-oriented markets.
"Entering a multi-year earnings recovery as a broad cyclical inflection converges with improving mix, substantial operating leverage and a more durable secular growth profile."
Not sure we can read much into the defense certification and new product launches for the near-term numbers — those are housekeeping. The r/r here is simple: 146% YTD means the easy money is made, but Needham's $45 PT implies the market still isn't paying for the 31-33% gross margin target. If the operating leverage shows up in the next two prints, the stock goes higher. If the book-to-bill rolls over, this thing gives back a chunk of the move fast. Cyclical recovery + structural margin story is a solid combo; just respect the tape.
Beat-and-raise lands with two thumbs up. Cantor doubles down at Overweight with a $25 PT (54% upside) after the FQ1 print, and Morgan Stanley upgrades to Overweight — PT trimmed to $23 but that's just optics against a $24 prior. The combo tells you the street sees this as a credible inflection, not a one-quarter mirage.
Numbers check out: FY27 revenue guide ¥465-473B, adjusted EBITDA ¥149-155B. MS pencils FY27 EBITDA ABOVE company guidance and pegs a 3-year CAGR of ~19% for both revenue and contribution profit. That's not a hopium pedal — that's visibility.
The one wrinkle: InvestingPro flags PAYP trading above its own Fair Value. Pays for the multiple only if execution compounds.
Cantor views PayPay's FQ1 beat-and-raise as a credible first step.
Two houses, one thesis — the guide-up is real, and the market's still pricing in skepticism. At $16.27 vs a $23-25 PT cluster, the r/r skews fine for a name with profitable LTM and an accelerating narrative.
Q2 was a miss, Q3 looks worse, and the stock's already paid the price. Freedom Broker cuts PT to $185 from $225 (keeps Buy) after GAAP EPS printed -$1.36 vs -$0.44 consensus, adjusted -$0.40 vs +$0.12. Revenue $1.22B, -14% QoQ, roughly 5% under the $1.29B Street bogey. Shares -12.8% on the week, -54% YoY.
The forward look is the real problem. Transaction revenue run-rates to ~$460M for Q3 — 39% BELOW prior consensus. Subscription & services midpoint of $500-580M sits ~15% light. Crypto tape is the culprit, pure and simple.
"Transaction revenue of $130 million generated through July 26 run-rates to approximately $460 million for the full quarter, about 39% below prior consensus forecasts."
There's a bull case hiding in the wreckage. MARKET SHARE HIT AN ALL-TIME HIGH OF 10.3% (from 9.1% in Q1) despite spot volumes falling 25% QoQ. And management is cutting: 2026 adjusted opex guide down to $4.20-4.45B, Q3 midpoint ~4% below consensus, with a 14% headcount reduction in the bag.
The rest of the Street reads it the same way — Mizuho $155, Baird $130, Needham $177, Benchmark $230. Wide PT dispersion, one theme: transaction revenue is hostage to the crypto tape, cost control is the only variable they own.
Rosenblatt starts LSCC Buy with a $175 PT (stock at $127.18) — and the call is a beat-and-raise into tonight's Q2 print. They see AI server rack adoption, broadening industrial recovery, and AMI accretion driving the quarter. Not a lonely call (consensus is Buy, targets $105-$175), but the PT lands at the very top of the range.
The bull math is clean: server units = 35-40% of revenue, AI ~25%. Direct AI server beneficiary with an FPGA twist. Rosenblatt models FY2028 revenue/EPS of $1.4B/$3.77 vs consensus $1.14B/$3.37 — that's a 23% revenue gap in three years. Worth noting the stock's already ripped +161% over the last year, so this isn't a "cheap" pitch. It's a second-leg inflection pitch, built on 42x FY2028 next-gen EPS.
"Rosenblatt sees parallels between the current Lattice setup and conditions that preceded the company's 2019 to 2021 outperformance."
That's the right frame. LSCC spent five years in the wilderness on destocking and industrial softness — that cycle is now in the rearview. Q1 already showed it: EPS $0.41 vs $0.37 est, revenue $170.9M vs $164.89M. The 2019-2021 analog is apt, except now you get AI server attach on top.
Risks: InvestingPro flags it as overvalued on fair value, and 42x a 2028 number is a demanding multiple for an FPGA name — if the recovery stumbles, the derating is fast. But the r/r into an earnings print with this setup skews positive. No position at the open; watching the print.
