Good morning.
Futures are a sideshow; tape’s churning around AI single names — SPCX -12% erasing $205B in market cap — while semis sit firm after a monster SITM print. SITM beat Q2 estimates by ~$11M on revenue and guided 1.6T optical to double in 2027, with synchronization adding several hundred dollars per rack. That’s the cross-layer signal: not a replacement cycle, an additive one. AMD delivered a record quarter — revenue $11.54B at 56% GM, DC AI set to more than double in 2027, and the $30B Instinct estimate is “probably too low.” Storage is in the green too: SNDK/WDC demand beating consensus, though SNDK’s revenue guide looking soft. Asia’s quiet, but the AI infra bid hasn’t broken — TSMC reportedly outsourcing a key CoWoS front-end step, packaging still the bottleneck. Theme one: optics are the new pigs — SITM, TSEM, ANET, ALAB, CRDO, MRVL, AVGO all get a lift from rack-level synchronization and 1.6T pull. Theme two: compute supply is flooding in — META shopping ~400k GPUs (H100/H200/Blackwell) into merchant compute, pressuring neocloud economics and GPU rental rates, exactly why SPCX got hit. But ETN says no slowdown — operators are expanding in proven locations, so the read-through is consolidation, not cancels. Theme three: GOOGL is bleeding AI talent — Jeff Dean and Oriol Vinyals out to launch Discovery Loop, Noam Shazeer already at OpenAI — that’s not a narrative, that’s a structural problem for frontier model credibility. Theme four: storage has legs but not without warts — LTA coverage expanding to 80% by 2028, but SNDK’s guide says the market isn’t pricing smooth sailing. We’ll hit up SITM and AMD first, then get to optical/networking and the Google brain-drain trades.
AMD printed a beat-and-raise quarter and still got sold 8% after hours. That's not thesis breakage — that's positioning. Stock up ~198% in a year, +15% into the print, everyone already owns it. The bar moved past what a beat-and-raise can clear. The operating story, by every metric in this quarter, got stronger.
Nine shops in this file, one Equalweight, eight Buy-equivalents. PT range $465 to $685. The cluster sits at $650-$675: Jefferies, Bernstein, Roth/MKM at $650; TD Cowen $675; Benchmark $685. Stifel $635. Truist $594, DA Davidson $550. Morgan Stanley's $465 Equalweight is the outlier and the one worth respecting.
Five firms moved targets: Jefferies $640→$650, Bernstein $600→$650, Truist $478→$594, DA Davidson $425→$550, MS $410→$465. KeyBanc sits at $725 on the high end, cited across multiple files. Consensus baseline is the 2027 framework: data center more than doubles, server CPU >70%, Instinct clears $40B CY27. The bulls think that's conservative.
Management gave an actual 2027 data center framework — more than double YoY, server growth >70%, and an implied near-tripling in AI GPU revenue. That's not fluff; that's guide-able conviction.
H2 2026 server CPU growth got raised to >80% on improved supply. Helios production shipments start Q3, accelerate Q4 into 2027. Management says server demand stays supply-constrained through 2027 on wafer, packaging, and substrate bottlenecks — that's a demand problem, not a supply one.
Embedded is entering recovery. Client is better than expected. Base business firming alongside the AI story is the quiet tell here.
One caution: Q3 opex guide of $3.65B vs $3.54B Street. They're spending into the opportunity. That's fine if the growth shows up, and a margin question if it doesn't.
Bull: This is a demand signal, not just a beat. DC revenue $6.718B, +107% YoY, 58% of total revenue versus 42% a year ago. EPYC +70% on double-digit unit and ASP growth. Instinct ~$2.9B, +190%. GM up 200bps with CPU growing faster than GPU — margin-accretive mix working exactly as advertised. Truist's CY27 EPS went from $10.61 to $14.49 in one quarter. TD Cowen thinks AMD hits $20+ EPS in 2028, ahead of the company's own November 2030 model. Rate of change is the story.
Bear: The bull case is priced for 2027 already. Stock trades at 160-176x trailing. Morgan Stanley's warrant math is the most important table-pounding point in this file — AMD is issuing >$15B in warrants against $15-20B in revenue from two customers. Treat that as cash expense and it eliminates profitability over the next three to four years. And Musk just went exclusive NVIDIA on the SpaceX call. The "AMD wins #2" narrative is real, but the default hyperscaler choice is still NVIDIA — and AMD trades at a premium to NVDA and AVGO on MS's numbers. Crowded longs don't need bad news; they just need "not enough good news." We saw that after-hours.
"The reaction reflected a crowded long and a financial bar that had moved materially higher after AMD's Advancing AI event, Intel's stronger-than-expected report and a roughly 15% one-week gain in AMD shares." — Benchmark's Cody Acree
"The warrant issuances, while potentially effective for stimulating customer interest, represent a cost that would eliminate profitability if treated as a cash expense." — Morgan Stanley
"Going forward, we've decided to build exclusively on NVIDIA because we think the Vera Rubin architecture is the best architecture. So, we're exclusive to NVIDIA." — Elon Musk, SpaceX earnings call
Three things for PMs. One: Intel's stronger print raised the bar, AMD cleared it, and the tape still sold it. That's what a crowded $845B long looks like — discipline the position, not the thesis. Two: MS's warrant point is under-discussed across these nine articles; if that math flows into consensus models, the "overvalued" debate gets louder even as fundamentals inflect. Three: AMD is the highest-beta way to express AI capex — the second derivative. If hyperscaler spend wobbles, AMD has the most air beneath it. If it holds, $40B Instinct in 2027 is the baseline, not the bull case.
ANET DIDN'T JUST BEAT — IT REWROTE THE CONTRACT. Q2 was the first $3B quarter: $3.04B revenue vs $2.83B consensus, EPS $1.02 vs $0.88. The headline is the guide, though: FY26 GROWTH TO ~40% FROM 27.5% three months ago. That's a 12.5-point re-rate in one quarter. Q3 guide of $3.3B = 43% growth — Street was at 28%. The beat is getting bigger, not smaller.
What changed? Supply. Last quarter Arista had demand it couldn't ship. This quarter management locked in "critical chip supply" via PRE-PURCHASE COMMITMENTS. The visibility is absurd: product deferred revenue +$600M q/q to ~$4.2B, purchase commitments +$800M to $9.7B. That's a roadmap, not a quarter.
Five firms raised PTs to $240-$260 from a prior $187-$210 cluster. All Buy/Outperform. Nobody's fading this print. UBS went $187 to $259 — a 38% PT hike in one morning. Needham on top at $260. BofA the low at $240, and they had to expand the multiple to 49x 2027 EPS (from 46.5x) to get there. At $197, the range implies 22-32% upside — and the stock already ran 20.6% this week.
Bull case: This is a 40%-growing franchise at a ~$250B market cap. Deferred revenue growth still ~70% — decelerating from 90-100%, but from absurd to merely great. Management flagged "high likelihood" of adding one or possibly two new 10%+ customers — Needham names Google and Oracle. Strength is broad across hyperscalers, cloud, campus. Every AI cluster buildout is share-take from legacy networking. You pay up at 49x 2027 EPS, but you get visibility most AI names can't show.
Bear case: The stock ran 20.6% in a week. Deferred revenue growth decelerated 20-30 points — the rate of change is peaking. Supply catch-up flatters this quarter: pent-up demand shipping late, not a new run-rate. At 66x trailing earnings, perfection is priced. If the deceleration narrative builds into Q4, that multiple doesn't cushion — it amplifies.
Known: AI demand is strong. Arista was supply-constrained. Growth was accelerating.
