Good morning. The positioning squeeze got its second confirmation overnight — Asia AI complex rallied across the board, Nasdaq futures higher, and SPCX +8% PRE-MARKET after a +161% revenue print with the AI segment now bigger than rockets and Starlink combined. When one print lifts the entire complex, that's a market looking for reasons to get back in, not repricing fundamentals. T AT&T beat and raised (fiber the standout, "shrugs off SpaceX fears" per the tape), and SMCI's Monday preliminary continues to ripple through the neocloud complex — NBIS, IREN catching the beta while CRWV, CORZ, APLD keep wearing the credit stain.
Asia handed us the read we asked for: Taiwan and Korea absorbed the rally, TSM bid, and memory stays hot with Samsung upgrading Xi'an NAND lines and MU's Taiwan strike risk basically dismantled (Taichung signed, Taoyuan a symbolic one-dayer). Macro is the one fly in the ointment — OIL ~$110 with Hormuz mines blocking tankers, UK CPI 3.5% cementing a BOE hike, France 10Y above 5% for the first time since 2002. Rates are the clock ticking on every neocloud refi.
Four themes framing the day. ONE: it's Meta's verdict night — the whole week has been pointing at this print, and a beat alone doesn't cut it. The market needs 2026-27 capex discipline PLUS monetization detail; beat-and-raise-capex gets vetoed by the credit side all over again. TWO: power is now official management-speak — Microsoft's UK CEO confirmed on the record that Nscale died on grid connections, which makes GEV, ETN, VRT and the turbine chain the cleanest way to play the bottleneck (suppliers too scared to expand = the durable kind of shortage). THREE: demand keeps supply-constraining, not demand-constraining — SpaceX's 440k Vera Rubin cluster plan and full-rack testing as the output bottleneck says buy the picks-and-shovels (TER, FORM, AVGO, CRDO, COHR) alongside NVDA. FOUR: the market is separating energized capacity from announced pipelines — DayOne's halved IPO and NextDC's implosion vs. SpaceX's sub-1-year paybacks. Own commissioned assets, fade narrative names.
We'll hit up META, NVDA, and the SPCX/TSLA merger-study circus first, then get to memory and the power complex...
NFLX $68.69, down 45% over the past year, sitting 6% above its 52-week low of $65.08. This is a broken chart and the Street just spent the morning telling you why the bounce isn't coming soon. Print was fine. Guide was not. And the ad narrative — the thing that was supposed to be the second act — just got pushed out again.
Q3 revenue of $12.87B, +12% y/y, +11% FXN — in line, nothing heroic. That's the SLOWEST GROWTH PACE IN NEARLY THREE YEARS, and the decel is sequential and real: 14% FXN in Q1 → 12% in Q2 → 11% in Q3 → 11% implied in Q4. Operating margin of 34.6% beat, but the beat was expense timing, not opex discipline — read the footnotes before you celebrate it.
The damage: Q4 revenue guide of $13.4B (+11.5%) vs consensus 12.2%. FY26 ad revenue cut to ~$2.7B FROM ~$3.0B — the second trim on a business that was supposed to double this year. Management says delays are "months, not years." We've heard that movie before.
Oh, and buried in the print: Peter Rice replaces Bela Bajaria as Chief Content Officer, plus a 6% RIF. That's not a tweak. That's a content strategy admission.
Ten shops weighed in, and the message is remarkably uniform: cut PTs, hold ratings, nobody's pounding the table either way. PT range now $71-110, with most of the Street clustered at $77-90:
"Recent reductions in force and content chief changes demonstrate decisive action by the company." — Morgan Stanley
"It will take approximately two years to fix Netflix's $20 billion content slate, with material content write-offs expected before then." — Needham
"Management discussed slower ad product rollouts rather than demand, suggesting a delay rather than disruption." — Bernstein (the bull pushback — but note Bernstein still cut its PT $15 and sees no positive catalyst before year-end)
New and incremental: the ad cut to $2.7B (that $2.7B implies ~80% y/y growth — healthy in a vacuum, but the bogey was a double); the Rice-for-Bajaria swap, which tells you the board agrees U.S. content execution is the problem (Sarandos said the quiet part out loud — the slate needs more breakout hits and talent is taking deals elsewhere); and the 6% RIF with no quantified savings. Also new: management explicitly deferred ALL 2027 commentary to the Q4 call. That punts the most important debate — whether double-digit 2027 growth is credible — by 90 days while estimates bleed lower in the interim.
Already known: UCAN decel, ad ramp slower than hoped, YouTube winning the viewing-time war with lower content costs. The Needham structural point — NFLX vs YouTube/Disney/Sky on cost structure and revenue diversification — is the bear framework gaining converts, not a fresh idea.
This is a show-me story until the Q4 call, full stop. The core sub business is fine — churn improving through price hikes is the single best data point in the whole print, and it means pricing power is intact. But the multiple doesn't re-rate on "core is fine." It re-rates on ads inflecting or content hits returning, and management just told you both are 2027 stories at best. R/R at $68.69 with a $71-110 Street range and a live $57 bear case is roughly balanced-to-slightly-negative until we get 2027 guidance. If you own it, you're owning the Q4 call as a binary. If you don't, there's no reason to be early — Bernstein's point stands: no obvious positive catalyst before year-end.
Read-through: watch the 2027 consensus ad number (~$5.2B) — Guggenheim just called it $4.0B and nobody's defended the high side. If that estimate capitulates, the "cheap on PEG" argument gets a lot thinner. Second-order: NFLX's content reset and talent-relations problem is a quiet positive for the studios feeding it licensed content, and the YouTube comparison keeps strengthening the GOOGL narrative on the other side of the CTV wallet.
This is the most important print in TMT right now, full stop. Q3 smashed — rev $14.6B vs $12.9B consensus (+161% y/y), adj EBITDA $8.8B vs $6.6B (+306%), EPS $0.35 vs $0.13e. But the quarter isn't the story. The story is the Q1'27 guide of $33-37B vs $23B consensus — a ~50% above-Street guide — and Musk putting a $500B-$1T ARR bogey on year-end 2027 when the Street's 2028 estimate sits at $186B. Stock at $147.79, still ~25% below its post-IPO peak close and well off the $225.64 52-wk high. The market does not believe the ramp yet. The buyside debate is whether that gap closes via the stock or via Musk's credibility.
JPMorgan threw the kitchen sink at it — Anmuth to $500 from $240 (yes, +108% raise) on 20x 2028 GAAP EPS of $24.41, modeling 10GW end-2027 compute and $254B 2027 revenue. Morgan Stanley holds OW/$300 with AI revenue now modeled at $146B for FY27 (from $81B) and capex ramped to $278B. Goldman to $245, UBS to $230, RBC/Wedbush at $225, Mizuho $205, Stifel $190. KeyBanc the lone holdout at Sector Weight. Full range runs $118 to $800 (Raymond James, off the Starship FAA filing), consensus implying ~51% upside. Translation: everyone raised numbers, half the street still won't underwrite Musk's bogey.