Needham says the AI-driven WFE supercycle is real, and UCTT is the pure-play pickaxe. Firm raised PT to $125 from $92 — 22x CY28 non-GAAP EPS — Buy maintained. Stock sits at $89.17, so ~40% implied upside, but shares are UP 253% YTD. The move is the message: this stopped being a value story months ago.
Q2 crushed. EPS $0.70 on $644.9M sales — 10.3% above Street — and Q3 guide of $700-750M / $0.83-$1.30 blows past consensus. Management calling for DOUBLE-DIGIT Q/Q GROWTH for the next several quarters. Not a guide — a statement of intent.
The kicker is capacity. $4B/yr by mid-2027 supports a $200B WFE run rate, and they're already scoping $5B/yr in phases as customers lock LTAs. This isn't spot demand. The 10-15% operating margin flow-through stays intact as revenue compounds.
Management expects the company to grow in double-digit percentages quarter-over-quarter for the next few quarters.
Risk: 253% YTD means the front-run money is banked. But Needham rolling to CY28 tells you they see the cycle extending well past the current order book. Long-duration AI capex + rising semi content per wafer. We'd buy pullbacks, not chase this print.
Earnings in TWO DAYS and Canaccord steps up to $95 (from $78) — Buy maintained, stock at $82.40. Street consensus is Strong Buy, targets $65-$110. But this isn't a direction call. It's a quality-of-growth debate.
The bull case is usage. GITLAB'S 49% CODE PUSH GROWTH AND 38% CI PIPELINE GROWTH are the leading indicator — ramping usage forces bigger customer commits over time. JFrog's own LTM revenue growth is 25% at 77.5% gross margins. Canaccord hearing limited pushback on overage pricing, strong Curation pipeline, and whispers of a fourth frontier-model customer. That last one is the AI optionality the market keeps paying for.
Bear case: Morgan Stanley cuts to Equalweight on valuation — while raising PT to $95. "Good company, wrong price" in one move. Also, some customers are still in experimentation, not scaled production. Overage that never converts into ARR, RPO, or sustained NRR is noise dressed as revenue. Plus the late-July Artifactory zero-day — patched, but customer impact and self-hosted upgrade clarity are still open questions.
"Investors should look through a headline beat and focus on durable cloud commitments, security-led platform expansion, and the quality of any fiscal 2026 guidance raise."
That's the playbook. The beat is table stakes. The zero-day gives management a forum to pitch security-led platform expansion as the wedge into frontier-model companies. Watch the FY26 guide and RPO quality — not the headline revenue print.
The market prices this like a crypto trading book. The data center platform says otherwise. Wednesday's print is the first real look.
Benchmark holds Buy and $57 PT into Q2 — stock at $21.02, DOWN 17% YTD. That's a 2.7x gap between tape and target. Big mispricing or big trap; the call answers it.
First quarter with rental revenue from 133MW of critical IT load delivered to CoreWeave at Helios. That's AI infrastructure cash flow, not crypto volatility. $3.5B in senior secured notes just closed to fund Helios Phase II, plus a 500-acre second campus in McGregor. Management keeps saying "multi-client, multi-campus data center platform." The call should put a timeline on it.
Rosenblatt trimmed PT to $35 from $39, keeps Buy. Blame weak crypto prices and volumes — the trading book drags, the infrastructure story doesn't.
Side noise: Texas Tech naming rights (Galaxy Stadium, 2026 season), Galaxy Curator on Morpho, $500M Grove warehouse facility. All fine, none move the needle Wednesday.
R/r skewed positive into the print. CoreWeave revenue recognition + platform update > crypto tape noise.
Piper Sandler is the first adult supervision on the tape — Neutral, $224 PT vs $212.58 spot. Not a bear call. They see the same $25B ARR pathway everyone else sees, they're just unwilling to pay for it before Vineland de-risks.
The model is genuinely differentiated: asset-light "Soperator" scaling, custom rack/server design, Token Factory, end-to-end platform, and customer prepayments funding the buildout. Piper's contracted-to-active power math gets them to $25B+ ARR BY 2028. That's the bull case in one line.
"Piper Sandler's analysis of contracted to active power suggests Nebius could build an annual recurring revenue business exceeding $25 billion by 2028."