New: Supply secured via pre-purchase commitments. FY guide raised 12.5 points. Deferred revenue +$600M, purchase commitments +$800M. Potential Google/Oracle as 10% customers. And the one data point bears can hang a hat on: deferred revenue growth decelerated to ~70% from the 90-100% range.
"This quarter's results likely exceeded expectations of bullish investors." — TD Cowen
"Management said the company secured critical chip supply during the quarter after demand had exceeded its ability to supply in recent quarters." — Needham
"The company affirmed a high likelihood of adding one or possibly two new customers representing at least 10% of revenue." — Needham
This is the supply-side loosening of the AI trade. AVGO and MRVL read positive — Arista securing "critical chip supply" means merchant silicon is flowing through the ecosystem. ORCL and GOOGL read positive if they're the new 10% customers — hyperscaler AI capex is broadening, not rolling over. CSCO reads negative — every campus/core sale Arista books is share-take. The thematic signal for PMs: the market is rotating from "who has AI demand" to "who can secure AI supply." Arista just proved supply gets the re-rating.
Verdict: Q2 beat and a margin guide raise are real, but the Q3 guide is the tell — growth steps down to 13-15% and one-time tailwinds did the heavy lifting. UCAN acceleration and Performance+ traction are genuine. But the market prices PINS at ~50x earnings while management guides to mid-teens growth — that's a compounder multiple on a moderate-grower trajectory. PTs cluster $32-$34; DA Davidson's $27 is the realistic holdout. At $25.58, r/r skews positive — just don't confuse a clean beat with a thesis change.
Clean beat across the board: REV $1.18B (+18% YOY, ~3% ABOVE CONSENSUS), ADJ EPS $0.43 vs $0.36 est, ADJ EBITDA $311M — 15% above consensus. MAUs hit a record 640M, 12th straight quarter.
Decompose the growth and it gets messier: FX +1pt, Prime Day shift +0.5pt, World Cup +1pt = 2.5pts of the 18% from one-time boosts. The real signal is UCAN, which ACCELERATED FROM 13% YOY IN Q1 TO 18% IN Q2. EU stayed weak and management openly says it stays weak into Q3 and potentially Q4. Mizuho's read: normalized growth accelerated slightly, and the company raised 2026 adjusted EBITDA margin guidance by 1pt.
Bull: The UCAN re-acceleration is product-led, not luck. Performance+ now runs 30% of lower-funnel revenue, and P+ advertisers grew lower-funnel spend MORE THAN 2X FASTER than non-adopters. ROAS +28% in SMB testing. That's the same algorithmic-lower-funnel playbook Meta ran with Advantage+ — and PINS is early innings. The margin guide raise proves flow-through. BMO's kicker:
"Pinterest achieves cost per transaction at 8% the cost of comparable proprietary models." — BMO
If that open-source AI edge holds, the margin story compounds into 2027.
Bear: One-time events inflated the quarter, not durable strength. Q3 guide implies 13-15% YOY — a FOUR-POINT STEP-DOWN from Q2. DA Davidson frames it perfectly:
"The guidance suggests low-to-mid-teens growth in coming quarters rather than durable high-teens growth." — DA Davidson
EU is a drag with no clear fix timeline. At ~50x trailing P/E and a decelerating top line, the multiple does all the work. Mizuho said it best: the quarter is "not thesis-changing for bulls or bears." Polite way of saying the beat doesn't buy durable re-acceleration.
Four PTs, all Buy/Positive/Outperform: BMO $34, Mizuho $32, Susquehanna $32, DA Davidson $27 (from $26). Range $27-$34, cluster $32-$34. BMO flags the stock at $25.58 sits below both its target and InvestingPro fair value of $34.20. DA Davidson's $27 is the lonely stick — same positive read on Q2 execution, but pricing in the Guide Reality.
PINS is the high-beta pure-play on the digital ad tape. Performance+ is a direct echo of Meta's Advantage+ narrative: algorithmic lower-funnel, SMB-driven incremental dollars, ad-platform-enhancement ROI. Watch SNAP and RDDT prints for corroboration. The regional split also rhymes with the broader market — US dollars are healthy, EMEA is the laggard. And the step-down guide is a warning for anyone paying high-teens multiples across social: even the winners guide to moderation.
Two analysts, $56 apart. Barclays at $49 Equalweight. Needham at $105 Buy. Barclays wrote VSAT at $86 pre-market; the stock closed $81.12 — that's a 305% trailing return and the Street still can't agree on what it owns. The spread IS the debate. I lean Needham.
Q1 FY27 (June): revenue -1% YoY, 3.6% below Street. Service revenue held up, but PRODUCT REVENUE MISSED BY 10.3% — that's the quality tell. Group EBITDA -7% (DAT -20%, Comms Services -3%). Fixed and Maritime still bleeding at -7% YoY; Aviation and Government SatComm strong. Awards $1.3B, +10% YoY. Net leverage down 0.4x to 3.2x. Backlog grew. Cash flow improved.
They kept FY27 guidance — H2 meaningfully stronger than H1. Everyone accepts that.
The real news is the fleet: VS-3 F2 in orbit testing, service entry by September. VS-3 F3 over Asia-Pacific late Aug/early Sep. Management says these TRIPLE GLOBAL CAPACITY. In four words, that's the bull thesis.
Needham went to $105 from $90, keeps Buy, with estimates "largely unchanged" — this is a sum-of-parts markup, not an earnings upgrade. They're paying for defense segment value, spectrum holdings, and the pending Equatys global JV. They also see commercial-DAT synergy accelerating; management telegraphed exactly that on the call.
"The price target increase reflecting the high value of the defense segment and optionality for spectrum holdings and the Equatys joint venture expected in coming months."
Barclays holds $49 Equalweight — miles below the tape. They're not disputing the satellite story. They're anchoring to the operating print: product revenue -10.3%, DAT EBITDA -20%, no actual EPS figure printed (consensus had a -$0.33 loss). And the strategic review — the supposed unlock — has management saying NO INTENT TO SEPARATE THE BUSINESS for now. That kills the breakup catalyst. Optionality without a catalyst is just a carry cost.
This is binary into September. Bull case doesn't need estimate revisions — it needs narrative confirmation: F2/F3 enter service, Equatys closes, DAT inflects in H2. Bear case is pure valuation: after 305%, you're still printing adjusted EBITDA declines and sub-consensus product revenue.
The r/r isn't clean at $81. But Needham's getting paid to price optionality, and the optionality here is real — spectrum monetization plus a global JV plus tripled capacity. Watch the F3 service-entry date. That's the trigger, and it's a month away.
APTV got buried on a beat-and-cut. Q2 print was fine — EPS $1.63 vs $1.42E, revenue $3.3B in line — and it didn't matter. The guide cut is the only number the tape cares about: FY26 EBITDA down ~3%, H2 guided ~7% BELOW prior. Stock -17% in a session where the S&P was +2%. Now trading near the 52-WEEK LOW OF $46.41. The spin-off halo is officially dead.
Management did themselves no favors copping to it. They said they "were not conservative enough" on launch and ramp assumptions. Then CEO Clark started talking about "continually evaluating our business portfolio" — which is portfolio-review speak. Markets read that as optionality they'd rather not price.
Three firms moved Tuesday, none of them upgrades:
Bull: The valuation is getting interesting. 7.8x 2027 EPS vs an 8.7x supplier average, with ADAS/interconnect content growth and a half-of-FCF buyback cushioning the downside. UBS's math says the discount is overdone.
