Good morning. The bond market is running the show again — 30Y TOUCHED 5.683%, 10Y PRINTED A 2002-HIGH, UK 30Y GILT BROKE 6% for the first time since the Euro crisis, and DXY at a 16-month high — yet Nasdaq futures pared the damage and still closed in the green. That's three straight sessions where the rate shock ate every intraday rally attempt and AI refused to crack. "Checked" is not "broken," but each day the complex absorbs this, the bar for the next shock gets higher. Brent back above $100 (China product export halt + Russia diesel ban) is the accelerant, and 16 US trucking bankruptcies in a month is the first corpse from energy inflation hitting the real economy.
The single biggest story on the tape is AVGO reportedly extending Anthropic up to $42B in credit to lease its own chips — banks got picky on DC loans 48 hours ago, and the chip guys just became the banks. Vendor financing is back, and everyone who's read the Lucent chapter knows it; the r/r hinges entirely on the loan terms (recourse, rate, accounting treatment), which we don't have yet. ORCL catches a bid on the Tencent $7B/5-year SEA deal (~100K chips, ~30% PREPAID — real cash, real demand), but you're now underwriting a BIS rule patch on offshore leasing as a line item. MU FLAT FOR A SECOND DAY after a $5B beat, and DIGITIMES handed the bears ammo overnight: the highest-margin lines are PC and handset, not AI datacenter — the margin-mix question on the shortage trade just got louder.
Asia overnight gave us more signal than the US session did: Huawei's Xu publicly claiming Ascend has overtaken NVDA in China share (discount the source, not the direction), DB HiTek hiking 200mm foundry prices up to 30% with 8-inch near full, and Japan planning a $140B DC build with Dell and Jera. China's compute trade is officially two-tracked now — offshore leasing plus domestic substitution — and both tracks are spending.
Four themes framing today: (1) FINANCING, NOT DEMAND, IS THE CONSTRAINT — AVGO-Anthropic is the thesis we wrote Tuesday arriving faster and more aggressively than expected; watch who else's balance sheet starts absorbing this risk. (2) MEMORY'S VERDICT IS PENDING — MU's tape is telling you the marginal buyer is gone until Q4 contract prices (DRAM +10-15%, NAND +15-20% bogeys) actually print. (3) CPU REVENGE IS REAL — Meta's Muse booked AMD EPYC orders into 2028; second engine for AMD, another share-loss warning for INTC. (4) AWS raising reserved GPU prices a FOURTH STRAIGHT QUARTER (H200s +75% YTD) says zero slack in the system — supply-constrained, not demand-constrained, full stop.
We'll hit up AVGO and the vendor-financing question first, then MU and the memory mix debate, then get to the semicap group and the China two-track reads...
The print settled the debate: this is a structural cycle, not a cyclical one. Beat-and-raise, 2027 already 75% SOLD OUT, supply tight through 2028, and a buyback bazooka unlocking December 9. The stock went FLAT after hours on a "moderating beat" and a GM trough guide — that's your tell on positioning, not fundamentals. Street PTs now span $1,200 (Morgan Stanley, still Overweight) to $2,200 (Melius), with the cluster re-anchoring at $1,400-1,650 and the high conviction names at $1,900-2,200. Consensus rating 1.37 — effectively unanimous strong buy ex-Goldman. MU at $1,065 trades at 5-6.5x out-year numbers against PTs implying 40-100% upside. Something has to give.
FQ4 revenue $54.2B (+31% q/q, 6% above the ~$51B bogey), EPS $33.42 vs ~$31.2-31.8 consensus. Guide: $61.5B revenue and $38.15 EPS for November — 7-8% ahead. DRAM pricing +HIGH TEENS q/q. NAND +30% q/q, with NAND revenue +42% SEQUENTIAL. The only blemish: GM guided down ~70-75bps to 86.25-86.3% on incentive comp and fab start-up costs. Management called that the FY27 floor — and 86% as a "floor" tells you everything about where this cycle is.
Three genuinely incremental data points, not just confirmation:
1. The SCA book exploded. 26 strategic customer agreements vs 16 last quarter, $32B in deposits/commitments vs $22B, >35% of revenue locked through 2030, first extensions into 2031. ~75% of SCA revenue carries floor/ceiling pricing bands. Customers are signing 5-year paper because they're scared of 2030 DRAM availability. That is not cyclical behavior. 2. 2028 commentary is NEW. Prior messaging was tight-into-2027. Now management says 2027 AND 2028 tighter than 2026, with DRAM bit growth decelerating to LOW-20% (from mid-20% in 2026) as HBM trade ratios eat wafers. Melius flagged the 2028 comment as the key new info — industry capacity adds get absorbed as HBM scales. 3. Capital return has a date. December 9 (CHIPS Act two-year anniversary). 100% of excess cash over time, buyback authorization expansion coming off a measly $2.2B current. Melius models $120B+ OF BUYBACKS IN CY27 — >10% of market cap. RBC sees ~$130B FCF from FQ2'27 through FQ1'28; Mizuho has FY27 FCF >$140B. Rosenblatt: >$100B FCF despite capex stepping to $55B.
Bull: The SCAs convert a commodity cycle into contracted revenue with pricing floors — Goldman itself admits that could expand the multiple investors pay at peak. HBM pricing resets 70-100% HIGHER in 2027 (Mizuho), Rubin ramp plus ASIC accelerators push HBM GM toward corporate average (Susquehanna), and enterprise SSD emerges as a second AI growth engine (Raymond James). At 5.6x/4.5x Cantor's numbers, you're paying cyclical-trough multiples for structural earnings. Melius sees EPS power >$200 and the cycle extending into 2029 on agents, robots, AVs.