The incremental stuff, because the "AI compute is big" narrative was known: (1) SI (the rebranded AI segment) did $7.6B rev / $5.0B EBITDA in the quarter — a 65% segment margin — vs $1.1B EBITDA in Q2. That's not a ramp, that's a step function. (2) Compute path raised to ~13GW cumulative by end-2027 from 5-10GW, backed by a 20GW power/cooling buildout, with Musk flagging upside to 20GW on power visibility. (3) Spot pricing inflecting — incremental contracts now above $50/watt, UBS sees payback periods compressing to ~6 months. (4) Q4 capex guide of $35-40B vs $19.6B in Q3. UBS pegs the 13GW build at ~$350B of spend. (5) SpaceX/Tesla special committees formed on a potential combination — Musk recused, "merit but not inevitable." (6) Starship Flight 14 done, ship+booster catch attempt next month; first revenue-generating Starship flight delivered 26 V3 sats with ~10x Falcon 9 downlink capacity.
Bull: Demand is supply-constrained, not demand-constrained — Musk says they're "hounded by every hyperscaler" and could sell out capacity several times over. SI economics are already proven at scale (65% EBITDA margins), connectivity accelerated +69% y/y with ARPU up 5% q/q to $69, and the company flipped to $4.4B net income from a $1.1B loss. Stifel's line captures the bull framing — fundamentals are tracking more than a year ahead of the company's own pre-IPO targets, and the market can't size the earnings power. Buy the disbelief.
Bear: $278B of FY27 capex (MS estimate) against negative levered FCF today means the funding question never goes away — CDS costs on AI-linked debt are already rising, and Wedbush explicitly wants detail on how the Apollo-led debt fits the plan. KeyBanc's point stands: FCF stays pressured as far as the eye can see. Then there's the Musk bogey problem — Stifel itself notes the gap between Musk's ARR target and Street 2028 estimates equals Amazon's entire LTM revenue of $776B. Execution, capital, power, and a potential Tesla merger overhang all sit between here and $500. RBC captures the consensus posture: estimates raised substantially, multiples cut to compensate.
"The market will remain skeptical of the steepness of the ramp, but the direction of travel is hard to dispute... acquire shares while the market struggles to size the earnings power of the company." — Stifel
"Hounded by every hyperscaler" to lease compute, with demand that "could sell out several times the available capacity." — Musk, on the call
JPMorgan acknowledges "substantial execution is required" — buildout, capital, and sustained demand all have to land.
SPCX is now the purest public proxy for the AI infrastructure supercycle — more levered to compute leasing economics than NVDA is to GPU units at this point. The $466B of AI-linked debt issuance in 2026 (vs $101B in 2025) and rising CDS costs on SPCX/ORCL/NVDA credit tells you the market is starting to price funding risk into the whole complex, not just this name. If SPCX's payback-period math (6 months, per UBS) holds up across the group, the neocloud/hyperscaler capex debate resolves bullish. If power or funding slips, this is the first domino — it has the most aggressive buildout and the least margin for error. Watch Starship catch attempt next month and lock-up expirations (Goldman flagged flows) as near-term catalysts. Positioning-wise: if you missed NVDA at the inflection, this is the market offering you a second chance at the same trade with 10x the execution risk. Size accordingly.
Zoomtopia did its job. The street woke up to a fresh Oppenheimer upgrade this morning — Outperform from Perform, $130 PT — and the bull cluster is now stacked at $120-130 while the skeptics sit at $104 with Cantor. Stock at $97.46. The bogey bulls need to clear is ~34% upside to the top of the range, and honestly the math doesn't look crazy given the setup.
This is a crowded post-Investor Day tape. PT range runs $104-130:
Bull: ZM just put out its first-ever long-term guide — 6-8% revenue growth by FY30 vs the 4% consensus bogey, with operating margin lifted to ~40% from 33-36% and GAAP op margin at 28-30%. Oppenheimer came back from Zoomtopia converted, pointing to the "system of action" pitch (Conversation, Context, Completion) and real uptake in the AI platform growing ARPU and retention. Mizuho's key point: the acceleration relies mostly on products already proven in market — not vaporware. Underneath it all: 78% gross margins, a 9x P/E, $28.4B market cap, and churn data that says AI adopters churn 70% less. Oh, and a possible ~$3B Anthropic stake that StoneX thinks gets a fair-value bump if the Anthropic IPO prints.
Bear: Citizens won't get off the sidelines. Their read: FY30 targets assume enterprise growth accelerates to double digits from 7.5% in FQ2 FY27 — the fastest pace in three years — which is a big ask mid-transition from single-purpose app to platform. Competition from MSFT Teams and NICE stays brutal. Cantor's $104 Neutral echoes the caution: they like the Enterprise acceleration and churn reduction, but expect stable-to-flat Online performance through FY30. Translation: the growth story is all enterprise, all execution.
"integrating communications with AI applications should deepen customer relationships and raise the growth rate to double digits on stable margins" - Oppenheimer
The upgrade itself (Oppenheimer flipping is the incremental signal). But the real newness is the framework shift: first long-term guide ever, the Unified Context Layer positioning, the margin raise to ~40%, and explicit capital-return language with the CEO signaling patience on M&A. The Anthropic stake math (~$3B per Citizens) is also getting fresh airtime — that's a free-ish option on a 9x P/E stock.
This is the legacy-comms-platform-into-AI-agentic-workflow trade, and ZM is now the poster child alongside what CRM has been trying to sell for two years. If ZM can actually get credit for a 6-8% top-line algo at 40% margins, it pressures the whole "COVID beneficiary left for dead" cohort to re-rate. Watch MSFT reaction in Teams bundling commentary — that's the swing factor for the bear case. R/R at $97 with a $130 top-of-street and a $104 floor: bulls have the better setup, but this needs enterprise acceleration to show up in prints, not just at conferences.
Salesforce is taking Braze out at $47.50 in cash. That's the whole ballgame. $5.66B EV, 58% premium to the pre-deal close of $30.04, and the stock ripped 61% on the week to $45.57 — sitting ~4% below the offer. The sell-side responded the way the sell-side always does when a deal prints: UBS, Mizuho, and TD Cowen all moved to Neutral/Hold and parked their PTs at exactly $47.50. Not analysis, just math. The rating actions tell you nothing new — what matters is where the Street disagrees on what happens from here.
TD Cowen is the interesting voice. They're holding out for optionality:
"The firm expects the proxy statement to reveal a competitive bidding process from other strategic buyers."
Mizuho took the other side — called the price "a strong outcome for Braze shareholders" and said flatly they don't expect another buyer to emerge. That's your bull/bear in miniature. The bull case is a bump via proxy-season drama or a second strategic (the INFA comp TD Cowen flagged — CRM paid ~5.1x TTM sales there vs ~6.8x for BRZE — suggests CRM stretched to win this one, which means they wanted it). The bear case is you're clipping ~4% into a deal that doesn't close until CRM's fiscal 2028 with a long regulatory and shareholder runway.
Valuation context: ~6.2x CY26 sales ($913M bogey), ~5.3x CY27 ($1.07B). Full for a company with Braze's growth decel, but strategic value is strategic value — CRM gets real-time engagement and AI decisioning to bolt onto Customer 360 and Agentforce, and TD Cowen sees low antitrust risk.