Bear: "Show-me" is the right frame. The execution discount persists because the execution keeps not showing up. China domestic is a multi-quarter problem, export wins don't scale fast enough, and you're asking PMs to trust guidance from a team that just admitted it wasn't conservative enough. The burden of proof is on management.
"The company has entered 'show-me' territory as investors assess its growth resiliency." — Deutsche Bank's Edison Yu, which sums up the whole setup in one sentence.
Worth the risk? Only if you believe the non-auto story compounds faster than China bleeds. Right now the market is betting on the bleeding.
Double PT raise this morning — DA Davidson to $32 (from $30, Buy), KeyBanc to $27 (from $22, Overweight) — and the market's already lapped KeyBanc. Shares trade north of $28. DA Davidson's target is the only one with room, and it's built on 24x 2027 FCF. That's a growth multiple, not a value pitch.
Revenue $443M vs $420.7M consensus — a beat consistent with the trailing eight-quarter average. Ex-political organic growth at 28% y/y for the SECOND straight quarter (vs 27.4% in 2Q25). ARPU expansion plus more super-scaled customers doing the heavy lifting. Adjusted EBITDA +56% y/y. Company is now GAAP profitable.
But the EPS print is the asterisk: $0.03 vs $0.20 expected. An 85% miss. That's the kind of divergence that gets growth names sold, not bought. Tape hasn't cared — yet.
Management raised Q3 guidance and still embeds a 2-5% cushion in the FY26 outlook against the 2028 framework. Translation: beats are modeled in. DA Davidson expects outperformance through year-end, pointing to recent partnerships as unmodeled upside. KeyBanc's angle is sharper — the Palantir partnership has produced two signed agreements with more in the pipeline, and Athena is seeing early traction in accounts pulling >$1M ARPU.
Best line of the morning from KeyBanc's Jack Nichols:
"Growth rates exceeded those in management's 2028 model, driven by scaled growth and cross-selling."
Bull: organic growth accelerating at 28% while flipping to GAAP profitability — Palantir sales cycle just starting to convert, 2-5% guide cushion means beats keep coming. The 2028 model keeps getting exceeded.
Bear: 52% run in six months, KeyBanc's PT already in the rearview, EPS missing by 85%, and fair value models flag the stock as slightly overvalued at current levels. The r/r at $28 is not what it was at $18.
The bull/bear fight here is whether Spotify's soft Q3 guide is a strategic reset or a growth stall. Bernstein says investment. The tape says otherwise — stock -8.75% on the week. Cantor sits in the middle with a Neutral rating and a PT raise that's basically a shrug in dollar terms.
Q2 was a mixed print that traded like a coin flip — shares swung from -6% to +3% intraday before reality bit. EPS came in light at $2.61 vs $2.76 consensus, but that's noise. Revenue $4.78B, in line. Op income 3% above Street. Record gross margin 33.4% (+180bps YoY) — though some of that is a one-time Canada tax benefit, so don't annualize the whole thing.
The milestone: 300M PREMIUM SUBS — 7M net adds in the quarter. Premium revenue +16% YoY. Ad revenue only +3%, but automated ad channels hit ~40% of ad revenue, up from ~30% in Q1. That's the underappreciated part of this story. FCF +14% YoY to EUR 797M, and a perfect Piotroski Score of 9. The balance sheet is not the problem.
The market is fixated on one thing: Q3 MAU and margin guidance came in light. Bernstein frames it as deliberate — a quality-control move in under-monetized developing markets, plus accelerated product spend (Prompted Playlists, SongDNA, DJ). They expect the user weakness to reverse in Q4.
"Spotify's user weakness is limited to under-monetized developing markets, reflects a targeted initiative to right-size and quality-control its installed base, and is likely to reverse in the fourth quarter."
That's the strongest bull line you'll get this week — from a firm at $625, no less. Cantor's less convinced. Raised PT to $530 from $510, but that's just rolling the valuation forward to 22x FY27 EBIT. Neutral. They see steady fundamentals, not accelerating ones. The Merlin AI remix/covers deal extends the UMG-style licensing push, but it's not a needle-mover yet.
Bull: Q3 guide still implies >40% incremental gross margins. That's the number that keeps the 2030 targets alive. Bernstein says the softness is contained and controllable — user weakness is limited to under-monetized developing markets, and the product investment is a feature, not a bug. Subscriber engine hums — 7M net adds. Ad automation shift is a margin story in the making. PEG of 0.24 makes the valuation hard to hate if you believe the growth rate.
Bear: MAUs decelerating, and management's response is to spend more. EPS miss plus a light guide says the growth narrative is being managed down, not up. Cantor's $530 PT — roughly flat to spot — sets a low ceiling. If Q4 doesn't deliver the reversal Bernstein promises, this stock has further to fall from a week that already hurt. The record gross margin isn't clean, either.
The 3p: The bogey is Q4 MAU reversal. Bernstein's right that user softness fits the broader internet platform trend this earnings season — but that cuts both ways. When the whole sector guides light, "strategic" is often just the polite word for "demand is soft." If MAUs reverse in Q4, $625 is live. If not, Cantor's $530 is the top of the range. The r/r skews long only if you trust management's control over the narrative — and the tape is telling you it doesn't.
ON's bull story sounds great on a call. The prints just aren't matching the pitch. Truist cuts PT to $91 from $101 (Hold) while raising CY27 EPS to $4.35 from $4.21. That's the entire battleground in one line: estimates up, targets down.
Truist's math: 21x on the new EPS, a 6-turn discount to recently lowered peer multiples. They see the same signals as management — book-to-bill climbing, lead times stretching, pricing firm, AI/data center traction real. But then they look at actual results: REVENUE BEATING BY JUST 1-2% PER QUARTER, GROWING HIGH SINGLE DIGITS Y/Y. That's the disconnect. The narrative says "cyclical recovery + AI supercycle." The P&L says "slow grind higher."
"Cyclical recovery and artificial intelligence and data center traction continue for the semiconductor maker, but actual results appear less strong than the company’s messaging suggests."
The rest of the Street isn't screaming either. Needham keeps Buy but trims to $116 from $130. Stifel cuts to $90, stays Hold. Cantor sits Neutral at $110. Q2 did beat — $0.74 non-GAAP EPS on $1.6B rev vs. $0.71/$1.59B estimates — but the beat is small and the margin story (39.3% GM, fourth straight up) is steady, not explosive. (Truist also flags business exits as part of the gap between messaging and revenue.)
Bottom line for PMs: this is a "show me" stock. PT cluster hangs between $90 and $116; no one's going aggressive until the quarterly surprise starts measuring in cents, not pennies. Wait for the narrative to catch the print.
Benchmark keeps Buy at $335 after a news-driven vacuum — TSEM gapped down on FCC chatter about restricting Chinese-made optical transceivers, then reversed to close green. The perceived exposure runs through China-based module customers InnoLight and Eoptolink. All unconfirmed, management declined to comment, and the tape decided the policy risk is noise.
The real anchor is the Q2 print: EPS $2.68 VS $2.21 EST, revenue $1.4B vs $1.32B — record profitability, record revenue, and the SiPho conversion + margin leverage story intact. No new strategic news (FY28 model and Japan expansion came out in July), so the move was pure signal on the operating trajectory, not headline hype.
"The operating result was better reflected by the shares' positive close than the opening move."
STOCK +27.5% ON THE WEEK. Trades at 85x P/E — that's the bull case priced for perfection. The policy overhang is a known unknown, but until the FCC actually moves, this is a momentum tape with fundamentals behind it. Not chasing here; watching for a digestion pullback.
Cantor RAISED to $80 (from $65) on the Q2 beat. Evercore CUT to $95 (from $120). That divergence tells you everything: this is a valuation debate, not a fundamentals debate. The underlying business is fine — the market just stopped paying for it.