"The threat of a supply-driven cyclical downturn appears to be a year or more away." - Wedbush
"This manufacturing and operating efficiency is not fully understood by the market." - Susquehanna, on DRAM wafer fungibility across server/HBM (differentiation is back-end packaging — cost sharing the market underappreciates)
Bear: Goldman's Neutral at $1,250 is the only real dissent, and it's thin — GM came in slightly below projections, and the beat magnitude is compressing (Morgan Stanley: 20-40% EPS beats the last three quarters → 5% this time). Morgan Stanley also noted industry wafer adds across the big three plus CXMT push industry bit growth to 20% next year vs 11% this year — supply is coming, even if MU's own piece is visible. And the SCAs cut both ways: Baird expects only MODEST GM expansion precisely because contracted pricing caps the upside capture. If 2028 capacity lands into slowing AI capex, the floor gets tested. Not sure we can read too much into the flat after-hours reaction — but at +486% in a year and a $1.2T market cap, the bogey is now "beat by more than last time."
Positive for the whole memory complex — Citi put a Buy and $2,100 PT on SNDK off this print. HBM pricing reset + Rubin/ASIC commentary is a direct read-across to NVDA's BOM costs (margin question for them, revenue question for memory). The "supply tight into 2028 with construction-heavy capex" line means WFE names don't get the full capex benefit — this is a buildings cycle, not an equipment cycle (construction growing meaningfully faster than WFE in FY27). And if 40-50% of hyperscalers are now under supply agreements (Baird), the spot market is thinner and more violent in both directions. Own the contracted names, respect the tail.
Bottom line: 18 of 19 voices say buy, the lone Neutral admits the LTAs could re-rate the multiple. The r/r at 5x out-year earnings with a $120B buyback unlocking in 10 weeks favors the bulls. Watch December 9 and the 10-K for buyback sizing — that's the next catalyst.
CEO leaves for Meta, stock pukes 21% in a week, and the Street's answer is... mostly a shrug and a trim. That's the tell. Nobody downgraded. Six firms in the feed, every single one still Buy/Outperform, with PTs now spanning $385 (Stifel) to $525 (Evercore) — cuts clustered at the low end (Stifel $465→$385, DADA $465→$415, BMO $450→$420) while the bulls (RBC $465, Piper $470, JMP $519, Evercore $525) held their numbers and pointed at the math. Consensus rating sitting at 1.63. This is a "multiple compression on leadership risk" tape, not a fundamentals break.
Investor day delivered a real raise, not a canned one: total revenue growth target moved to >20% from high-teens, Atlas to mid-20s from 20%+, with a shorter 3-year horizon (vs 3-5 prior) and $5B+ revenue over that period. RBC notes the midpoint implies >$1B in FY30 FCF — stock trades at ~22x that number and 7.5x EV/CY27E revenue today. Product side: MongoDB 9.0 now GA (faster querying, better price-performance vs 8.0), plus Atlas Infinite, Atlas Agent Engine, and a Devin partnership for agentic workloads and legacy migration. Piper raised out-year estimates post-event and flagged AI benefits "beginning to appear" in the financial model.
The genuinely incremental nugget came from Stifel's .local NYC channel checks: customers view MongoDB and Postgres as different workloads, not substitutes, and plan to keep legacy estates for the foreseeable future. That cuts both ways (more below).
Bull: You're getting 25% current growth, raised long-term bogeys, and $1B+ FY30 FCF at 22x — RBC calls the framework likely conservative. The CEO exit reads as comp, not code: BMO's take is Desai left because Meta paid up in the AI talent war, not because something's broken inside. 30 analysts revised earnings upward into this. AI-native products (Infinite, Agent Engine) give the second-derivative story legs.
Bear: No permanent CEO + Stifel explicitly saying no re-rating until you get one AND Atlas shows sustainable acceleration. That's the bogey — and it's a two-key lock. Postgres momentum is "growing" per Stifel, and their own customer work says legacy modernization is a slow drip, not a near-term catalyst: share gains "accrue gradually." The 21% weekly drawdown says the market agrees there's no hurry.
"We do not expect a meaningful re-rating of the stock until Atlas shows sustainable acceleration and a permanent chief executive officer is named." — Stifel
"We believe the CEO departure had less to do with MongoDB business and more to do with compensation from META in the AI talent war." — BMO
r/r is decent but not urgent. At ~$350-365 you're buying a raised guide at a compressed multiple with an interim CEO — classic "dead money with upside skew" until the two Stifel catalysts hit. The $385-525 PT spread is unusually wide and tells you exactly where the debate lives: transition risk vs FCF math. Not sure we can read too much into one week's -21% given the stock is still +38-40% over six months — this smells like positioning flush on a headline, not a thesis change.
Two threads. First, the Meta AI talent tax is now hitting software C-suites — watch for more "poached CEO" headlines across mid-cap SaaS; that's a sector-wide overhang nobody's pricing. Second, Goldman's software work (median ARR growth 21.3%, EPS beats of 14.2%) keeps the AI-driven software bid intact — MDB's Postgres coexistence data point is actually mildly positive for the database cohort broadly ( SNOW, legacy ORCL estates): migration TAMs are real but slow, so model them as tailwinds, not catalysts.
The street can't agree on much about CoreWeave — except that the 10% price hike dropped at Fully Connected this week is the story. Stock at ~$87 vs. a 52-week high of $153, and you've got buys clustered at $151-250 with a single Neutral at $115 doing all the skeptic work. That's a wild spread for a name growing 115%.
Range is $115 (Mizuho, Neutral) to $250 (Rosenblatt, Buy). Consensus sits heavily constructive — Truist $165, Cantor $176 OW, Jones $160, Freedom Capital $151, plus William Blair and JPMorgan both fresh to the bullish side (Blair initiated Outperform, JPM upgraded to OW citing "favorable pricing conditions" and premium short-term contracts). Even at the $151 low-end buy you're looking at ~73% upside. Rosenblatt's $250 implies 187%.