Our read: r/r here is thin. ~$1.93 of upside to the offer against deal-break downside back to the low $30s (where the fundamental guys had PTs — Citizens at $35, Cantor at $38, DA Davidson at $40 tells you where fair value lived pre-deal). The FY2028 close timeline means the annualized return on the spread is underwhelming unless a bump shows up in the proxy. If you want to own the arb, you're betting on TD Cowen's competitive-bidding call. Not sure we'd pay up for that ticket.
SMCI preannounced a monster Q1 and the sell-side came in behind it this morning. This one has real legs. Revenue of $18.5-19.5B — 71% Q/Q, ~280% Y/Y at the midpoint, BLOWING THROUGH prior guide of $15.0B — plus gross margins of 12.0-12.5% vs. 10.4-10.8% guided (favorable mix doing the work). That's the margin print bears said couldn't happen while the top line was ripping. It happened.
Rosenblatt went to $57 from $51 (Buy), pushing FY27 revenue to $80B from $68.5B and EPS to $6.40 from $5.13 — and they think consensus isn't done moving:
"Consensus fiscal 2027 earnings per share estimates have risen from $3.17 to $4.34 over the past three months. Rosenblatt expects estimates to move at least $1 higher following the preannouncement."
Needham held at $46 (Buy), 7x their new CY28 EPS, leaning on the $60B+ backlog and next-gen AI ramp. Note the gap in frameworks here — Rosenblatt at ~9x FY27, Needham at 7x CY28 — both cheap-looking multiples because the denominator keeps exploding. That's the story. Even the hawks (Bernstein at Market Perform) walked their PT up to $42 from $37 earlier this cycle.
Bull: $60B in orders from 20+ customers, Vera Rubin NVL72 racks already shipping with DLC, Cisco Secure AI Factory partnership expanding the channel, and the independent review cleared management on export controls. Trading at ~10x P/E with a PEG of 0.1 while growing 280%. The estimate revision treadmill (15 analysts raising, per InvestingPro) is the fuel.
Bear: Margin beat was mix-driven, not structural — LTM gross margin is 10.8%, so one favorable quarter doesn't rewrite the model's low-margin DNA. Export-control investigations remain ongoing even with the board review clearing current management (an overhang, not a verdict). And at some point the $60B backlog has to convert — last quarter already showed delays in large customer projects clipping revenue.
Net: The r/r tilts long while estimates are chasing the print upward. Watch the first call under new CFO Mark Long — guidance credibility with the street is the next bogey. If revisions stall at $4.34 and the multiple stays pinned at 10x, you're renting, not owning.
CRM stays on the right side of the Street, but the Braze deal just gave bears fresh ammo. Consensus sits in a $275-315 PT cluster — Evercore at $290, Cantor at $300, Citizens at $315, Oppenheimer at $275 — with BofA recently upgrading to Buy from Underperform and lifting its bogey to $285 on AI platform strength, effectively retiring the "AI disintermediates SaaS" bear case. The debate now isn't whether the core works. It's whether Benioff can stop buying things.
The Braze acquisition is the flashpoint. Evercore flags the timing — right on the heels of the Fin deal — and the 80% PREMIUM TO THE 30-DAY VWAP as unusual absent a competitive process. Their concern is real: more M&A muddies visibility into core organic growth precisely when the tape is rotating back toward clean secular growth stories with strong FCF. Their sharper observation — NOW (ServiceNow) is the flow beneficiary of CRM's deal appetite. That's a relative value tell, not a rating change.
Cantor takes the other side and makes the cleanest bull case on the deal itself:
"The acquisition helps jump start the revival of Marketing Cloud and adds more value to the entire CRM ecosystem at a reasonable acquisition price."
Fully funded out of FCF too — $15B+ LTM LEVERED FCF, an 8% FCF YIELD on a stock at 22x earnings and a 0.36 PEG. That's the crux of the bull argument: even if you hate the M&A, the cash machine absorbs it.
Bull: AI offerings (Fin in service, Breeze in marketing) are landing inside existing footprints with "universally positive" Dreamforce feedback, Marketing Cloud gets a genuine revival catalyst via Braze, and the whole thing self-funds. Citizens' customer spend checks and Oppenheimer's Dreamforce survey both show AI budgets shifting from adoption to ROI — good for the incumbent with the data moat.
Bear: Serial M&A at fat premiums kills the GARP re-rating story. Every deal defers the clean "organic growth + margin + buyback" narrative that gets this multiple expanding, and the 80% premium on Braze suggests either a contested auction or discipline issues. Neither is great.
Net-net, constructive. The fundamentals case keeps strengthening and valuation is cheap against FCF, but the stock trades on narrative — and "CRM does another deal" is the one narrative that keeps multiple expansion capped. Watch whether management signals a digestion period. If they do, the path to the $300+ PTs opens up. If another deal drops before year-end, expect the Evercore relative-value call (long NOW vs CRM) to get more airtime.
Roth/MKM reinstated APLD at Buy with a $35 PT — 81% upside from $19.33, and yes, the stock is DOWN 38% over six months, so there's plenty of room. The thesis: execution certainty from secured equipment plus a repeatable "AI factory" design gives you asymmetric upside the market isn't pricing. Revenue grew 218% LTM. The shift from signing leases to actually delivering capacity is doing the work here.
The note spends most of its energy swatting away the three bear bogeys — financing risk on remaining builds, headline yields below peers, and bipartisan NIMBY opposition to data centers. Roth/MKM's counters: investment-grade offtakes broaden capital access, yields are at market with upside embedded in contract structures, and the ND/LA/AL footprint dodges contested markets.
"Record, together with the company's land position and power agreements, gives it a stronger path to scale than the market recognizes."
Context PMs need: the rest of the Street is WAY above Roth/MKM. Texas Capital at $71, Needham at $70 (cut from $83 on regulatory risk), Citizens at $60, Wells at $55, UBS at Buy flagging a path to 3.5-4.0GW by 2030 vs. 1.4GW leased today. So a $35 reinstatement reads conservative — almost a soft tell that the risk case has teeth. Also flag the FQ1 print: $341.9M vs. Needham's $90M bogey, but largely one-time fit-out services, so don't extrapolate the beat. Net: the bull case is intact and cheap, but the dispersion in targets ($35 to $71) tells you nobody has conviction on the right multiple for this model yet.
RBC goes Underperform on MBLY, PT to $5 from $7 — and the thesis is structural, not cyclical. Tom Narayan's call: base ADAS gets displaced as OEMs shift to centralized compute, and the premium products (SuperVision, Chauffeur) have a customer concentration problem, namely near-total dependence on VW — a customer RBC thinks actively walks away. Incremental premium wins haven't shown up. Stock at $7.30, DOWN 51% over the past year, -30% YTD, and RBC still sees downside. The kicker is the framing:
Mobileye increasingly resembles a mature hardware supplier rather than a software-driven autonomy platform.
Even after the de-rate, it trades at a premium to auto suppliers — which is the multiple compression case in one sentence.