WIX printed a clean beat — revenue +2% vs consensus, operating income +5%, and ADJ FCF OF $61M VS $36M ESTIMATED (nearly 2x the bogey). Bookings decel to +12% from 15% in Q1, but that was already guided and already in the numbers. Q3 guide at low-double-digit growth keeps the decel narrative alive. The Street wanted more.
"Wix.com has underperformed the Nasdaq by 50 percentage points year-to-date and trades at 6 times fiscal year 2027 estimated free cash flow."
This is a classic "good print, bad tape" situation. The stock is DOWN 36% YTD, down 48% over the past year — the market has already repriced the AI narrative to "show me, don't tell me." Cantor's 7x FY27 FCF multiple is the new reality. Evercore's $95 PT is the old reality. Neither is wrong — it depends on whether you believe Base44 margins hit 60% in 2H26 and the FCF compression to high-teens margins is transitory. That ~12pp of FCF margin compression is the entire bear case in one number. Management says it's investment, not deterioration. Stock says otherwise.
At 6x forward FCF with a 15% FCF yield, the risk/reward is finally getting interesting for a name with this much AI optionality. But "cheap" and "catalyst" are different things. Wait for bookings re-acceleration or a Base44 proof point before stepping in front of this train.
UBS cuts DOCN to $140 from $155, stays Neutral — and the cut is about capacity disclosure, not the quarter. Q2 was clean: EPS $0.45 vs $0.25 est, revenue $281M vs $277M est, +29% YoY. But the stock fell premarket because the crowd wanted more from FY27 guidance. Management kept ">50%" growth — no raise. No new MW beyond the 20 already announced. Investors were positioned for >55% and another 10-20MW. Got neither.
Here's the frustrating part: management admits the outlook has improved materially since issuing guidance. A ~30% GPU price increase isn't even baked into the original forecast. So the conservatism is real, but at $122.97 (up 270% in a year) you don't get paid to wait for management to catch up to its own math.
"60 megawatts of capacity in 2027 at $17 million per megawatt, plus an additional 10 megawatts in the fourth quarter, would imply roughly 70% fiscal 2027 growth from capacity additions alone."
UBS's math points to ~70% growth from capacity alone versus the 50%+ guide. That gap is the whole story. Either management is sandbagging hard or something's missing. A stock that's already run 270% prices the optimistic scenario — the market wanted the company to de-risk the story with hard numbers and a revised guide. They didn't. This one's a wait-for-clarity, not a chase.
Erste cuts AMAT to Hold — the classic "great company, expensive stock" call. The thesis is brutally simple: AI exposure is real, but 207% in one year and a 51x P/E mean the market already paid for it. Revenue growth of just 3.3% LTM versus the sector is the problem — the multiple can't expand if the growth doesn't show up.
The bull counter is equally simple: AMAT is the picks-and-shovels play for advanced packaging, DRAM, and high-bandwidth memory, and the capex cycle has legs. Stifel and Mizuho set $650 PTs, TD Cowen goes to $700, all citing product portfolio strength and multi-year fab equipment spend. So you've got a valuation stall vs. an order-cycle story — and today, the valuation guys are winning the tape.
"The stock's price-to-earnings ratio is higher than the sector average while growth lags peers, limiting potential for further stock price gains."
Worth noting the tape context: SOX dropped 4.8% the prior session, SK Hynix ADRs bouncing 6.8% premarket — semis are churning hard and AMAT's beta to that chop is high. Downgrade is a valuation signal, not a business deterioration signal.
Needham's the bull — PT to $160 from $140, Buy, and the tape's already +22% on the week (+146% over the year). The fundamentals back it up: Q2 REVENUE OF $201M (+62% YoY) beat cons by 8%, EPS OF $0.53 doubled and cleared by 19%, September guide is 25%/11% above the Street.
The momentum story is broad-based. Attach rates, new applications, ASPs all inflecting higher, with backlog visibility through 2027 and an exit rate near $1.2B ANNUALIZED. AMI is the kicker: $200M+ base revenue in 2026, +25% in 2027, gross margins in the mid-to-high 70s, EBITDA north of 40%. Accretion begins Q4 2026.
"Momentum continues to grow across the business, with attach rates, new applications, and average selling prices inflecting higher."
BofA's the counterweight — Underperform, PT to $125 from $105. They see the strength, they just hate the multiple (Needham's at 40x its $4.00 2028 EPS). The channel inventory shift from drawdown to buffer is the tell worth watching: customers aren't destocking, they're pre-buying into a constrained environment. That's durable, not a one-quarter fluke.
Needham jacks ALAB to $425 from $260 (50x 2028 EPS of $8.50) after a beat-and-raise that has the stock +45% in a week. Revenue +104% TO $392.4M, EPS $0.80 vs $0.69 cons (15.9% beat), gross margin 76%. This isn't a one-product story — every line grew.
Scorpio X switching ramps materially in H2 2026, and Scorpio becomes the largest segment by Q3. The kicker — content per XPU keeps expanding:
"The content opportunity for Scorpio X solutions alone to grow well beyond $1,000 per XPU in future generation AI platforms."
Also landed a fresh Leo CXL design win at a US hyperscaler. That's the diversification bull case: switches today, CXL controllers tomorrow, all riding the same AI capex wave.
$425 PT is a premium multiple, but PEG of 0.48 says growth clears the bar. Stock faded slightly after-hours following the run — some digestion at these levels wouldn't shock. The real bogey into year-end: any slip in Scorpio X ramp timing. That's the entire thesis.
Twenty-six percent in a week. That's the real story here, and Tigress just threw gasoline on it — $690 PT, Buy maintained, prior level undisclosed (cowards). Stock's at $491, so that's a 40% flag-plant from the street's biggest bull.
The fuel underneath is legit. Azure +43% vs 41% est, guided +45% vs 42% consensus for Q1. Copilot at an inflection point, record commercial backlog, 18% LTM revenue growth against a 27.2 P/E. PEG 0.87 — you're not paying for the AI dream, you're paying for compounding execution at a discount. That's the whole bull case in one number.
"Copilot adoption is reaching an inflection point, creating a significant long-term AI monetization opportunity across Microsoft's installed base."
Worth the flyer? The installed base is the moat — every seat is an upsell. But check the comps: UBS at $525, Cantor at $522. That's the rational cluster. Tigress at $690 sits 40% above the next-highest target, which means they're either seeing something nobody else sees or they're renting attention. Backdrop helps: Morgan Stanley has hyperscaler capex at $1.2T by 2027, 30% YoY, all major clouds supply-constrained. Demand isn't the problem — capacity is. That's the kind of scarcity that accrues to the incumbent with the deepest pockets.
Bottom line: great company, great print, late entry. 26% in five sessions is a hype run, not a risk/reward setup. Let it consolidate, buy the next Azure scare instead.
D.A. Davidson lifts PT to $16 from $14 but keeps Neutral — and the stock already trades THROUGH that at $16.62 near the 52-week high. That's the whole setup: the cheap-and-ignored story has flipped to cheap-but-priced-for-acceleration, and the fundamentals underneath don't yet support the multiple re-rate.
Q2 beat was real ($0.30 vs $0.23 est; rev $60.5M vs $58.25M) on overages revenue, enough to nudge FY26 revenue/EBITDA guides slightly higher. But the quality metrics tell the bear story: ARR organic growth DECELERATED to 5% YoY from 7% last quarter, dollar-based net revenue retention FELL 2pts QoQ to 103%, and FY26 ARR guide is reiterated at just ~3% organic growth. One overages-driven beat vs. a decelerating recurring base — that's the tension.