The glue holding the bull cluster together: PRICING POWER. CRWV just took another 10% price increase ON TOP OF the 25% hike flagged at the August print. That's the incremental datapoint from Fully Connected, and it's what separates this tape from the neocloud skeptics' "commodity GPU rental" narrative.
Bull case: Demand outruns supply, and CRWV is monetizing it. The company is selling THREE-YEAR access to A100s — silicon from 2020 — while NVDA says even 2017 Voltas are still running AI workloads. That's not a depreciating asset problem, that's a scarcity rent. Add SemiAnalysis/Gartner top-tier rankings on managed-cluster quality, MLPerf wins at rack scale, enterprise ACV tracking toward 20% of revenue via the Forge software stack, and Truist's read that pricing flows into 2027 exit ARR with a bigger kicker in 2028 as renewals roll. Financing handled: $3.7B converts plus a 35M-share ATM, and management says they're comfortable through 2026.
Bear case: Mizuho is the lone holdout and their ask is fair — show me margin leverage. 67% gross margins on $7.59B LTM revenue is nice, but the company is burning cash to scale, rate environment isn't helping, and the 12-month revenue upside magnitude is still fuzzy. There's also the mix question Freedom Capital raised: discounted bare-metal middleman deals to hyperscalers aren't the same animal as premium managed capacity, and the gap between leaders and challengers narrows over time. The double equity/convert raise also tells you the capex beast needs constant feeding — that's dilution plus leverage against a thesis that requires pricing to hold.
"CoreWeave demonstrated a leading technology stack and strong demand that enabled the company to raise prices." — Rosenblatt
"We would like to see evidence of more margin leverage." — Mizuho (the honest bear in the room)
"Growth driven by providing discounted bare-metal capacity to hyperscalers in a middleman model is not an apples-to-apples comparison with providing managed capacity directly to end customers." — Freedom Capital
The pricing data point matters beyond CRWV. If neoclouds can stack a 10% hike on a 25% hike while renting out 5-year-old silicon on 3-year contracts, the "GPU depreciation cliff" bear case on the whole complex (NBIS, IREN, the neo-cloud basket) takes a hit — and it quietly supports NVDA's installed-base longevity argument too. Power and land remain the bottleneck, which keeps the scarcity rent alive into 2027-2028 renewal cycles. Positioning-wise: CRWV at $87 with a $153 high and a wall of $150+ PTs smells like a name PMs got shaken out of post-lockup/post-raise. If the next print shows any margin leverage at all, Mizuho's last bear leg falls and the path back toward consensus is straightforward. R/r favors the bulls here, but you're underwriting capex intensity and financing risk the whole way. Size accordingly.
The Street is lining up behind one thesis: AWS is monetizing AI scarcity, and it's working. Wells Fargo reiterated OW with a $338 PT after AWS announced a 15% GPU reserve price hike effective Oct 7 — the FOURTH CONSECUTIVE quarterly increase, with H100/H200 and B200/B300 reserve prices now UP 40-75% YTD. That follows a 20% bump in July. A100s got included for the first time this year, meaning the demand wave is broadening beyond the newest silicon, not just the frontier stuff.
Meanwhile Goldman added AMZN to its Director's Cut conviction list (20-25 most differentiated Buy ideas, refreshed monthly), and Rosenblatt pushed its PT to $360 on the view that agentic commerce fears around advertising are overdone. Add it up and the bull case has two legs: AWS pricing power on the cloud side, ads resilience on the commerce side.
The caveat Wells flagged matters — reserve capacity is contracted up to six months out, so there's a lag between announced price hikes and realized revenue. This isn't a Q4 number; it's an early-2027 acceleration story.
"The fourth quarterly price increase in a row highlights sustained demand for AI compute and supports hyperscaler ability to pass through supply chain inflation expected to impact 2027 capacity additions."
Bull: 40-75% cumulative GPU price hikes with no visible demand destruction = real pricing power, which supports the AWS revenue acceleration bogey through early 2027. Trainium2 pricing held flat, so AMZN is also steering workloads to its own (cheaper) silicon margin lever. At ~20x earnings, PMs aren't paying a frothy multiple for this.
Bear: The monetization lag is real, and these hikes could equally be read as supply chain inflation pass-through rather than pure pricing power — cost defense, not revenue offense. Plus a UK class action over the 2018 Apple sales agreement just got cleared to proceed. Noise, but it adds to the regulatory drip.
This is a "quiet grind higher" setup into earnings (28 days out). The Director's Cut add plus consecutive analyst PT raises ($338, $360) signals positioning is rotating back toward AMZN as the cleanest big-cap AI compute play. Watch Trainium3 listing status — still absent from the reserve price sheet, and that's the custom silicon tell for 2027.
Investor Day delivered. The OpenAI + AWS one-two punch gives SNPS the AI monetization story the stock has been missing, and the FY27 guide came in ABOVE the street — $11.15B rev / $19.08 adj EPS vs. Stifel's $10.80B / $17.34 bogey. That's rare for a company still digesting Ansys. Stock sits at $434.94, DOWN 7.4% YTD, with consensus targets clustered $415-$700. The gap between the print and the price is the opportunity — or the tell.
Three things, in order of importance:
1. The AWS deal is real money, not vapor. Multi-year license exceeding $1B across Graviton, Trainium, and Nitro. AOIP projected ~$1B by FY30, royalties ramping FY28. That's a custom IP royalty stream layered on top of the license — the kind of recurring, high-margin revenue PMs pay up for.
2. OpenAI partnership opens the agentic AI revenue line. Post-trained OpenAI model bundled with Synopsys tools, agents, and workflows — tool licenses plus outcome-based revenue share. AgentEngineer drives 5-10X more EDA tool calls. Agentic AI revenue starts FY27. This is the "EDA as AI toll booth" thesis made concrete.