Street is genuinely split. Post-Q2 (strong print, EPS bogey of $0.48 for the year, first profitable year), Berenberg went to Buy/$11 calling mid-term growth mispriced, Piper went OW/$12 on the robotaxi push, Tigress at Buy/$18. Barclays and UBS sit at $9 on lack of advanced wins and the Shashua departure overhang. So you've got a $5-$18 range on a $7.30 stock — that spread IS the story.
R/R: bears need VW defection to confirm; bulls need one premium win to break the "hardware supplier" narrative. Until one of those prints, this is a show-me story with a leadership transition on top. Not a short I'd press after a 51% drawdown, but zero reason to catch it either.
Mizuho stays Outperform with a $2,050 PT into the Oct 29 print, and the call here isn't really about the quarter — it's about 2027-28 NAND tightness. Stock at $1,365, off 2.7% this week, but up a silly +815% over the past year. Buy the pullback, per the desk.
The thesis in one line: supply stays flat ex-China while AI demand (KV cache, agentic AI, enterprise SSD ramps at META/GOOGL/ORCL) accelerates into 2027-28. The real kicker is the contract structure — non-binding MOUs with eight datacenter/edge AI customers representing ~$94B in minimum revenue floor, with half of bits on MOUs through FY27 and two-thirds into FY28. That's visibility this industry has never had.
"Would be a buyer on the pullback with NAND demand accelerating driven by KV Cache and Agentic AI while supply remains contained."
Context for the r/r: memory prices already ripped ~20% in Q3 on shortages (Bernstein), Citi sees undersupply through 2031, and Moody's upgraded to Ba1 after SNDK killed its funded debt. RBC sits at Sector Perform / $1,600 — the lone "it's run a lot" voice. Fair. After +815%, the bogey for Oct 29 isn't beat/raise, it's whether management validates the MOU-to-binding conversion path. If they do, Mizuho's $2,050 math (50% upside) starts looking less crazy than it reads. Watch the Kioxia XL-Flash / SLC mix commentary too — that's the wildcard for 2027 tightness.
Piper goes OW from Neutral on DOCN — but here's the twist — they CUT the PT to $126 from $142 while doing it. Classic "valuation reset, story intact" move. The stock has pulled back ~9% over the past week after a monster run (still UP 150%+ over the past year), and Piper says the dip more than compensates for near-term margin and FCF pressure.
The thesis: AI as an acceleration driver. Conservative ARR assumptions, middleware optionality, combined platform leverage, and direct exposure to AI-native startups. Piper even floats a 2027 ARR guidance raise as a catalyst.
"The recent pullback in shares outweighs likely near-term pressure on margins and free cash flow estimates."
Bull case has company — Truist (Buy, ease of use in AI apps), Citizens (Outperform post-Open Intelligence Summit), KeyBanc (OW, $200 PT), Stifel (Buy on the Advanced Edition database GA). That's a crowded long. The bear wrinkle Piper itself flags: DOCN needs additional capital. After a 150% run, raise risk on an AI story trading on acceleration hopes is the real r/r question — not the margin bogey.
Cantor is standing by the OW and $109 PT with the stock at $26.97 — down 13% on the week as the NEXTDC M3 Melbourne power-delay story rattled the AI-infra complex. Their read: this is noise for SHAZ specifically. The M3 delays hit expansion power only (existing ops untouched, no new power work until Feb 2027), and SharonAI's exposure is a rounding error — ~2MW operating at M3, nothing else contracted there or in the pipeline. The contract that actually matters — the NVDA 72MW deal — sits at M2, 11 miles away, and NEXTDC said nothing about M2 delays on the call.
Here's where I'd push back on the Cantor defense, though. The M3 story is a distraction — the Q2 print was genuinely horrid. Adjusted loss of $26.16/sh vs. the 47c bogey, revenue $1.93M vs. $7.54M expected. That's not a miss, that's a different company than the model assumed. Yes, contracted demand is the bull case — TCV up to $8.8B YTD from $2.2B at the start of Q1 — but the gap between contracted TCV and actual revenue is the entire debate here, and the print widened the skepticism gap, not closed it. Add NVDA pausing its revenue-sharing program with AI cloud providers (antitrust overhang, per reports) and you've got a name where the narrative damage is cumulative even if the M3 headline is overdone.
Stock's still ~14x off the $1.90 52-week low, so the market is paying for conversion of that $8.8B pipeline. At $27 vs. a $109 PT, Cantor is betting the M3 fear passes and execution shows up. R/r is tempting if you believe the contracted power converts — but this is a show-me story now, and February 2027 is a long time to hold a headline-risk name on faith.
"SharonAI currently has approximately 2 megawatts operating at the M3 site. The company has no other exposure to M3 in what it has already contracted to customers or in its total pipeline of contracted power."
StoneX reiterates Buy, $1,800 PT — but the real message is the setup, not the rating. MPWR trades at a P/E of ~77 and the multiple has contracted for six months. The bull case hinges on one variable: data center capex, which drives 35-40% OF REVENUE but is the leading growth contributor (LTM revenue growth 29%). In other words, the stock is a levered proxy on whether hyperscaler capex holds up — and we find out in the next 10 days.
"For Monolithic Power Systems' valuation multiple to recover from the contraction witnessed over the past six months, investors will need to grow more confident in the current data center capex environment."
That's the whole debate. META reports this week, GOOGL/MSFT/AMZN next week, then MPWR's own print OCT 29. The Street is already positioned bullishly — Truist at $1,889, KeyBanc at $2,100 on Enterprise Data +160% Y/Y, and the company just hit its $4B revenue capacity target and raised to $6B+ (long-term $8B bogey). Estimates likely drift up post-print if 2H delivers.
R/r here is binary-ish into a catalyst stack. If capex commentary from the hyperscalers stays constructive, the multiple re-rates and the $1,889-2,100 cluster looks reachable. If any of META/GOOGL/MSFT blink on capex, 77x earnings has no floor. Positioning call: this is a name you size around the prints, not through them.
Street is uniformly bullish into the Oct 28 FQ1 print — Guggenheim reiterates Buy at $586, and the PT cluster has marched UP to $600-675 (UBS $675 from $525, BMO and StoneX at $600, Piper at $610 looking for Azure +45% y/y). Consensus sits at 13% upside. Nobody's fighting this tape.
But read Guggenheim's segment detail carefully — the conviction is really only Azure. M365 Commercial Cloud has LIMITED upside with some guide risk, and Windows OEM & Devices carries risk to BOTH numbers and guidance. Segment reorg muddies things further (folding higher-growth businesses into M365 makes that line look better optically, but gives cleaner Azure read — net positive for transparency).
The real bogey is capex. MSFT reports early among hyperscalers, so its capex commentary becomes the AI demand tell for the whole complex. And here's the tell on sentiment: MSFT CDS AT RECORD HIGHS as investors hedge AI-linked debt exposure. Market loves the growth story but is quietly paying up for insurance on the spend. Watch the capex guide, not the beat.
"Microsoft will be closely watched for capital expenditure commentary as an indicator of overall artificial intelligence demand."
PLTR -7% OFF HIGHS INTO THE NOV 2 PRINT, and StoneX's Hold says the quiet part out loud: at 41.6x 2027 EV/REV and a 166x P/E, a beat is table stakes. The bogey isn't the number — it's the magnitude of the raise and what it implies for Q4/FY26 trajectory, NRR, and large-deal flow. Stock's still +54.5% since the Q2 print in August, so positioning stays long and expectations stay fat.