Bull case: P/E 9.1, PEG 0.39, new unified product platform rolling out, +55% six-month tape says momentum is real. Bear case: the market is paying up for a reacceleration narrative that ARR data doesn't confirm yet, and the PT below the current quote shows the sell-side isn't buying it either.
"D.A. Davidson said it will continue to await signs of improved execution before changing its rating."
Translation: they want to see ARR reaccelerate and NRR stabilize before blessing this move. At $16.62 with a $16 PT, the risk/reward is no longer asymmetric long. Pass until the recurring metrics confirm the story.
Rosenblatt reaffirms Buy, $250 PT — stock closed at $92, so that's a ~2.7x r/r call. The fear trade: Meta building a cloud business to offload excess AI compute (CRWV -14% on July 1, Nebius -17%) plus a forced liquidation at shareholder Situational Awareness Fund. Rosenblatt's checks say the market's wrong — demand for AI training/inferencing still exceeds supply, no idle capacity piling up. They did nudge up interest expense and capex estimates after a deeper dive on the $35.1B debt load, but that's a tweak, not a thesis change.
"Checks from its June 25 initiation report remain valid for the second quarter, showing more demand for AI training and inferencing than supply."
Street's lining up behind them: Piper launched at OW/$151, BofA keeps Buy while raising FY26 capex est to $34B from $29B. Add the chatter of NVDA guaranteeing $250B in OpenAI financing — that's a mega-structure tailwind for hyperscale AI names. Earnings Aug 11; CRWV has a history of overshooting implied moves, so buckle up.
Truist is holding the line. Reiterated Buy with a $215 PT while the stock sits 41% off the June 22 peak — despite the recent week's 30% bounce back to $131.30. The call: fundamentals haven't changed, this was a positioning shakeout, and Q2 earnings is the reaffirming event that resets the narrative.
The bull math is straightforward. Data center/AI, defense, and industrial recovery drive EPS growth >60% in 2026 and >35% in 2027. Margin expansion comes from board complexity and favorable supply/demand. Truist keeps CY27 EPS at $5.65, applies a 38x multiple, and lands at $215 — a PEG of 0.49 that makes the multiple look less aggressive than it first appears.
One pushback: the platform's own Fair Value model flags TTMI as overvalued at current levels. So this isn't a consensus "buy the dip" — it's a conviction call from one house against the tape.
Context worth carrying: S&P just lifted the outlook to positive on >25% revenue growth next year from AI PCB demand. Russell 1000 inclusion goes effective June 26. Two European hardware acquisitions close in June. None of that moves today's tape, but it stacks the deck if Q2 confirms.
No blockquote — the source didn't give us a line strong enough to pull.
Piper bumps PT to $12 from $10, keeps Neutral — and the stock already sits at $12.10. That's the entire note in one line: good execution, no room left. Credit the EX engine all day, but Piper won't chase until the CX albatross stops weighing on the mix.
"The customer experience business represents approximately 39% of annual recurring revenue and is growing 4% in constant currency, while the artificial intelligence business remains nascent and the employee experience business could face increased competition."
PM read: quality compounder trading at a fair price. Wait for CX re-accel or a dip on AI digestion. Neither happens today.
DA Davidson just raised QLYS to $165 from $135 — and the stock is already 14% ABOVE that at $188. That's the tell. Sell-side is chasing a tape that's run ahead of fundamentals. Neutral maintained, which is Wall Street for "nice quarter, but I'm not paying this multiple for it."
The quarter was legit, though. Revenue +11% YoY — ACCELERATING 1PPT Q/Q, the first real reacceleration this name has shown in a while. Calculated billings came in well ahead of consensus at +16% YoY. Gross margin 83%. Dollar-based net expansion ticked up to 105%. EPS $1.98 vs $1.78 estimate. Revenue $182.2M vs $178.6M est. Beat and raise, clean print.
"Qualys raised its 2026 guidance for revenue and EBITDA. The company now expects current calculated billings growth of 9% to 10% year-over-year for 2026, up from prior guidance of 7% to 8% year-over-year."
Bull case: QFlex and enterprise threat management momentum is real. Upsell activity is accelerating. The guidance raise confirms the demand inflection isn't a one-quarter fluke. At 105% NRR, existing customers are spending more — that's the quality signal.
Bear case lives in the back half. Q2 billings grew 12% YoY; H2 is guided to just 7-8%. That's a sharp deceleration. The stock is up 19% in a week and 21% YTD — the market is pricing the H1 strength as durable when management is explicitly telling you it's not linear. DA Davidson's $165 target is just the math catching up to the guide.
Stock's at $188. You're paying for the reacceleration narrative while the H2 guide suggests a growth fade is built into the plan. The quarter was the first green shoot in a while, but the easy money was the pre-print dip, not today's chase.
Oppenheimer finally throws in the towel on the bear case. Upgrade to Outperform from Perform, PT to $180 from $160. STOCK SURGED 35% IN A WEEK TO $147.72 — the Street was waiting for someone credible to bless the deleveraging story.
The bogey was always the CMC acquisition overhang. Oppenheimer had been cautious on internal constraints post the levered 2022 deal. Now they see visible execution, margin acceleration, and faster deleveraging. CEO Reeder and new CFO Nagesh get the credit.
Q2 was the trigger: NON-GAAP EPS $0.93 VS $0.82 ESTIMATE, REVENUE $883.2M +11% Y/Y, ADJ NET INCOME +42%. The beat wasn't massive — it's the trajectory.
Firm bumps 2026 rev +3% to $3.6B, EPS +9% to $3.85. 2027 rev $4.1B, EPS $4.95 (+8%). The Nov 9 investor day is the real catalyst: AI materials strategy formalization could validate a higher long-term growth framework.
"Oppenheimer cites visible operational execution, rising confidence in CEO Dave Reeder and new CFO Sukhi Nagesh, accelerating growth and margins, a potentially stronger 2027, and faster deleveraging."
Caveat: InvestingPro fair value flags the stock overvalued here. 35% in a week prices in a lot of good news. But for a name shedding its balance-sheet overhang, momentum is the trade into investor day. Deleveraging stories don't die on valuation — they die on execution slips. Nothing slipping yet.
Stifel's the first to say it plainly: the semi/AI rally ran ahead of the revenue curve. SEMI/AI ORDERS UP >50% — but long lead times delay recognition, so the P&L can't keep pace with the narrative. Meanwhile academic demand stays weak. That's the setup behind the fresh $50 PT (from $45), a Hold that's basically pinned to the current $50.30 tape. Translation: the re-rating is done, now the business has to deliver.
Q2 showed the split perfectly. EPS $0.49 beat the $0.3832 bogey on margins and cost cuts; revenue $838.5M missed the $853.4M consensus. Organic growth miss, stock -18.5% in a week. Management left FY guidance untouched — they're implicitly saying the back half pays off, but the market is done paying up front.
Stifel's framework: low single-digit growth this year, mid-single next, low-20s multiple for tools. Peers have started chirping more constructively on the academic segment, but nobody's backing it with numbers yet. The bull case rests entirely on the semi/AI kicker flowing through revenue — and that's a 2027 story masquerading as a 2026 title.
"The stock had outpaced fundamentals prior to the recent decline, creating an opportunity to reassess potential upside."
Fair. But "reassess" isn't "buy." Hold rated, PT at spot, ball's in management's court on the H2 conversion.