3. LT targets with a floor. Mid-teens revenue CAGR through FY30 with a 15% FLOOR (management doesn't give floors unless they're confident), ~50% adj op margin, ~$35 adj EPS by FY30. Backdrop helps: Q3 was a clean beat ($3.91 / $2.48B vs. $3.67 / $2.44B bogey) with a raised FY guide.
Consolidated view: six firms moved or reiterated bullish, PTs now spanning $500 to a street-high $700. Rosenblatt to $620 from $575, HSBC upped to Buy at $700, Stifel holds Buy at $600, Benchmark reiterates at $570 post-print, MS upgraded to OW at $500 (Ansys integration + Design IP recovery), Baird to Outperform at $560 on double-digit organic growth returning FY27. Rosenblatt's read captures the mood:
"On Wednesday, we attended a very positive Synopsys Investor Day in New York, at which the company announced a product development partnership with OpenAI, a significant custom IP contract with AWS, and provided a new FY27 and long-term growth outlook that was ahead of our forecasts and consensus."
Bull: You're getting a 15%-floor grower with 83% gross margins, ~50% target op margin, and two new AI revenue vectors (outcome-based OpenAI share, AWS royalties) that aren't in anyone's numbers yet. The MS/Baird upgrades argue the Ansys synergy and Design IP recovery de-risk the organic story. At $435 vs. a $600+ consensus cluster, the r/r skews hard if FY27 prints anywhere near guide.
Bear: That 46% LTM revenue growth is Ansys inorganic — strip it and the organic recovery is still a promise, not a print. Royalties don't ramp until FY28, agentic revenue until FY27, so you're underwriting 2-3 years of execution on faith. And note the low end of the target range: $415 sits BELOW the current price. Someone on the street thinks this is fully valued already. Stock down 7.4% YTD while semi software ripped says positioning hasn't forgiven the post-Ansys indigestion yet.
This was the catalyst SNPS needed and the guide backed it up. Not sure the OpenAI revenue share matters near-term, but the AWS $1B and the 15% floor do. Watch FY27 organic growth ex-Ansys on the next print — that's the number that decides whether $600 targets or $415 targets are right.
Classic good-print-bad-reaction setup. JBL beat on everything — Q4 CORE EPS $4.40 VS $4.06 STREET, REVENUE $10.6B VS $9.69B (+29% YOY), above-consensus FQ1 guide — and the stock is DOWN 8% ON THE WEEK anyway. Still +34% over the past year, so this smells like positioning/froth-coming-off rather than thesis break. Both covering desks (Stifel, Argus) used the weakness to reiterate Buys at $460 and $475 — that's 60%+ implied upside from $287, and neither flinched.
The number that matters: FY27 GUIDE OF ~$44.5B REVENUE (+24%) AND CORE EPS $17.55 (+34%). For an EMS name. That's not an EMS multiple story anymore.
AI revenue guided to ~$22.1B IN FY27, UP ~54% AFTER +60% IN FY26. Half the company is now AI infrastructure — hyperscalers, networking, power/cooling, semi cap equipment and test. Second hyperscaler customer crossing 10% of revenue in FY27, third one still modest (that's the embedded upside if it ramps). Demand exceeding supply in parts of the portfolio, with inventory building sequentially through the year — read that as backlog confidence, not bloat.
Also worth noting the non-AI tail: defense/aero, healthcare, energy infra, robotics all growing. This isn't a one-legged AI stool.
Bull: PEG OF 0.62 on a business compounding EPS at 30%+ with AI revenue still accelerating into FY28 catalysts (additional hyperscaler, healthcare, India CPO, physical AI). Sell-off is a gift.
Bear: Execution and margin conversion. Core op margin guide is only 6.1% — this is still an EMS P&L, and the aggressive U.S. capacity build is exactly where these stories go wrong if demand pauses. The 8% post-beat drop tells you some holders are already underwriting that risk.
Stifel's framing is the right one — near-term this is an execution story, not a demand story:
"Stifel views execution and margin conversion as the key near-term factors, with fiscal year 2028 opportunities including additional hyperscaler, healthcare, India CPO and physical AI projects."
Take: the r/r here favors the bulls at $287 with two desks at $460+. The quarter proved demand; the guide proved durability. What it hasn't proven is that JBL can build this much capacity without stepping on margins — that's the bogey every print from here. Not sure the 8% dip is all-clearable in a tape that's punishing capex-heavy AI names, but on fundamentals alone this screens cheap for the growth.
The $350 PT stays, and the bogeys keep getting bigger. Cantor hosted Jensen, Colette Kress, and IR in NYC last week and came away reiterating Overweight with a $350 target — roughly 13x their updated CY28 EPS of $26. Stock sits at $228, $5.51T market cap, so the target implies ~53% upside if you want to do the math on what the Street thinks is still left here.
The takeaway from management: demand is broadening beyond the usual hyperscaler suspects, Agentic AI is opening new verticals, and NVDA keeps capturing a bigger slice of total IT capex with each system generation. Cantor now sees upside to the guided 70%+ CY27 revenue growth — which is a sentence I'd have laughed at two years ago — and Jensen is out there talking $3-4T in 2030 compute spend with semis tracking to $20T over time. Yes, $20 trillion. Vision-casting, sure, but the man has a decent batting average on these calls.
Street's with him: 35 analysts revised earnings up for the upcoming period. When the estimate revision momentum and the narrative are running the same direction, you don't fight it.
Cantor sees "upside to the company's guided 70%+ calendar year 2027 revenue growth" and raised its model on "the more robust demand backdrop and incremental upside from Nvidia's supply chain work."