Near-term risk is specific: the Vineland datacenter. LTM REVENUE $877.9M WITH +575% GROWTH — explosive, but off a tiny base, so one facility slippage punctures estimates. The rest of the tape is all gas: $775M GPU-backed debt facility (high-credit customer, matures '30), Baird Outperform, NORTHLAND AT $410, Freedom Capital to Buy. And the Nvidia/OpenAI $250B financing guarantee chatter involving SoftBank's energy arm — if that lands, every PT on the street goes out the window.
Bull vs bear: bulls say the training-to-inference transition funnels workloads to their full-stack platform. Bears say 575% growth off $878M means every quarter is do-or-die, and Piper's right that the r/r at $212 isn't compelling until Vineland prints. Not sure we read too much into one Neutral start — but it's the first price-realistic PT on a tape that's gone full cult.
Davidson is staying patient, but this is a trust-me trade, not a momentum trade. Reiterated Buy at $40 PT (adjusted from $2 for the 1-for-20 reverse split) — implies 6-7x EV on 2027E EBITDA of $453M. That's roughly a full turn of multiple expansion from where we sit today: 5.04x on LTM EBITDA of $485M.
The core argument: 2026 is a transition year. New senior management, broken track record, expectations need resetting. Davidson explicitly wants Q3 guidance WITH results and more full-year color — they're practically begging management to hand the Street a roadmap.
STOCK DOWN 38% OVER SIX MONTHS. That's the market pricing in exactly that management chaos. Bull case: at 5x forward EBITDA, you're not paying for any recovery — downside is cushioned. Bear case: this is a value trap until the new team proves it can execute, and reverse splits + guidance delays don't scream confidence.
"The team needs to re-set expectations and start to re-build their track record."
The BNY retirement solution collab is a sensible strategic datapoint — recordkeeping plus custody/payments is a legit bundle. New SVP for broker/channel partnerships on the leave side, fine. Small moves, directionally okay.
No earnings print here. This is a "guidance is coming" setup. PMs should wait for Q3 before adding or fading — nothing in this name moves the needle until management actually speaks.
KeyBanc's the only voice in the room, but it's loud: PT to $430 from $305, Overweight maintained. That's a 41% hike — they're saying the stock's 52% YTD run isn't the trade, the re-rating is.
The Q2 print beat, but read past the headline. Non-GAAP EPS $6.35 vs $4.36 bogey, revenue $1.56B vs $1.50B. ~$1.20/share came from IEEPA tariff refunds — strip that out and they STILL cleared. That's the quality-of-earnings tell.
Margins, sales, and EPS would have exceeded expectations even without the tariff refund benefit.
Forward story is memory. ZBRA mitigating cost headwinds and locking up supply into FY27 — that's the direct answer to Street margin anxiety. Add improving short-cycle dynamics and a multiple still below historicals, and KeyBanc's math gets simple: growth accelerating, valuation not yet extended.
Stock sits at $368.83, kissing the 52-week high ($369.79). $17.4B cap. One raise, one thesis, no pushback. The risk: after a 52% run, everyone's already in the boat — next leg needs the FY27 memory story to show up in actual orders.
AMD — CAPEX TRIPLING TO $808M VS $298M EST buried the beat-and-raise. Q2 REVENUE $11.54B +50%, DATA CENTER $6.72B +107%, Q3 guide above — stock drops 8% anyway. Market wants ROI proof, not just a beat. At 200x forward, the bar is uncompromising. Musk committing xAI to Nvidia exclusively makes AMD's second-source narrative harder — and the claimed 20% CPU edge over Nvidia lacks model number, wattage, and a place on AMD's own price list.
NVDA — BLACKWELL RENTAL PRICING UP 50-60% IN 6-7 MONTHS — mid $2 to just under $4 per GPU hour, one inference cloud plans to pay 100% more at contract expiry. Hyperscalers under-earning on AI infra; demand still outstrips supply. VERA RUBIN NVL72 PER-RACK COST $8.34M, +97% VS GB300 — accelerator cost curve steepens, good for NVDA pricing, inflationary for hyperscaler ROI. Rubin Ultra HBM4 8HI despec is supply-driven, not demand loss.
INTC — 18A yields match TSMC by year-end; EMIB-T yield >90% at ~50% of CoWoS cost. MediaTek ordered; Broadcom and Meta evaluating. Market prices Intel at 9X P/S VS AMD 22X — even valuing fabs at zero, that gap is wrong. 2027 CAPEX RAISED ABOVE $20B signals the filled-fabs path to $13-15 EPS. Expert call still blames the 10nm EUV rejection and IDM coupling starving volume — but TSMC at 100% utilization opens a second-source window.