Two PT hikes, one story: the WFE upcycle is finally hitting the margin line and the Street is paying up. DA Davidson to $85 (from $82), Needham to $88 — both Buy. Q2 delivered: revenue $294.8M (+24% YoY, vs $268.2M consensus), EPS $0.34 vs $0.31. But the real tell is non-GAAP GROSS MARGIN 14.1% — ABOVE the guidance band — with a SECOND CONSECUTIVE quarterly guide-up. That's operating leverage arriving early.
Both firms are on the same page: manufacturing investments are converting to earnings leverage. Ichor now says it grows at the HIGH END of WFE growth (~30%) this year, with industry visibility improving into 2027. Seven analysts revised estimates higher post-print. So why the pullback on the day? Because after +287% YTD, "beat and raise" is table stakes. Q3 revenue midpoint $330M vs $314.5M consensus, EPS midpoint $0.45 vs $0.41 — clears the bar, doesn't break it.
"Manufacturing investments are beginning to translate into higher margins and earnings leverage."
Stock's up ~3.9x on the year at $67.88, market cap $2.53B. InvestingPro's Fair Value flags overvaluation at current levels — fair enough; a lot of 2027 is in the price. But the margin inflection is real, WFE momentum is real, and a THIRD straight guide-up next quarter would reset the narrative. PMs: lean Buy on pullbacks, not chases. The easy alpha is already banked.
Desjardins throws in the towel — Hold from Buy, PT C$13.50 from C$19. BofA already there: Underperform, C$13 from C$22, 2028 FCF estimate cut ~30%. Two shops, one message: THIS ISN'T A 2026 BLIP, IT'S A STRUCTURAL RESET.
"The guidance implies a reset of free cash flow expectations rather than a 2026 blip."
Management guided 2026 organic EBITDA DOWN 3%. Q2 service revenue -1% to CAD 4.4B, EPS CAD 0.16 vs 0.22 a year ago. The 55% dividend cut is the tell — they need the cash for capex, and the market read it correctly. Stock fell 11% on July 31 and sits at $9.53, down 34% over the past year.
The weird part: the valuation premium to peers barely compressed despite the drawdown. P/E still 22. Desjardins calls that "increasingly difficult to justify." Steelman the bull case — InvestingPro Fair Value flags it undervalued, and the dividend still yields something after the cut — but when two analysts slash PTs by 30-40% on the same FCF guidance reset, you don't catch that knife. The r/r is bad until the multiple actually breaks.
Cantor upgrades BTDR to Overweight from Neutral, PT to $18 from $15. The catalyst is one real deal: a 16-year, 121MW AI/HPC colocation agreement at Tydal, Norway. $4.7B in contracted payments over the initial term, $8.0B with the 8-year renewal option. NVIDIA GPUs, Dell as tech provider, Volta subsidiary as the lessee — and reporting names Anthropic as the ultimate off-taker.
THE STOCK IS +27.87% OVER THE PAST WEEK. Market cap $2.77B. Revenue growth 146% LTM. This is the first AI colocation deal Bitdeer has signed with actual favorable economics and real time-to-revenue. Not a land play, not a "we're exploring options" press release. A signed 16-year lease with a named hyperscaler-grade counterparty. Cantor calls it thesis-changing — and they're right to use that word.
"The muted reaction in shares following the deal announcement suggests very little, if any, of the pipeline is being priced in."
That's the bull case in one line. The market has given Bitdeer credit for the mining business but no credit for the AI pivot. Cantor's $18 PT still doesn't embed the remaining unleased Norway capacity or the Rockdale site — those are the next catalysts to watch. Land and lease matters at Rockdale, specifically. Get those signed and this thing re-rates again.
The bear case? Execution. 121MW is a fraction of the total capacity they've talked about. The lease is "subject to certain conditions." Dell, NVIDIA, and Anthropic attached means credibility, but it also means Bitdeer is now a vendor in someone else's supply chain — integration risk, delivery risk, and dependency on the AI capex cycle that could reset if the lab spend pulls back.
For the divergence watch: Citizens initiated at Market Outperform with a $35 PT. That's nearly 2x Cantor's target. Different lens — Citizens is underwriting the vertically integrated Bitcoin mining platform (self-mining hash rate at 70.2 EH/s, up from 13.6 EH/s a year ago; May production +370% to 921 BTC). Two analysts, two very different frameworks. The market is still deciding whether BTDR is an AI colocation comp or a mining hardware story. The Tydal deal says the former. The $36M Sparks, NV manufacturing facility says the latter. Right answer: it's both, and the stock hasn't priced either fully.
Bottom line: the upgrade is deserved but the PT math still lags the narrative. The re-rating comes when the remaining Norway MWs get signed. Watch the filings.
Bull camp isn't flinching. Evercore sits at $3,100 on supply deals + pricing power through 2027. Lynx at $2,450 on hyperscaler capex driving memory/flash demand. 7 analysts raised estimates into the print. This reads like a positioning reset, not a thesis break — but the guide miss says the reset isn't done.
"Salesforce continues to increase revenue and operating margin through advanced software solutions in customer service, marketing automation, and data analysis using AI." — Hans Engel, Erste
But this isn't a clean tape. Raymond James stays Strong Buy at $290, while Morgan Stanley downgraded to Equalweight at $185, flagging a disconnect between AI progress and subscription growth. The VA contract ($1.6B, three-year, Missionforce) gives Agentforce real enterprise credibility — shares +1.3% on that print. Net: Erste sees a skewed r/r at current levels, but the Street's still split on whether AI monetization catches up to the multiple. Watch subscription revenue like a hawk.
Evercore ISI lifts PT to $175 from $135 (Outperform) — and the stock at $147 leaves ~19% on the table. The estimate revisions are modest (revenue +2%/+4%, EBITDA +7%/+8% for FY26/27), but FCF projections got a 14%/13% boost. That's the tell: they're underwriting the margin and cash conversion story, not just top-line growth.
Q2 delivered: EPS $0.42 vs $0.40 expected, revenue $3.58B vs $3.45B, and GMV +32% YoY to $116B — FIFTH CONSECUTIVE QUARTER OF GMV GROWTH ABOVE 30%. This is a scale compounder firing on all cylinders, and the breadth (merchants, regions, channels) argues it's not a one-off pull-forward.
The long thesis in one quote:
"Overall, Q2 results strongly amplified the SHOP Long Thesis," Evercore ISI said. "We view SHOP as one of the few Large Caps in our coverage with a still premium growth outlook, no major capex cycle, and very likely an AI beneficiary on both the merchant-productivity (Sidekick) and demand-generation (Shop App / agentic surfaces) sides."
New PT rests on 45x 2028 EBITDA ($4.75B) and 50x 2028 FCF ($4.66B) — they're paying up for a rare bundle: premium growth, zero heavy capex, and AI upside on both cost-per-merchant and demand gen. The 69x EV/EBITDA multiple (TTM) still gives PMs pause, but with growth at this clip, multiple compression isn't the risk — narrative stall is. Watch the agentic commerce angle. That's where the next leg comes from.
SPCX. THE 10GW TARGET DOESN'T FOOT. Disclosed pipeline sits ~2.4-2.5GW; at ~$40B/GW, the missing 7-7.5GW needs $280-300B nobody has modeled. Q2 capex $18.4B, AI op loss $1.3B, stock -12/-13%. The Rubin exclusivity and $6.7B contracted revenue are real (management says payback in under a year), but this is a merchant GPU cloud with an unlimited balance sheet and a 20% lockup overhang hitting in 48 hours.