Bear case, steelmanned: at some point the law of large numbers bites, and a 70% growth guide for CY27 leaves zero room for a digestion quarter. Racks shipping (SMCI's Vera Rubin NVL72s are moving) means supply catching demand — great for revenue, but it's also how margins eventually normalize. Not a today problem. Probably not a this-quarter problem either. Keep riding, watch the revision trend as your tell.
Networking investor day did its job — but the stock's already at the finish line.
Wells Fargo lifted its PT to $63 from $54, kept Equal Weight, and the stock sits at $63.89 — ABOVE THE NEW TARGET after a +169% YTD run, a stone's throw from the 52-week high of $67.10. Rakers flagged the Juniper integration tracking better than expected, and the networking-for-AI bogey got another lift: cumulative FY25-26 orders target now >$3B exiting FY26 (from $2.5-3B prior), implying >$2.1B in FY26 orders vs ~$900M in FY25. Helios trays have already pulled in $200M+ of orders, with HPE sizing the two-year Helios opportunity at >$1B.
Wells Fargo analyst Aaron Rakers noted the Juniper integration is performing better than expected.
The catch: FY27 is back-end loaded on Helios shipment timing and the Oracle deal, so expect lumpiness and some dead quarters on the way. And note the positioning split — Goldman at $75, Raymond James at $86, Truist at $70, while Evercore downgraded to In Line purely on appreciation. That's the debate in a nutshell: nobody disputes the AI networking story anymore; they dispute whether you get paid to own it HERE at ~11x FY28 earnings with the good news already printed. R/r looks balanced at best until the revenue conversion actually shows up in the P&L.
Citi stays Buy, $2,100 PT — and the MU print gives the NAND squeeze thesis real teeth. MU's NAND revenue jumped 42% q/q with pricing up ~30% versus the Street's 20% bogey, and Citi reads that straight through to SNDK (closed at $1,739.89, so meaningful room to target if the setup holds).
The crux: industry NAND bit shipments projected mid-20% growth in CY27-28 with supply constrained BOTH years. MU itself signaled NAND supply growth below industry growth in CY26 — that's a producer telling you they're not adding wafer starts to chase price. Citi flags AI datacenter demand as the driver, specifically KV cache offloading to SSDs (a genuinely new NAND demand vector, not the usual smartphone/PC cyclical story). One honest tension: MU's NAND supply growth runs above SNDK's own mid-teens industry CAGR assumption, but Citi waves that off on demand durability and non-binding LTAs with customers.
"Industry NAND bit shipments are projected to grow in the mid-20% range in calendar years 2027 and 2028, with the industry expected to remain supply constrained during both years."
Pure-play NAND leverage keeps working as long as pricing outruns expectations — and a 30% print vs 20% consensus says it is. Watch whether those non-binding agreements convert to committed volume; that's the gap between a squeeze and a structural upcycle. Not sure one MU read-through makes a new thesis, but the direction of revisions is your friend here.
GOOGL keeps grinding on the AI model cadence story. KeyBanc reiterated OW / $435 PT (24.3x their 2028 EPS) after Gemini 4 "Argon" dropped earlier than expected, with progress across most benchmarks and a compelling cost-per-task profile — though it still lags on coding, and the paced rollout means nobody can really gauge true performance yet. Piper's also in the bull camp at $400 PT, keyed on TPU sales they see topping $100B by 2028.
The firm framed the three debates that matter from here: speed of assistant improvement, how Gemini 4 + assistants feed Search and Cloud, and the capex bill to support it all. That last one is the real tension in the story — Gemini progress helps sentiment near-term, but the 3p model race keeps pushing the spend bogey higher (TSMC advanced capacity sold out through 1H28 tells you where the industry's wallet is going).
KeyBanc acknowledged the announcement could bolster near-term sentiment but said the paced rollout makes it difficult to gauge true performance.
Net: announcement is a sentiment positive, not a thesis changer. Stock at 17.5x with a $4.19T cap — the r/r still hinges on whether Gemini actually moves Search/Cloud economics, and one early model drop doesn't answer that.
Needham's Laura Martin staying at Hold after the Muse launch — and honestly the reasoning is hard to argue with. Stock up 27% SINCE SEPT 1 vs a FLAT S&P, trading right at InvestingPro's $725 fair value, and the Street now has to underwrite >$135B of FY26 capex with NEGATIVE FCF. That's a lot of AI optimism already in the price.
The product side is genuinely impressive though. Muse pulled ~5M downloads in three weeks, #1 free app in the US and Canada, and the Connectors list keeps ballooning (Walmart, Expedia, Instacart, Shopify's entire catalog). Truist thinks it pressures OTAs; Jefferies notes Meta jumped to #3 on the Artificial Analysis intelligence ranking. The agent commerce narrative is real.
But two yellow flags: Amazon already blocked Muse from shopping its site (walled gardens gonna walled garden), and OpenAI fired back with Dots agents on Sept 29. When your distribution edge is "we have the apps," a competitor with 1.2B weekly users is not nothing.
R/r at fair value with a Hold cluster (Needham + StoneX) forming — this feels like a name where you wait for a pullback or a Muse monetization proof point rather than chase. Not sure we can read too much into download numbers yet — downloads aren't MAU, and Meta says Muse data won't feed the ads engine, which begs the question of what the model actually is.
VICR's licensing model is working, and the AI royalty stream just went from thesis to numbers. Needham took the PT to $350 from $320 (Buy) after the company raised Q3'26 revenue guidance FOR THE SECOND TIME IN THE QUARTER — now calling for >30% QoQ growth, up from >20%, up from the original ~10%. The driver: a non-exclusive VPD royalty deal with an unnamed AI OEM, with Needham flagging the upside likely includes catch-up payments from prior quarters (consistent with how these licensing agreements have historically recognized revenue).
Needham expects this to be Vicor's largest revenue quarter from a single licensee to date.