"Investors will focus less on whether Palantir beats estimates and more on the size of any upside."
That's the whole ballgame. U.S. Commercial and Government demand aren't in dispute — the r/r debate is purely valuation versus growth durability at scale. The bull camp keeps pressing (GS to Buy with a $230 PT, Yorkville Ives and Baird both Outperform at $250 on enterprise AI deployment and the open-weight shift), but $230-250 targets against a $195 stock aren't exactly pounding the table. Nobody's underwriting much multiple expansion from here — they're underwriting execution.
One wrinkle worth a glance: HSBC AM weighing a PLTR removal from its $9B Islamic index fund over pension-scheme pressure. Not thesis-changing, but a reminder the ESG/government-exposure overhang never fully leaves this name. Into the print, this is a "how big is the guide raise" trade, not a "do they beat" trade.
The TPU story is now the bull case. Full stop. Mizuho reiterated Outperform with a $460 PT (35% upside, rolled to 2028) and the call isn't really about search anymore — it's about custom silicon. They're projecting $100B IN TPU REVENUES BY 2028 and pushing Google Cloud estimates 11%/9% ABOVE CONSENSUS on 2028 revs/segment op income. Their line: upside from TPUs not in consensus outweighs the downside risk to core search.
"Mizuho sees more tangible upside from the TPU business not reflected in consensus estimates versus downside risk to the core search business."
The interesting wrinkle — they're cutting the multiple while raising the target. Down to 23x from 30x, citing cyclicality, growing search risk, and group multiple compression. And their own search estimates sit BELOW CONSENSUS through 2028. So this is a "search is a melting ice cube, own the compute" call dressed in an Outperform. R/R skew: +62%/-31%, bull $550 / bear $235.
Worth flagging: the article cites a 17x P/E and 0.15 PEG — take those with a grain of salt, they look like stale data vendor prints, not something to anchor on. (TD Cowen also raised on expectations Cloud revs more than double in Q3 off that 82% Q2 growth, but that's a recycled data point.)
Positioning take: the marginal buyer of GOOGL here is buying the TPU externalization thesis, not search stabilization. If you're long, you want to see third-party TPU wins in print — that's what closes the gap to $460. If search decel accelerates, Mizuho's own below-consensus numbers tell you the Street's search bogey is too high. Both things can be true, and that tension is the whole debate in the stock.
ALGM down 50%+ from its 52-wk high of $71.77, sitting at $33.89 near the bottom of its five-year valuation range — and StoneX says the market is pricing an estimate cut they don't think is coming. They reiterated Buy, $60 PT, and expect a rally off the October 29 print even on in-line numbers, so long as Dec-qtr guidance shows a real path to gross margin expansion. Consensus is Buy across the board, PTs clustered $40-62, so the Street broadly agrees with the asymmetry — this is the consensus long that's been destroyed, not a contrarian call.
The fundamentals back the setup. Q1 FY27 beat both lines ($0.23 vs $0.21 EPS, $259.24M vs $250.96M rev), sixth straight quarter of sequential growth, sales +27% y/y. RBC also initiated Outperform at $50 recently, leaning on ALGM's leadership in magnetic position and current sensing.
"The firm said it would view any sell-off on continued concerns over auto production or data center capital expenditures as a buying opportunity."
Bear steelman: auto production and DC capex are exactly the two macro bogeys that could miss, and "rally on in-line" is a hope, not a catalyst. But at 5-yr trough valuation with six quarters of sequential growth, the r/r skews long into the print. Watch the gross margin guide — that's the whole trade.
PEGA is a broken growth story right now, and management basically admitted it. Stock DOWN 39% YTD after missing on nearly every key metric — revenue, Pega Cloud revenue, Cloud ACV, EPS, FCF. Pega Cloud ACV grew 20% vs 21% consensus (not a horrid miss in isolation, but it's the guide language that kills you). Management said 3Q YTD ACV growth "did not reflect the opportunity" and flagged that GTM execution issues will weigh on growth FOR THE REST OF THE YEAR. When the C-suite pre-commits to a weak exit, you believe them.
Citizens stays at Market Perform, and the broader tape is messy: D.A. Davidson cut to Neutral with a $32 PT on frontier model competition risk, JPMorgan went to Neutral and trimmed PT to $47 from $54 on execution, while Yorkville Ives bucked the trend with an Outperform initiation at $43 on the workflow automation angle.
"Blueprint appears to provide a structural advantage for Pegasystems."
That's the bull case in one line — plus genuine confidence in Trefler and Stillwell from the Street. The bear case is simpler: ACV growth has slowed significantly, the sales motion is broken, and the frontier-model-providers-eat-the-app-layer narrative has a real ticker to attach to now. R/r here hinges entirely on whether GTM fixes show up in 1H27 prints — nothing to do until then. This is a wait-for-proof name, not a buy-the-dip name.
Deutsche Bank goes to Buy from Hold on Wartsila, PT to EUR37 from EUR34 — implies 33% upside. The stock has been horrid: DOWN 27% over six months, sitting just 5% above its 52-week low ($5.90). That's the whole pitch. DB thinks the market got too cautious on whether Energy order momentum lasts, and the AI-sentiment lull gives you the entry point.
This is a behind-the-meter data center power story, not a marine story. US power availability is the binding constraint on the AI buildout, and DB says Wartsila's engines are one of the few viable bridging solutions before end of decade. Portfolio optimization done, production capacity ramping materially, higher-margin backlog converting next year — they cut 2026 EBIT 3% but RAISED 2027-28 by 5-11%. Classic "right numbers, wrong year" setup.
One caveat: OTC ticker, single broker, and the "record backlog converts next year" line is doing a lot of work. But if you believe the power-shortage trade has legs, this is a cheaper, beaten-down way to play it than the names everyone's already crowded into. Worth a look on weakness.
Fresh initiation from Erste Group at buy, and the pitch is simple: demand is massively outrunning shipments. Book-to-bill at 1.23, revenue up 54% LTM to $29B, and Q3 guided to roughly +51% y/y with EPS growing the same. This is the AI interconnect trade working exactly as advertised — and Bernstein SocGen jumped on last week too with an outperform and $99 PT, calling out APH's copper interconnect leadership in data centers.
"Given the positive outlook, we expect the stock's upward trend to continue."
The pushback writes itself — 40x P/E isn't cheap — but the 0.65 PEG and 38% ROE (vs sector average) make the growth-justifies-multiple case pretty clean. Q2 already delivered the proof point: $1.35 EPS vs $1.16 bogey, $8.8B revenue vs $8.19B expected, +55% sales. With datacenter demand still accelerating and a foldable iPhone 18 cycle (>$2,000 ASP) adding content tailwind, hard to fight the tape here. New money initiation at these levels reads as chasing, but the backlog says the chase is warranted.