MU — A SUBSTANTIAL PORTION OF 2027 AI DRAM CAPACITY IS ALREADY COMMITTED; customers get only 60-70% of requested volumes. Rubin Ultra HBM4 despec cuts 2027 HBM demand ~10%, but it's a supply constraint, not demand loss — shortage extends through 2027. Final ASPs set under LTAs. Extraordinary FCF year shaping up. Counterpoint says SK Hynix is only ~1pp ahead of Micron — supports Micron's relative position.
SNDK — Reports Wednesday: NAND LTAs and mid-80s margin commentary are the flags. Kioxia already printed 45x net profit on AI datacenter NAND — flash pricing power is real. Utilization low-80s vs DRAM high-80s, but storage is becoming another memory tier. LTA durability likely holds on game theory.
WDC — Reports Wednesday; nearline exabytes and CY27 coverage are the numbers. HDD/AI storage demand extends visibility. Flash-memory shortage pushes storage closer to the GPU tier.
ALAB — Q2 BEAT CLEAN: EPS 80C VS 69C EST, REV $392.4M VS $361.3M, GROSS MARGIN 73.3%. Scorpio overtaking Aries is the key product-level signal — Gen6 already >1/3 of revenue. If Scorpio dominates, ALAB transforms from retimer vendor to rack-scale fabric company. UALink optionality adds another leg.
AAOI — Trump admin drafting a ban on new Chinese optical transceivers; AAOI named a next-gen certification beneficiary, but 3P capacity can't fill the gap. Pull-forward orders likely. Aug 6 earnings: key is whether AAOI moves up the value chain from modules to critical optical components. If customers buy lasers/components rather than modules, the market should re-rate beyond transceiver comps.
LITE — INP SCARCITY WORSENING; TSEM notes shift from EML to CW/SiPho transceivers. SiPho InP intensity is ~50% lower than EML — LITE benefits from CW light source demand. The bottleneck migrates to lasers, not finished modules. Chinese ban would shift 1.6T/3.2T certs to LITE/COHR/AAOI — but capacity is the constraint. NPO adoption delay adds upside if Chinese supply freezes.
COHR — Named a next-gen certification beneficiary of the Chinese optical ban. Vertical optical integration positions COHR for the InP laser and component shortage. Capacity gap remains the bear case.
AXTI — InP substrate is the likely bottleneck if the ban shifts constraints from modules to raw materials. But China controls InP export risk — if Beijing retaliates, AXTI gets squeezed. InP can't be the straw that breaks the AI camel's back.
FN — Named a beneficiary of the Chinese optical ban, but like LITE/COHR/AAOI it lacks capacity to meet demand. Pull-forward and NPO delay could still drive order flow.
DOCN — Q2 beat with Q3 guide raise; INFERENCE ENGINE CUSTOMERS DROVE TOKEN CONSUMPTION +30X IN 60 DAYS. 85% of AI customer ARR now from inference/core cloud. But EBITDA margin guide disappointed — market wants evidence AI attach converts to margin, not just gross adds.
ORCL — Needs to raise equity — helps credit markets and therefore equity. Pays for itself. Source flags ORCL as part of the AI balance-sheet funding debate.
GOOGL — Google assembled ~$200B OF ANTHROPIC COMPUTE FINANCING: >$150B chips, 2.4GW TPU power backstopped across 10 developments, UP TO $44B LEASE GUARANTEES. Largest AI infrastructure financing ever assembled. But Alphabet is the least transparent hyperscaler on MW/GW capacity and asset-level returns; FCF went negative for the first time since 2004 against $195-205B 2026 capex. Next-gen TPU rumored on Intel 18A/14A.
MSFT — Adds ~1GW/quarter; closest hyperscaler to delivered fleet capacity reporting. Qualitative ROIC support. Custom Arm chips deploying across Azure.
AMZN — AWS Q2 REVENUE $42.2B +37%; OPERATING INCOME $16.6B VS $10.2B; MARGIN ~39.4% VS ~32.9%. AWS now ~60.5% of consolidated OI. Capacity, not hesitation, constrains demand. The less-than-3-year server break-even framework is directly testable — most credible hyperscaler ROIC disclosure.
META — Custom Arm chips deploying across Meta; external price discovery encouraging but not realized economics. AI monetization still unproven in reported financials.