NVDA. Demand keeps accelerating. SPCX builds exclusively on Rubin; Melius says that alone could add $100-200B to 2027 revenue. NVDA takes 50%+ of 2026 CoWoS, and TSMC now outsources a front-end packaging step — shipments are capacity-limited, not demand-limited. BlueField-4 and double CX9 per Rubin push 1.6T module ratios from 1:2.5 to 1:5 — connectivity content compounds. The bear case is AMD Helios claiming 30% more tokens per dollar; NVDA defends strategic accounts with price and networking discounts. Absolute growth intact, margin umbrella under attack.
META. The ~400K merchant GPU headline (H100/H200/Blackwell) is the signal. Direct pressure on lower-end neoclouds renting older silicon. Muse Code launching under OpenAI/Anthropic pricing is the first real AI monetization attempt after investor pressure — but the compute sale execution is the swing factor for GPU rental markets.
ETN. "We're just not seeing the slowdown." Operators expand in proven locations instead of greenfield. Favors incumbents with existing site relationships and power agreements. Electrical content stays intact.
TLN. Management talking ~$50/share in FCF and the market doesn't pay for it. Best r/r in the group if that number holds. Cash flow funds the buildout — deep-value AI power, not story stock.
GEV. SPCX targets 15-20GW of power/cooling by end-2027, and management explicitly wants more power/cooling equipment than GPUs. That's a grid-equipment supercycle, multi-year and multi-customer. The binding constraint just got a demand stamp.
VRT. Secondary winner from buildout diversification. Every new operator — neocloud, sovereign, enterprise — needs power distribution, thermal, and electrical infrastructure. SPCX's 20GW plan validates cooling/power demand through the decade.
PWR. 15GW+ of builds missing gensets/UPS creates construction and grid-delivery demand. Electrical infrastructure is the gating item for AI revenue recognition. Same tape, different layer.
APLD. Recurring rent of $44M against $5B in debt. That's the overleveraged neocloud profile. SPCX entering merchant compute compresses scarcity premiums. The bear side of the AI infra trade, and it's winning right now.
NBIS. SPCX's $6.7B contracted revenue validates neocloud demand — and simultaneously rolls in a new entrant with ~$100B liquidity and direct NVIDIA supply. Validated and threatened in the same print. Differentiation matters more than ever.
PLTR. GAPPED 19% ON US COMMERCIAL +149% Y/Y. Market re-rates from consulting to AI infrastructure platform. Now at a discount to DDOG/SNOW/CRWD on EV/EBIT. The re-rating has room.
POWL. RECORD $934M ORDERS, 3.0X BOOK-TO-BILL. But 12-month backlog conversion drops to ~54%. Duration extension, not near-term earnings acceleration. Power supercycle intact, timing pushed out.
MU. THE MEMORY CYCLE IS NOT ROLLING OVER. CXMT refused Apple price cuts and prices LPDDR5X at/above Samsung/SK Hynix — that cements commodity DRAM pricing power and frees Korean majors from low-end clearing. Musk says supply grows ~20% vs demand ~200%. PSMC confirms HBM back-end packaging for Micron — mini line 2026, mass production 2027. Memory is the binding constraint.
WDC. Clean beat-and-raise: adj EPS $3.56 vs $3.31, GM 54.1%, Q1 guide $4.1B vs $4.06B. But the stock trades as a SNDK shadow. The key cross-read: if SNDK's guide softness is consumer-specific, WDC's HDD/enterprise mix is insulated. If it's cycle fatigue, nobody is immune.
AAPL. THE MEMORY SHORTAGE JUST BIT APPLE. CXMT refused Apple's LPDDR5X price-cut demands and prices at/above Samsung/SK Hynix. Apple's scale cannot force price down in a shortage cycle. Margin headwind on the next iPhone, not just a component line.
LRCX. Samsung/SK Hynix testing AMEC Chinese etch as contingency — long-term threat to Lam's memory franchise. Near-term, memory capex and packaging ramp increase etch/deposition demand. All roads lead to semicaps.
KLAC. Process control intensity rises with advanced packaging and memory complexity. The Chinese etch threat barely touches KLAC's franchise. Same memory capex wave, cleaner moat.
ASML. Memory shortage plus the all-roads-semicap trade. Samsung Foundry 4nm full, 2nm inquiries continue. DUV multi-patterning extendibility to 2nm/1.4nm supports Chinese foundry process extension. Gating supplier for leading-edge capacity — full stop.
SITM. STRONGEST PRINT IN THE WINDOW. Q2 revenue $157.4M vs $146.5M bogey, EPS $0.66 vs $0.58, GM 67.1% UP 8.9PTS — that's pricing power, not volume. Management guides 1.6T revenue +100% in 2027 with 800G also growing; combined timing SAM ~$450M. Rack-level sync adds several hundred dollars per rack (CED +181% Y/Y). Customers order 12-18 months out into neocloud/sovereign/enterprise. This is an additive cycle that just got longer.
MRVL. Best-positioned in CXL and a direct 1.6T DSP/PAM4/SerDes beneficiary. Tower-Marvell hit multi-million coherent optical-module shipments — real volume, not design-win theater. The contested piece is fabric switches: Astera's Scorpio X-Series is coming for that merchant layer. Custom silicon and optical DSP carry the story.
AVGO. Core AI networking franchise — 1.6T DSP, PAM4, SerDes, switch-interface. The SiTime read-through supports differentiated content pricing. The $31B guarantee on Blackstone's $36B Anthropic financing is the overhang — one balance sheet carrying massive credit risk. The AI arms race is creating off-balance-sheet contagion.
CRDO. Direct read-through from SITM: rack-level sync adds hundreds of dollars per rack, and CRDO owns key connectivity pieces. Challenges the compute-density-reduces-networking-BOM thesis. The additive cycle helps.
MPWR. SiTime's GM/operating leverage proves differentiated AI infra components command premium economics. MPWR participates in power/analog where performance limits constrain the system. Pricing power holds.
LITE. FCC restrictions on Chinese optical transceivers could structurally reshape supplier lists — accelerated second-source qualification, larger LTAs, better pricing. But non-Chinese capacity is insufficient. The bottleneck moves to InP lasers. This is a trade about module demand, not laser content per module.
COHR. Direct U.S. transceiver beneficiary if the FCC ban lands. Also did a 400G silicon-modulator development with Tower. Low-loss PICs cut laser counts per link, and Coherent sits on both sides of that trade. Policy-driven demand shift is positive.
AAOI. Named as a U.S. optical module beneficiary. The problem: non-Chinese capacity cannot close the gap. Supply shortage creates pricing power if hyperscalers qualify fast. Watch qualification speed.
VIAV. Grouped with SNDK/SITM/TLN as an AI infrastructure print after the close. No specific numbers in the window — treat as a sentiment signal. The market now trades VIAV as an optical/test infrastructure name.
CSCO. Rack-level sync and the 1.6T transition mean networking content grows faster than server units. CSCO benefits if it holds share in an expanding pie. Read-through only — no specific print in the window.
NOK. Acquiring NXP's Arizona fab to convert into an InP semiconductor facility. Vertically integrates optical component supply ahead of 2027/2028 constraints. Smart strategic move — if AI optics demand inflects, NOK's supply chain moat grows.
NXPI. Selling the Arizona fab to Nokia. Frees capital, concedes the InP/optical opportunity. Likely neutral if the price is right — watch where the proceeds go.
MCHP. SiTime post-Renesas is a full-portfolio timing threat — oscillators, resonators, clocks. MCHP already carries inventory/industrial exposure; timing competition is another headwind. Avoid until the inventory cycle turns.
AEHR. Silicon carbide and advanced packaging test remains important. Board member selling $2.5M at $82 is poorly timed but not fundamental. Test demand scales with packaging complexity.