Sell-side is split on GLW into the OCT 27 print, but the tell is JPM's framing: the quarter looks like a beat-and-raise vs consensus and STILL not good enough for where the buy-side bar sits. Stock's +5% since last earnings vs optical peers DOWN 1%, management's tracking to the high end and hitting the $20B Springboard revenue target a quarter early (high-teens growth, sustained through '27) — and JPM still cut PT to $165 on valuation, saying the buy side needs something bigger, like a raise to the 2028 Springboard targets and a 20%+ op margin commitment.
"Buy-side expectations would likely need a more material catalyst, such as a raise to the internal Springboard 2028 revenue target and operating margin target of at or above 20%."
The rest of the tape is noisy: Truist upgraded to Buy ($175 PT) on the pullback, Mizuho trimmed to $180 (kept Outperform) flagging capital needs after the $2B equity distribution agreement with GS — dilution overhang worth watching if the AI/optical capex build accelerates. Options pricing a 7.2% MOVE into the print. Also note the Apple all-glass iPhone cancellation (Jefferies supply-chain checks) — stock shrugged it off, which tells you the market's treating GLW as an optical/DCI story now, not a consumer glass story. DCI target of $1B by 2030 could get pulled forward on the AT&T/VZ/Zayo deals. Setup: good quarter, high bar, funding questions. R/r feels balanced — we'd want the Springboard raise before adding.
Baird downgraded ACN to NEUTRAL from Outperform — but raised the PT to $252 from $245, which tells you everything. This is a multiple call, not a fundamentals call. Stock at $239, ~6% to the new bogey. Meh r/r, hence the downgrade.
The bear math: FY27-29 organic cc growth guided to 3-4% vs the 8% average from FY14-23. FCF conversion dropping to 110%+ vs 125% historical. Leverage flips from 0.8x NET CASH to 0.3x net debt. And Baird thinks M&A costs more FCF than it used to just to generate 2% of annual revenue benefit. The AI-services flywheel is real but the organic growth engine has structurally downshifted — that's the thesis.
The counter: the stock trades at 16x NTM earnings vs a 23x FY14-23 average, and the rest of the Street didn't flinch after Q4. Constant-currency rev grew 7% — TOPPED THE HIGH END of 1-5% guide. UBS at Buy/$275, Stifel Buy/$242, Mizuho OP/$241. Even TD Cowen (Hold) only at $201, so the spread is wide. Three-year roadmap targets ~3% organic in FY27-28 accelerating to ~5% in FY29 — bulls are buying the back-half acceleration, Baird is saying prove it.
Net: downgrade is a yellow flag on growth durability, not a red one on the franchise. Investor day Oct 14 (first in four years) is the catalyst — management either validates the FY29 acceleration story or Baird looks early and right. Sitting tight makes sense into it.
The SpaceX merger narrative just got real — and JPMorgan is giving you a framework to size it. Musk disclosed on the SpaceX Q3 call that special committees have been formed at BOTH companies to discuss a combination, and JPM sees "strong strategic logic" in the tie-up. With SpaceX valued at ~$2T (up 10% since IPO, per the note) and TSLA at $1.48T, a 30-35% premium would make this a merger of equals — and critically, would keep Musk above 50% voting control of the combined entity. His June move to lift his TSLA stake to ~20% reads like pre-positioning, not coincidence.
The strategic case writes itself: shared AI infra, the Terafab facility in Texas, cross-purchases (Megapacks, Cybertrucks), and TSLA's $2B in xAI already inside SpaceX. Expect a SpaceX-led all-stock structure if this happens.
Two wrinkles. One, this is a LONG way from a deal — committees are formed, nothing more. Two, the actual TSLA print underneath is mixed: Q3 deliveries BEAT at 486,532 (vs. 461,974 consensus) but production MISSED at 464,391 (vs. 486,761). Sell-side hasn't re-rated on the M&A chatter — JPM stays Neutral at $415, Cantor OW at $485. The merger premium is the bull case; the core auto business still has to earn the multiple in the meantime. Watch for any language on the special committee process — that's the tell on timing.
Small-cap AI infra pivot keeps stacking contracts. DUOT signed a 10-year colocation deal with a "matured cloud platform" for 20MW of IT load — $400M+ of total revenue, and the longest-duration agreement the company has ever signed (prior modular data center deals ran 5 years). Cantor stays OW, $24 PT vs. stock at $5.80 — that's a 4x implied upside, so the street clearly thinks the market hasn't priced this yet.
The pipeline matters more than the single deal. Company says it's negotiating 100MW of additional powered, data center-zoned land across multiple US sites, with 25MW ALREADY UNDER CONTRACT. Stack that on the $500M five-year lease for 55MW announced earlier (which took Cantor's PT from $26 to $27 before this new deal kept it at $24 — yes, the PT walked back, not sure we can read too much into the sequencing there).
One flag: Q2 EPS of $1.61 vs. a $0.02 loss expected looks optically massive, but revenue MISSED at $6.18M vs. $10.7M bogey — that EPS print smells one-time in nature. Don't annualize it. The real thesis here is rate of change on contract wins in a name doing 99% LTM revenue growth, and whether land/power negotiations convert.
Long-duration contracts on a sub-$10M-revenue company = binary-ish small cap. Position size accordingly.
Raymond James stays at Strong Buy but trims the PT to $152 from $167 — and honestly, this one reads like a "we still believe, but the tape is ugly" note. NRG DOWN ~38% YTD, sitting at $97.84 just above its 52-WEEK LOW ($93.36), and the bogey list is long: weak ERCOT forwards, regulatory noise in Texas, and PM frustration that the data center contracting story keeps slipping.
The bull case rests on the LS Power deal closing the loop — RJ pegs Q3 adj EBITDA at $1.825B, ~4% ABOVE CONSENSUS ($1.751B), with $500-600M of that incremental from LS Power plus stronger Texas retail margins. The catalyst they're hanging the thesis on: year-end 2026 signing of the initial 1.2GW HYPERSCALER-BACKED CCGT, with PJM uprates, VPP/demand response, and a 4.2GW newbuild platform behind it.
"Raymond James believes the selloff more than reflects these risks."
Maybe. But worth noting the Q2 print was a miss ($1.49 vs $1.82 bogey) and Citi flagged NRG alongside PCG and EIX for ELEVATED IMPLIED VOL into Q3 end — the market is pricing this as a show-me story, not a power re-rating. R/r at 55% upside to the cut PT is interesting if you believe the hyperscaler CCGT lands by year-end. If it slips again, there's no floor near-term. Watch the contracting headlines, not the forwards.
Street's trimming numbers but nobody's flinching on the call. Raymond James cut PT to $176 from $182 (kept Strong Buy) on weaker ERCOT forwards and higher costs hitting the '27-28 outlook — and they're not alone. BMO went to $210 from $231, Wells to $212 from $220, Bernstein SocGen holding at $181, Mizuho initiating at $169. That's a wide $169-212 cluster, but every single rating is a Buy variant. The debate here isn't direction, it's how much ERCOT pain to price.
And the pain is real. STOCK DOWN ~7% YTD, -23% TTM, while RJ actually RAISED its 2026 EBITDA bogey to $7.269B — above the $7.2B guidance midpoint. Q3'26 est of $2.150B sits 2% above the Street's $2.110B and up 37% y/y. Texas gas margins, retail, PJM capacity, easy outage comps all working. The market is pricing commodity weakness; the numbers say contracted earnings power.