AAPL — LEGAL ESCALATION WITH OPENAI — Apple demands forensics and injunction; OpenAI fires back hours later. Model safety and liability vacuum becomes board-level risk. Not priced into the AI infra complex.
NOW — Subscription rev probably accelerated to ~21% y/y, but CRPO decelerated to 19.3% CC. AI revenue ~$250M/qtr (~6% of subscription), quality mixed. PF OM down; SBC up. At 50x 2027 EPS with downside risk, r/r is poor.
CRSP — Private fund exited CRSP entirely. AlphaFold solved the backend of the protein problem; the harder frontend remains. CRISPR sits on the earlier-generation side of the compute-data-native divide.
ZETA — Q2 REVENUE $443M +44%, BEAT $421M EST; super-scaled customers 197 +17%; ARPU +17%; FCF +73%; FY26 guide raised. AI-driven marketing platform compounding.
WYNN — Q2 REVENUE $1.86B +7%; ADJ EPS $1.24 +14%; WYNN PALACE $653.4M +21%. UAE property opens Sept 2027. Macau steady, no AI beta.
J — Q3 REVENUE $2.4B +8.3%; EPS $1.84 +13.6%; BACKLOG $28.9B +27.3%; FY26 EPS guide raised to $7.20-7.30. Infrastructure backlog durable.
SPOT — Q2 MAUs +12% y/y; net adds 16M missed 17M guidance; stock -4%. User growth decelerating. No AI narrative.
W — +30% after Q2 US growth strongest since 2020 and cost savings ahead of schedule. Crowded short squeeze triggered by the beat.
PINS — Beat but Q3 guide in-line; stock -7%. Gen Z user growth strong, but monetization rate isn't improving fast enough to justify the run-up.
MTCH — Missed. Tinder continues to drag. Dating turnaround not visible — no inflection in engagement. Horrid tape.
BKNG — Maintained full-year guidance; travel demand resilient. Middle East conflict pressures airfare pricing. No AI acceleration.
PSKY — Raised guidance; streaming growth continues. But Paramount-WBD merger antitrust trial set for March; 12 states sue to block. Merger overhang caps the multiple.
BWXT — Rep. Delaney filed purchases up to $80K while on the House Tech Committee. Nuclear/microreactor narrative, nothing moved on company fundamentals.
VSAT — +1,125% since prior call. No new fundamental data point today; momentum/positioning marker.
TSLA — CHINA JULY WHOLESALE 93,579 UNITS +37.8% Y/Y, YTD +30% — but domestic sales -2% y/y, export-led not a domestic recovery. SpaceX purchased $329M of Tesla Megapacks YTD for xAI data centers. Cross-subsidization formalizes — Tesla Energy is a beneficiary of AI data center power demand.
HPQ — HP, Asus, and Acer start using CXMT DRAM in budget laptops amid memory shortage. PC OEMs forced to Chinese memory because Samsung/SK Hynix/Micron redirect capacity to AI. CXMT fills the vacuum — and it's not cheaper than the top three.
LI — Alchip starts revenue from Li Auto's first self-developed AI inference chip, Mach M100, on TSMC N5A. Vertical AI silicon for smart driving.
SONY — Talks with TSMC over image sensor JV at Kumamoto new fab progressing toward definitive agreements; Sony majority; ¥10B earmarked. TSMC process technology boosts mobile/physical AI image sensors.
DELL — Dell assembles systems for the Anthropic-Volta $10B compute deal; Nvidia supplies chips. PLTR read-through positive for private AI infrastructure. AI server demand accrues to Dell.
HPE — PLTR positive read-through for private AI infrastructure; AI systems demand accrues. No direct HPE data point in the source.
BTDR — Volta Infra rents colo from Bitdeer for Anthropic's $10B deal — 121MW Norway site, 16 years, ~$4.7B contract. Bitcoin miner becomes AI neocloud landlord.
BX — Blackstone exploring $36B debt financing to fund Anthropic's Google TPU expansion. Private credit stepping in to bridge AI compute demand.
ARM — Google, AWS, Meta, and Azure all deploying custom Arm chips. Hyperscaler custom silicon validates Arm licensing; AI inference drives per-core economics.
MRVL — Nvidia invested $2B in Marvell. Bull: DSP influence acquisition, custom ASICs pulled toward NVLink, UALink blunted. Bear: Trainium 3 disappointing and custom AI silicon struggled. Not automatically bullish. CXL positioned well via Celestial; HBF needs Marvell/Broadcom/ASIC design-house partnerships.