Context matters on the setup. This is the same Needham that CUT the PT from $400 to $320 on valuation after Q2. So the $350 is them warming back up, not breaking new ground — and the valuation question hasn't gone away. Stock trades at 93x earnings; the new PT leans on 42x their 2028 "next-gen" EPS estimate of $8.37, which means you're paying for licensing scale that hasn't fully shown up yet. FY26 guide now >$600M (from $570M), consensus modeling 53% growth, 56.6% gross margins. Plus they're building Fab-2 and Fab-3 in New Hampshire — capex commitment signals they believe the product side follows the royalty side.
R/r here hinges on whether the AI OEM royalty is recurring run-rate or a one-time catch-up pop. Needham is betting recurring. Given they just cut the target two months ago on valuation, the fact they're back raising says the licensee scale surprised them. Watch Q3 print for how much of the beat is catch-up vs. forward run-rate — that determines whether $350 has legs.
Street is coalescing around the E7/Copilot consumption story as the next leg of the M365 monetization trade — and Piper just gave PMs the framework to model it. PT to $610 from $550, Overweight maintained, on the back of the "Super App" announcement and the new E7 tier. The math that matters: every 10% of seats shifting E5→E7 = ~$2B in annualized revenue uplift, and Piper sees Copilot/Cowork consumption scaling to a $2B run-rate by FY28-end. That's today's seat-based Copilot revenue scale in half the time.
The pitch underneath is worth internalizing: MSFT's value-add is the harness and auto-routing layer doing tasks cheaper than frontier models alone. If that holds, Copilot stops being a seat-count story and becomes a consumption story — which is exactly the Azure playbook bulls have been waiting to see repeat in software.
Not just Piper. Stifel upgraded to Buy ($575 PT) on AI positioning across Azure + Copilot. Citizens at $550 Market Outperform post the new Copilot features (Home, Code, Autopilot, Today). Street range is wide — $440 to $870 — but the direction of travel is one-way: 17 analysts revised earnings up into the print.
Oppenheimer flagged the shift toward measurable AI outcomes and disciplined spend — the "show me the ROI" phase. Piper's own note admits immediate E7 impact is minimal. This is a FY27-28 revenue story being priced today. R/r still fine at ~$3.8T given the multiple expansion is riding on an EV/OCF framework with real bogeys attached, but the seat-shift assumption is the whole ballgame. Watch E7 attach commentary next print.
(Side note: Nadella at the White House state dinner for Xi alongside Zuck and Jensen — geopolitical access trade alive and well for mega-cap tech.)
Cantor walked the plant, sat with CEO Cope and CFO Metcalf, and came away reiterating NEUTRAL / $235 PT. That's ~27% upside from $184, but don't mistake the implied upside for conviction — this is the same firm that whacked the PT from $320 to $235 after F3Q26 (rev $312M, EPS $1.42, slight miss on both despite +9% y/y growth). The visit itself reads constructive: management says NO CHANGE in the demand environment across data centers, utilities, and industrial, with utilities the long-duration focus, and capacity is not the bottleneck on executing the backlog.
"Powell continues to benefit from strong, broad-based demand across data centers, utilities, and traditional industrial markets."
The stock tells the tension. POWL +74% over the past year but down HARD off the $328 high — classic high-multiple power-equipment name where the narrative (record orders, record backlog, strong cash flow) is intact but the market stopped paying up for "beat-and-raise" that isn't beating. The r/r here hinges on one question: is the last print's miss a capacity/execution hiccup (mgmt says no, and the leased-facility throughput plan supports that) or the first crack in the backlog conversion story? Cantor's not willing to underwrite the bull case at these levels, and honestly after a -44% drawdown off the highs with the target already cut once, the bar for a re-rating is a clean quarter. Watch the next print — record backlog means the bogey is delivery, not demand.
The bull case here is a valuation argument, not a fundamentals one. Needham reiterating Buy / $15 PT on the DoubleO.ai tuck-in — that's nearly 4x upside from $3.79, which tells you more about how bombed out the stock is (-63% over the past year) than about the deal itself. DoubleO.ai is a ~10-person agentic orchestration shop with a few million in revenue. Needham's Reilly is calling it what it is: a small tuck-in that bolts agent orchestration onto ZI's GTM data. Fine. Doesn't change the story.
The actual story: Q2 was fine — $310M revenue vs $301.5M bogey, 35% adj op margin, $107M uFCF, 86.5% gross margins. The cash generation is real. The problem is the other side of the tape — DA Davidson cut PT to $4.25, Neutral, on exactly the bear thesis that matters: AI-native tools eating the legacy GTM software stack. That's the existential question, and one small acquisition doesn't answer it.
Bull: you're getting a 35% margin, $100M+/quarter FCF business at a price where the market assumes terminal decline. Bear: the market might be right — if AI-native GTM tools disintermediate the data layer, ZI's moat is the thing being disrupted, not the thing doing the disrupting. The DoubleO deal is management acknowledging that. Whether it works at this scale, not sure we can read too much into a 10-person tuck-in.
Net: r/r interesting for deep value accounts willing to underwrite the AI-disruption risk. Nobody's paid to wait on this one yet.
UBER near 52-WEEK LOWS at $69.15, trading at a 15.16 P/E — and Bernstein says the market still isn't paying you for the autonomous option. They reiterated OUTPERFORM / $95 PT on the AV rollout math: 15 cities live by year-end 2026 (6 US, 9 international), with Nuro in the Bay Area and Houston, Zoox in Vegas and LA, and Wayve in London/Tokyo as the deployments that matter. (Avride going driverless in Texas too, though the safety record needs work.) Citizens is in the same camp at $100, pointing out Waymo's rider-only miles grew just 2% in Q2 — a real slowdown from prior quarters, which takes some air out of the "Waymo kills Uber" bear case.