"Raymond James believes the selloff discounts considerable commodity weakness while assigning limited value to improving PJM fundamentals and additional contracting opportunities."
This is the classic IPP dislocation — forwards drag the stock while the contracting story (Meta, AWS, Cogentrix, the new 207 MW Luminant PPA starting Q3'27) builds underneath. Management agrees: buybacks have been aggressive into the weakness. At 25.7x on $8.66 of expected '26 EPS, r/r skews favorable if you believe PJM tightness and data center demand are structural rather than narrative. Not sure we can read too much into the ERCOT forward tape at these levels — but that's the bogey you have to underwrite. Watch for more hyperscaler announcements as the catalyst path; that's what closes the gap.
Ladenburg stuck with its BUY and $90.50 PT, and the interesting part is the model rebuild, not the rating. The firm now assumes BKH follows NiSource's GenCo structure for the Google data center generation build — that cuts equity needs by ~$576M vs. initial assumptions and pushes 2030 accretion to ~$0.79 from $0.51. Stock at $77.53 implies ~17% upside to target.
The accretion math splits into $35.1M net income from the 564MW generation investment plus $114.9M from the Microgrid Management Fee ($150M total). Worth flagging: most of the earnings come from the fee, not the generation asset itself — that's the part of the model I'd want more detail on before underwriting the full $0.79.
Street's lined up behind the deal: BofA at $97, Jefferies at $87, BMO at $86, Freedom upgraded to Buy at $78 (basically a market-perform in disguise). Nobody's fighting this tape. The cluster of PTs $86-97 vs. the stock at $77.53 says the Google/Cheyenne contract (effective Sept 2026, running to 2048) still isn't fully priced. Utility-with-a-data-center-kicker keeps working as a narrative — small cap, so position accordingly.
The street is telling you to buy the 15% pullback. B.Riley reiterated BUY with a $210 PT after Semtech's Oct 15 Data Center Analyst Day, and they're not alone — the tape says the market hated the event (-10% on the day, -15% over the week, -24% peak-to-trough per Northland), but the analyst corps is lining up on the other side of that trade. StoneX at $230, BMO at $230, Morgan Stanley raised to $205 calling the selloff "unwarranted," Northland upgraded to Outperform at $186 on the dip. Only Stifel blinked, trimming to $200 on valuation. That's a 5-vs-1 setup with PTs clustered $186-230 against a stock that just got cut down.
The bogeys are the story. Q3 FY27 midpoint: $417.5M revenue, +56% Y/Y, $1.14 non-GAAP EPS. Data Center business more than doubling in FY27 AND FY28, ramping to $2.1B by FY30. FY30 targets: $3.2B revenue, 70% GM, $12.40-13.00 EPS. B.Riley has FCF exceeding $1.3B cumulative over the next three years and approaching $1.1B in FY30 alone. At $210 they're valuing it at 14.8x FY29 and 11.5x FY30 EPS — not demanding for a name compounding data center revenue at this clip.
B.Riley raised its fiscal 2027 through fiscal 2029 revenue and earnings per share estimates.
The bear case is simple: stock was UP 96% YTD into the print, and "sell the news" on ambitious 5-year targets is what extended AI-adjacent names do. Optical/copper/laser guide-ups are only as good as hyperscaler capex follow-through. But 14 analysts have already revised earnings upward for the upcoming period — estimate momentum is running WITH the story, not against it. R/r favors adding into this weakness if you believe the FY27 double. Not sure we can read too much into one day's -10% reaction when the fundamental revisions are all pointing the same direction.
The AI Complex — demand vs. the moat debate
META — Tonight is the whole tape. Q3 numbers are probably fine; the stock trades on three things: 2026-27 capex口径, quantified AI ad monetization, and off-balance-sheet financing language. Beat + restrained capex keeps the complex bid. Beat + more capex hands the credit debate back to the bears. Two under-the-radar reads into the print: TPU inference is CONFIRMED IN PRODUCTION (real crack in the CUDA moat, procurement leverage), and EssilorLuxottica read-throughs conflict on AI glasses — WSJ says strength, FT says traditional frames offset a smart-glasses slowdown. Both true only if growth is decelerating. That's Meta's only hardware monetization proof, so it matters to the reaction function. Also flag: new commercial terms (Oct 30) shift AI liability to customers and grant Meta a license that SURVIVES TERMINATION — enterprise friction + regulatory headline, everyone copies it within a year.
NVDA — Demand signals keep stacking: Vera Rubin capacity at THOUSANDS OF RACKS/WEEK, a VR compute tray assembles in 1 MINUTE vs 90 for GB200, and SpaceX wants a 440K-UNIT Vera Rubin cluster by Q2 2027 (named demand, independent of Grok). China workaround is real money — ByteDance/BABA/Tencent ordered 3M+ RTX PRO 5500s, $35B+ sticker. Zero-China bear case is dead. The cracks are elsewhere: Meta TPU inference in production, AVGO bundling optics with custom silicon, 224G SerDes a two-player race. Also watch the KYEC prepayment — NVDA paying upfront to lock US test capacity tells you full-rack testing is the new bottleneck. Test is the pick-and-shovel trade here.
MU — HBM4 qualified and volume-shipping on Vera Rubin alongside SK hynix — even if model monetization disappoints, announced systems still need memory. Content is intact. Taiwan strike risk defused (Taichung signed, Taoyuan one-day only — immaterial unless profit-sharing sets a precedent). The structural worry: NVDA asking all three memory makers to build the NVHBM base die at TSMC (Samsung reportedly agreed) standardizes the differentiating layer and compresses memory pricing power vs NVDA. Spot is divergent — DDR5 allocation-tight, DDR4 up, NAND down 1.28%. AI mix pulling supply from legacy.
Neocloud credit — the tape is pricing contagion, pick your spots
CRWV — This is the stress point. 5yr CDS at a RECORD ~1,195bp, $35.6B debt vs $5.5B cash, market cap (~$33B) below debt, June par notes now yielding ~14.9%. $4.4B due rest of 2026, $6.2B in 2027. OCF $0.68B — negative without prepayments. Every macro hawkish tick tightens the refi clock.
NBIS — Down ~30% on CRWV sympathy, but the funding mix is cleaner: convertibles (~$9B this year), prepayments covered 21% of Q2 capex, $8.5B debt vs $8.0B cash. Mispriced relative to the sector — but not immune if the credit tape stays wide.
IREN — The cleanest contagion mispricing. Down 34% since Oct 2 despite equity + a $3B 1% convert + prepayments covering 87% of last quarter's capex. Net debt ~$1.7B, cash interest ~1% of OCF. Market is punishing the ticker, not the balance sheet.
CORZ — The real contagion vector is the landlord. CRWV is ~77% of CORZ revenue; it's also the tenant at Galaxy's Helios and leases 400MW from APLD. Down 17% since Oct 2 — less than IREN/NBIS, but if CRWV stress becomes lease renegotiation, it lands here first.