TSM. THE PACKAGING BOTTLENECK IS DEEPER THAN CONSENSUS. TSMC outsourcing a key front-end CoWoS step — previously kept in-house — is the tell. NVDA takes 50%+ of 2026 CoWoS capacity. OSATs get the overflow. TSMC's internal capex response is the next data point.
INTC. Server CPU share up in Q2 on the Granite Rapids supply wave, but JPM calls it one-time. AMD's unit-led server growth is the real negative. $30B capex next year sustains the fight. EMIB good news incoming — packaging becomes a weapon.
GFS. Q2 rev $1.786B +9% Q/Q, GM 29.9%. But Q3 EPS guide 32-42c vs 49c bogey — stock -5.8%. Margin expansion real, EPS guide horrid. SiPh momentum caps on supply constraints; real upside waits for the 2028 CPO inflection. 1H capex already exceeded 2026 capex.
IFNNY. THE NEW AI BOTTLENECK. Q3 rev €4.172B ATH +13% Y/Y; AI datacenter revenue >€1.6B FY26, up from €250M FY24 — >10x in three years. FY27 projection of €2.5B+ gets raised "materially" in November. Power semi TAM raised to $48B. Accelerating content story, not auto recovery.
ASX. TSMC placing additional CoW volume under license. ASE/Amkor/SPIL produce; equipment orders for laser processing and bonding underway. Structural capacity unlock for AI packaging.
AMKR. Incremental CoW work from TSMC outsourcing plus Intel EMIB good news. Advanced packaging is now multi-customer, multi-node. AMKR's capacity investments pay off.
CLS. SITM says AI infrastructure demand broadens into neocloud/sovereign/enterprise/captive; new OEMs/ODMs order 12-18 months out. Celestica is the direct beneficiary of outsourced design and manufacturing. Diversifies beyond hyperscalers.
JBL. Same read-through: new OEMs/ODMs bring AI infrastructure demand. Jabil benefits from outsourced design and integration. Watch demand quality — non-hyperscaler customers carry more credit and volume risk.
DELL. Benefits where enterprises, sovereigns, and smaller clouds choose complete systems over custom infrastructure. Component inflation and expedite fees pressure GM, but Dell passes costs in a constrained environment. Server demand is broader than AI GPU buildouts.
SMCI. Same complete-system demand trend. Exposed to component inflation and expedite fees. If enterprise AI orders decelerate, margin risk is high. But assembly is the current bottleneck — near-term pricing power.
HPE. AMD says 230+ Turin platforms available across HPE/Dell/Lenovo/SMCI. HPE is a share beneficiary of AMD's server CPU ramp. Enterprise server GM stays under pressure from component costs.
WDAY. Field work is "pervasively worse." The source's words: "Nice knowing you WDAY, NOW and FDS." AI cannibalizes seat-based software models. Revenue growth may hold, but the multiple compresses — that's the trade.
NOW. Per-employee pricing sits directly in the crosshairs of AI-driven workforce deflation. Multiple risk more than earnings risk. The SaaS bear case is crystallizing.
FDS. Same bear basket. The market wants AI-native data/workflow tools, not legacy desktop terminals. Field deterioration confirms the thesis.
XYZ. Block cut 40% of employees using AI and delivered a beat-and-raise. Stock flat/down — the market asks if this is one-time firing or sustainable leverage. GP $3.17B +25%, adj EPS $1.02. Need revenue growth to re-rate.
DUOL. Horrid guide — Q3 revenue below street, stock -12%. Management chose engagement over monetization. The market doesn't pay for optionality; it wants guided bottom line.
APP. Q2 sales slightly below, Q3 guide soft. EPS $3.76 vs $3.75, rev $1.92B vs $1.94B. Ad tech momentum cooling. The durability of the AI ad engine gets questioned.
UBER. Revenue in-line, Q3 guide below street, stock -5%. Profitability timeline patience fades. Delivery/travel demand stays resilient, but the market punishes any guidance blemish.
MELI. Beat consensus, but the spending plan spooked analysts. Fintech infrastructure investment compresses near-term margin. Growth strong — the market wants margin discipline.
DASH. Profit forecast topped, heightened spending warning. Stock not rewarded. Same pattern — spend-to-grow is out of favor. Deliver margin guidance, not just growth.
EBAY. Revenue $3.13B, lifted outlook, Depop close came in $200M higher than originally announced. Authenticated luxury and refurbished categories drive high-value buyers. Niche but improving.
Z. Revenue +18% Y/Y to $772M; rental is the bright spot. $36M restructuring cost drove a net loss. Market gives patience for rental structural growth. Profitability still a work in progress.
EXPE. Boosted full-year outlook; consumer travel demand not decelerating. Positive signal in a tape where every guide gets scrutinized. Online travel remains the cleanest consumer internet subsector.
NYT. -13% as subscriber growth slows. The market is extremely sensitive to growth premium in media. Paywall works but matures. No margin offset visible.
ETSY. Laying off 12% (~220 people). CEO denies cost-cutting or AI-driven — cites coordination speed. The market lost patience with the turnaround. Even with layoffs, the narrative is damaged.
DIS. Parks and streaming both beat; streaming profit +64%; Toy Story 5 strong. Ad sales +3% — linear TV drag persists. The beat is real, but ads stay a structural problem.
AMZN. Hyperscaler OCF growth 28-32% Q/Q. Cash flow funds AI capex — the earn vs burn debate lands on the earn side. Watch AI revenue disclosure; it matters more than AWS growth now.
FLTR. Earnings miss plus leadership change. FanDuel's US share dominance erodes. Online gambling competition intensifies. Share story, not market story.
HOOD. Prediction markets revenue $156M in Q2, 10x Y/Y. New high-margin revenue stream from event-driven retail activity. Bullish take rate.
AXON. Q2 EPS $1.88, revenue $904.3M. Lands in a tape where the software/defense premium holds. Need guidance for durability.
GOOGL. THE WORST AI BRAIN DRAIN FROM A HYPERSCALER EVER. Jeff Dean (27 years, 25 people to 190K) and Oriol Vinyals out; Hassabis steps back; Shazeer at OpenAI since June. Engineers reportedly keep personal blacklists — "places I will never work at for any price." Plus the Anthropic backstop: 2.4GW TPU capacity and $44B of lease guarantees. Google becomes the balance sheet, not the brain.
BX. Assembled a $36B Anthropic financing with Broadcom guaranteeing $31B. Credit enhancement turns junk into investment-grade paper for pension funds. This is the template for AI infrastructure finance. Scales, but concentrates risk.
TSLA. SpaceX bought $329M YTD of Tesla Megapacks for xAI data centers. Revenue for Tesla, formalized cross-subsidization within the Musk empire. Governance optics bad for SpaceX public shareholders; for TSLA it's a new energy storage customer.
LLY. Q2 revenue $23.0B +48% Y/Y, adj EPS $8.38. Mounjaro $9.9B, Zepbound $4.9B. GLP-1 demand still accelerating. Monopoly-ish growth — not TMT, but the tape keeps it in the premium basket.
GLW. Polysilicon Section 232 tariff/price-floor analysis puts Corning as the highest-quality U.S. beneficiary via Hemlock Semiconductor. Semi-grade poly production costs up to 30x solar-grade; Hemlock+Wacker ~75% of non-Chinese supply. The policy architecture decides the size of the prize.
SNAP / DVLT / AEIS / PLPC / ZBRA — Nothing in the window. No prints, no tweets, no channel checks. Dead tape is a signal too — rotation is elsewhere. PMs, don't force it.
Merchant compute & neoclouds