AVGO — Evaluating Intel EMIB-T packaging; HBF needs Broadcom/Marvell/ASIC design-house partnerships to escape commodity status. Broadcom remains key to scale-up/scale-out interconnects.
TSM — TSMC at ~100% utilization; building additional N3 and N2 capacity. Qnity confirms 3nm/2nm/18A commercialization across logic leaders. Content-per-wafer intact. EMIB-T costs ~50% less than CoWoS and is gaining ecosystem — but TSMC still far ahead in process know-how.
UMC — UMC manufactures silicon bridge components for Intel EMIB-T using mature 12-inch process; strong demand. Validates the advanced packaging ecosystem, adds foundry volume without leading-edge capex.
AMAT — WFE DEMAND ACCELERATING — Qnity Kalrez sees 'accelerated demand for wafer fab equipment.' HBM/3D/advanced logic add etch/deposition steps; content per wafer grows faster than wafer starts. High direct leverage.
LRCX — Same WFE acceleration; etch/deposition is disproportionate in HBM and 3D architectures. High direct leverage; installed-base service revenue grows.
KLAC — WFE acceleration and advanced-node process complexity raise inspection/metrology intensity. Process-step growth benefits the installed-base service model.
ASX — Advanced packaging, AI PCB, and thermal platforms growing >50% at Qnity; larger packages and tighter geometries increase material content. ASE as OSAT beneficiary.
AMKR — Same advanced packaging content >50% growth; larger formats increase material per package. Amkor benefits from outsourced advanced packaging demand.
TTMI — Qnity disclosed pulse-plating wins for high-layer-count AI PCBs. Revenue growth driven by process/content intensity, not board area.
CLS — Qnity/Powell read-through: complex AI servers, switches, and high-layer-count PCBs increase system complexity. Celestica's manufacturing value-add rises with AI rack content.
QNTY — 22% ORGANIC GROWTH; semiconductor tech +17% vs high-single-digit MSI; interconnect +28%; advanced packaging/AI PCB/thermal >50%. RAISED FY GUIDE TO ~18%; no Q3 pull-back; 2027 stacking up nicely. Content per wafer/package/board drives revenue, not units. Barriers to entry rise.
POWL — RECORD $934M ORDERS, 3.0X BOOK-TO-BILL, BACKLOG $2.4B. But 12-month conversion dropped to ~54% from low-60s — duration extension, not near-term acceleration. $400M+ behind-the-meter power island burns 2-2.5 years. Footprint +20% by end FY26.
CAT — Q2 beat; power-generation strong from data center spending; shares +8%. Data center power demand converts to orders.
GEV — Gas turbine orders extending into 2033 — peak-cycle bears keep getting pushed out.
BE — Behind-the-meter power islands expand TAM for on-site generation; fuel cells benefit. No direct order data.
CMI — Behind-the-meter power islands with on-site generation expand TAM for engine generation suppliers. No direct order data.
ETN — 800V DC and 1MW racks redesign the internal DC power chain; content per rack rises.
VRT — Thermal management critical as AI power densities rise; Qnity thermal portfolio >50% growth. 800V DC and 1MW racks increase cooling/power content.
MOD — Behind-the-meter power islands and thermal management demand benefit Modine. No direct data.
ABB — 800V DC and power islands increase electrical distribution/control content.
TSEM — Q2 REVENUE $460.1M +24%; EPS 88C VS 76C; Q3 GUIDE $520M RECORD +31%; 2028 targets raised to $3.6B revenue / $1.2B net profit. INP SCARCITY WORSENING — the optical supply chain bottleneck benefits the specialty foundry.
SNOW — PLTR comp chatter and Zeta collaboration shoutout only. No standalone fundamental data point.
CRWD — PLTR valuation comp only; Morgan Stanley debate questions whether security software wins in either AI scenario. No standalone data.
AI — PLTR's control plane and model abstraction pressure proprietary model monetization and subscale AI platforms. Negative read-through for C3.ai.
PATH — Legacy RPA named a loser in PLTR's control-plane world. Enterprises switch models and orchestrate workflows — traditional RPA gets disintermediated.
ADBE — Broad tech rally breadth only. Tape signal, not fundamental.
FISV — Same. Tape signal only.
CHTR — Same. Tape signal only.