Hyperscalers & custom silicon
AMZN — GPUs sitting in INVENTORY waiting for power and shells. Plus €16B+ Germany investment planned for 2027 (€4.8B AWS Frankfurt). Confirms the gating factor is power, not chips, not demand. Same story as Microsoft killing Nscale over grid connections.
AVGO — The custom silicon toll booth. 224G SerDes is a two-player race (NVDA, AVGO); want a custom chip, you call Broadcom. The kicker: AVGO can bundle PRIORITY OPTICAL SUPPLY with custom chips even for customers running NVDA GPUs. Anthropic's $160B+ Google chip obligation flows through Broadcom. The bottleneck is migrating from switch to optics and AVGO owns both ends.
AMD — Chips-for-shares at Anthropic leaked. Vendor financing outside cloud channels = the lab is spending equity because cash is tight. Anchor customer for AMD, red flag for the AI credit debate. New financing template — watch who copies it.
TSM — Two structural wins in one feed. A14 6T SRAM bitcell below 0.017μm² (~25% denser than the 0.021 stuck since N5) breaks the SRAM stall — a genuine design enabler. And NVHBM base die consolidating at TSMC adds wafer content + stickiness (at the cost of more NVDA leverage over memory makers). CPO is shipping — Spectrum-X in mass production on COUPE. Watch the Kaohsiung F22 accident (4 dead, steel work halted) — limited near-term capacity impact, but 2nm expansion engineering risk is now a headline risk.
INTC — SpaceX ASIC is optionality, not a deal. Grimes TeraFab cleanroom rental is the logical angle given no new 48D site beyond Ohio. Foundry silicon photonics positioning (NPO/CPO waves) is the sleeper. Call it a cheap call option on the foundry narrative.
ASML — Moat vs. stock. Expert call flagged structural limits to equity performance despite the monopoly — debate shifting to cyclicality and China domestic EUV substitution. No hard numbers. Keep it on the radar, not the book.
ARM / QCOM — The V10 ALA fight is the number for both. ARM proposing 13%-30% per chip (~$220 ASP) vs. a counterproposal of 5% on the ENTIRE SMARTPHONE BOM. If the per-chip rate holds, ARM gets a massive step-up and QCOM's cost base eats it; if BOM 5% wins, ARM broadens the base but compresses the rate. Escalating, unresolved, headline risk on both sides until settled.
Optics & networking
LITE — Sold out on high-power CW lasers THROUGH EARLY 2029 — near-term earnings can keep beating on shortage. The risk is architectural: scale-up shifting to wide-and-slow VCSEL/microLED arrays, and Lumentum's own OCP demo used VCSELs from its 3D-sensing platform. That's a 2027-28 problem, but own it knowing the bogey.
CRDO — Opened an SRMO standardization workstream with a hyperscaler + ams OSRAM, targeting 2027 approval. Own the standard, expand the TAM; commoditize it, eat the pricing. High-beta catalyst either way.
COHR — Leveraged way to own the optical shortage (lasers, signal-processing chips, light sources, fiber, testing all gating). Low conviction on the signal itself; architecture shift is the risk.
MRVL — Wrong side of the bottleneck. Consensus: weaker connectivity/networking, compounding system costs post-deployment. At 224G, custom customers go to Broadcom. Avoid.
TEL — Quiet beneficiary of USFP/SRMO formalizing 400G/lane and 12.8Tb/s in an OSFP footprint. Connector/cable content per rack keeps rising. Second-derivative AI networking.
Media & advertising
TTD — Netflix cut its ad outlook $3.0B → $2.7B on PRODUCT/INTEGRATION DELAYS, not demand — advertiser demand stayed strong. Don't port the $300M into a TTD estimate cut. The real read-through is Netflix expanding into 15 additional ad countries in 2027.
DIS — All five Ice Age films in Disney's top 10 ahead of the February theatrical release. Netflix licensing supporting theatrical franchise value. Small, positive, directionally right for IP monetization across windows.
CHTR / CMCSA — Netflix leaning into live (NFL Thanksgiving/Christmas, 2027 Women's World Cup; Sarandos: live drives membership). Incremental bundle pressure, slow-burn churn risk, not an estimate cut. Watch the next video sub prints.
LION / AMCX — Second-derivative winners of the Netflix licensing bid for library content. Positive on content value, no numbers to hang on.
IMAX — Direct beneficiary of Netflix experimenting with windows: ~2-week IMAX run for Cliff Booth ~4 weeks pre-Netflix, ~45-day windows + wide releases for Narnia and Charlie. Windows widening = IMAX.
Software / health
FIVN — Q2 revenue $312.4M, AI revenue +78% YOY, full-year AI growth outlook raised to at least 60%. Cleanest contact-center AI monetization print in the group. Small base, accelerating.
TEM — ESMO data is the near-term catalyst, binary clinical/regulatory validation. No numbers in the source. Trade the event, not the thesis.
Power, cooling & industrial — the model-agnostic trade
VICR — Beat clean: revenue $197.1M, +79% YOY vs $187M est; GM 68.3% vs 63.2% est. Management says capacity expansion is necessary. AI power delivery outstripping supply — high conviction.
ETN / VRT — Both publicly disclosed in NVIDIA Vera Rubin reference designs. Power and cooling install BEFORE GPUs, pulling equipment demand forward of compute monetization. Model-agnostic — wins regardless of which lab wins. This is the cleanest way to own the buildout without the credit tail.
GEV — Gas turbine shortage is the durable kind: orders surging, slots scarce, prices rising, and manufacturers REFUSING to expand because they remember the last overbuild. Blades/vanes lagging 12-18 months. Cautious supply = longer cycle. Microsoft's Nscale confirmation puts official language on the bottleneck.
MMM — Sneaky AI revenue line. Amazon warrant for up to 3,162,015 shares at $164.89, 85% vesting tied to payments for optical connectivity/datacenter products, full vest at $6B CUMULATIVE payments. Microsoft adopted Expanded Beam Optics first. That's a hyperscaler-validated product line with skin in the game.
Everything else
GM — Memory inflation hits the P&L: ~$2B full-year headwind from commodities/logistics/DRAM, ~$700M in Q3 alone. Console BOM economics broken by 5x RAM/NAND. The memory upcycle is now taxing non-semis earnings — breadth confirmed.
T — Beat and raise, fiber the highlight, MarketWatch framing it as shrugging off SpaceX fears. Starlink D2C/satellite substitution NOT showing up in user growth. Narrative, not data — and Starlink's geopolitics makes that safer, not riskier.
FORM / TER / ATEYY — The test bottleneck trade. NVDA prepaying KYEC for reserved US test capacity; full-rack testing is emerging as the Vera Rubin output constraint. Advantest first call, Teradyne and FormFactor secondary (FORM probe cards specifically contested). ATE capacity is now a strategic resource — levered picks-and-shovels on hardware complexity.
BULL — Structural, not a fine. House China committee says Webull has "structural" ties to Beijing and US customer data is accessible. Post-TikTok framework now hitting brokers. Lasting overhang on the US business and the listed entity.
ESLOY — The Meta glasses tell. WSJ vs FT conflict resolves to decelerating-but-positive AI glasses growth, analysts warning momentum tails off. Direct read into tonight's META reaction function.