Bernstein said the stock continues to trade at a "too cheap" multiple but believes a driverless deployment in the US with a newer AV entrant is needed to address the terminal value bear case.
That's the crux. The terminal value debate doesn't resolve until UBER proves a driverless US market with someone OTHER than Waymo. Until then, the bull case is you're buying the core mobility/delivery machine at 15x with free AV upside; the bear case is that "too cheap" multiple is the market pricing AV disruption risk correctly. Bernstein also flagged NVDA deepening its Uber relationship into 2027 — worth tracking as a tell on how serious the AV pipeline is.
One housekeeping item: the €4.5B senior notes offering (incl. €750M at 3.750% due '29) is routine capital structure management, not a signal. And on delivery — fees up 16% in Q3, now aligned with DASH, second straight quarter of hikes. That's pricing power, though it puts the consumer bogey in play if volumes wobble.
R/R here: you're not paying for AV success at $69. If one of Nuro/Zoox/Wayve prints a clean US driverless deployment, this re-rates. If AV stalls, you own a profitable compounder at 15x. Not a bad spot to be patient.
The bull case keeps building into the Q3 print, but the valuation math is getting harder to defend. Evercore reiterated Outperform at a $1,350 PT and raised '27/'28 estimates 3-4% — Electrification EBITDA up 6%, Power up 3-6% — leaving them 2-7% ABOVE CONSENSUS on orders, sales, and EBITDA through 2028. The key tell: they expect order strength to pull the $200B BACKLOG milestone into EARLY 2027. That's the number PMs are watching.
The rest of the street splits the difference. Baird ($1,400) and Bernstein ($1,298) stay Outperform on services and electrification; Jefferies sits at $1,185; Mizuho's the holdout at Neutral/$949 despite calling for a strong book-to-bill in 2H. Everyone agrees on the fundamentals. The fight is over what you pay for them.
Evercore: the services annuity and new electrification products — medium-voltage UPS, energy management, solid-state transformers — are "underappreciated by the market."
The other side: stock's up 57% in a year, trading at ~62x EV/EBITDA, and flagged on overvalued screens. Consensus bogey of 22% revenue growth is already in the number. R/r here depends entirely on whether backlog conversion accelerates into '27 — the estimate raises suggest the buyside whisper is drifting higher, not lower. Own it if you can stomach the multiple, but there's no margin of error left at these levels.
Morgan Stanley cut its PT to $355 from $360 — a nothing-burger of a trim — while holding Overweight. The tell is in the framing: the roadmap is "among the most exciting in over a decade" (first 4 straight years of iPhone unit growth since FY15, EPS compounding ~13% through FY28), but after AAPL'S +30% RUN IN SIX MONTHS, the iPhone launch changed the model almost not at all. Translation: great story, owned consensus, priced for it at 38.5X FY26 EPS of $8.93.
The margin cross-current: lower-than-expected new iPhone pricing modestly pressures ASP and GM assumptions, offset by better units, Mac upside, and Services pricing actions. MS still sits 0-3% above Street on FY27 (~$10 EPS) and FY28 (~$11).
Sept quarter looks clean — resilient iPhone builds, stronger Mac, Services pricing tailwinds, tariff refunds. December is the problem child: MS says Street numbers are stale and haven't incorporated the staggered launch, leaving iPhone shipment estimates "too elevated." That's your guide-risk bogey for the print, not a thesis breaker.
Elsewhere on the tape: Bernstein stays Outperform but flags memory-cost pressure on gross margins (strong international pricing as the offset), UBS holds Neutral with lead times steady in China/Japan, Needham stays Hold on Meta/AI ecosystem risk. Net net — nobody's leaning the other way, but nobody's paying up either. This is a "hold your winners, don't chase" name into the Sept print, with Dec guide as the real event risk.
Q4 Thursday. That's the story. UBS stays Buy, $275 PT, and the options market is pricing an 8.5% MOVE vs. a 4.3% historical average — so vol is rich and positioning is nervous into what could be a FY27 guidance print. Stock's trading at ~14x earnings with an 11% FCF YIELD. That's not a valuation, that's a capitulation multiple for a name doing 60%+ of revenue through its top 10 alliances (up from ~25% in 2018).
The bull case is simple: emerging AI alliances (NVDA, OpenAI, PLTR) are on pace to MORE THAN DOUBLE bookings, and the Anthropic deal — $1B+ each over five years on frontier model eval — gives them a real seat at the AI services table (Citi flagged it as a positive for the AI governance angle). Bear case: fixed-price revenue has crept to ~60% from ~50% in 2023, which is margin risk if AI delivery costs surprise. BMO's at Market Perform with a $200 PT, so the street is split.
Four things to watch Thursday per UBS:
R/r here leans interesting. Implied 8.5% move on a stock at 14x with double-digit FCF yield means bad news is largely priced — you'd need an ugly FY27 guide to get paid on the short side. Seasonals help too (Oct-Dec has averaged +3.6%, +4.6%, +2.9% over five years). Not a table-pounder, but if the guide merely meets consensus (BMO's bogey), the path of least resistance is up.
AVGO — The vendor financing era is officially open, and AVGO is the bank. $42B credit line to Anthropic to rent its own chips — roughly a third of the $125.2B five-year TPU commitment — with debt possibly convertible into equity. Banks stepped back (CCC spreads >1000bps), the chip supplier stepped in with its own balance sheet. Revenue and credit risk now sit on the same income statement. Bull offset: procurement roughly on par with NVDA, TPU volumes reportedly exceed NVDA in 2027. Watch the next print for recourse structure, rate terms, provisioning. If it's limited recourse at market rates, that's a re-rating event. If it's subsidized financing dressing up revenue, the 2000 Lucent/Nortel chapter is getting reprinted.
ORCL — Tencent signed a five-year ~$7B lease for ~100K AI chips across SEA DCs, ~30% PREPAID (~$2B cash up front). Tencent took its first negative FCF quarter in a decade (-RMB 13.8B) to do this deal — that's how scarce offshore compute is. Real cash-flow support for a stock beaten up on force majeure and financing doubts. The bear case writes itself: BIS only needs a "leasing-as-export" rule and execution on the contract is in question, and Huawei publicly claiming Ascend beats NVDA in China accelerates exactly that patch. R/r better than it was, not clean.
AMD — CPU second engine is now实证, not framework. Meta's Muse agent rollout pulled EPYC orders into DEEP 2027 AND 2028, lead times stretching — agentic workloads run control-plane tasks on CPUs, and Meta is the first hyperscaler proving it with deployment data. Macquarie's "Revenge of the CPUs" went from thesis to order book. The asterisk: Helios teardown confirms AMD still pays the AVGO toll on retimers and scale-up switch silicon — no in-house rack integration means ceding economics to suppliers and NVDA's full-stack moat. Retimer latency doesn't fix itself this cycle.
INTC — The Hynix Ohio DRAM JV rumor is a now-or-never Section 48D play — credits expire 12/31/26, projects must be announced and underway by year-end. Framed economics: ~$50B each pre-credits, ~$33B net. Trump admin holding $50B of Intel stock keeps the political put alive. This is industrial policy trading, not fundamentals trading. Meanwhile AMD EPYC stretching to 2028 is another share-pressure print for Intel's server installed base — the CPU revival is real, it's just not being booked in Santa Clara.
TSM — Two clean adds to the story. Weighing a tens-of-billions Texas campus for AI capacity, and Micron's HBM4E base die moving external opens a NEW DURABLE REVENUE LINE beyond accelerator logic (TSM already named the partner). Inventory math still says early cycle: non-NVDA fabless inventory at $0.99 per $1 of TSM quarterly revenue, down a third from the 2Q23 glut peak, bottomed at $0.81 in 3Q25. The 2021 analog ran 4-5 more quarters before correcting.
GOOG — Gemini 4 Argon launched the same day Bloomberg reported internal skeptics calling it benchmaxxed — struggles on real coding tasks, EXPENSIVE TO RUN = margin pressure. Trusted-tester-only rollout is the historical tell for a usability letdown. Watch developer vibes, not benchmark sheets. The silicon side is the offset: TPU volumes reportedly exceed NVDA's in 2027, and Anthropic's $125.2B commitment is the largest external validation of the stack. The TPU franchise is becoming merchant-scale; the model franchise is the question.
CRBS — Worst combination in the book: COO Dhiraj Mallick liquidated his ENTIRE $78M STAKE — to zero — days after SemiAnalysis reported GPT-6 ultrafast is NOT running on Cerebras. Insider selling into a thesis-breaking headline, with ~80% of the $25.4B RPO riding on the OpenAI token-resale model. The bull case now rests entirely on contracted take-or-pay — three tranches >$20B through 2028 plus a $45B+ option — but they're renting back G42 systems to fill Tranche 1, a gross margin headwind through 2026. Contracted revenue is real. Incremental wins are now questionable.
LITE — The investable fact is the 8-K, not the chatter: 6-YEAR InP substrate agreement with AXT, minimum annual purchase commitments, $87M IN DEPOSITS. Companies don't prepay $87M for substrates they can buy off the shelf. Management cites continuing laser shortages and favorable pricing; Greensboro 6-inch ramp lands mid-2028. The Vital Materials anecdote is unverified noise.
COHR — Second major buyer in the 6-inch InP transition. AXT disclosed a June 2026 development-and-supply agreement with a $22.2885M PREPAYMENT from Coherent. 6-inch promises 4X DEVICES PER WAFER and >60% DIE COST REDUCTION vs 3-inch. Two customers prepaying nine figures combined to lock the same supplier's capacity — that's the demand signal.
AXTI — The toll booth on the InP transition. Collecting deposits from both Lumentum ($87M) and Coherent ($22.3M) on multi-year reservations — pricing power and demand visibility confirmed simultaneously. China's Feb 2025 export controls on InP add a geography premium to non-Chinese qualified supply. Small cap, real leverage to the optics shortage.
FORM — Deutsche Bank Buy initiation: FORM has "cemented" second-source probe cards for NVDA GPUs at TSM — TAM EQUAL TO ITS ENTIRE 2026 LOGIC PROBE CARD REVENUE ($500M), first sales 2H26, path to 20% share by 2028. HBM probe cards can still hit $400M by 2028 even after ceding 40% share to Technoprobe. Stock traded down $80 in July; consensus FY27 revenue looks extremely light. Test intensity is the forgotten derivative of the HBM ramp.
COHU / AEHR / TER — Basket read-through, no company-specific catalysts this window. Rising test dollars per chip from HBM, advanced nodes, hybrid bonding moving to mass production. FORM's probe-card TAM expansion is the visible proxy. "Buying testing names has already made money" — the feed's framing, and the logic holds.
ASML — Micron's DeBoer: EUV is "critical to all advanced DRAM nodes going forward." Multiyear memory WFE demand is more durable than bit-growth models suggest — manufacturing effort per bit keeps rising even as bit growth slows. Timing discipline: MU's >$50B FY27 capex is majority CONSTRUCTION, so WFE orders land on a 2027-29 lag, not an immediate pop. Right thesis, wrong quarter if you're chasing the headline.
LRCX — Same Micron capex read-through, same 2027-29 lag. Differentiated front-end processes across HBM, low-power, DDR raise etch/deposition intensity per bit. Memory WFE durability is the thesis; timing is the discipline.
AMAT — Co-set the 3D IC/hybrid bonding agenda with TSM and ASE at OCP APAC, and the SEMICON follow-up moved from "future trends" to MASS PRODUCTION READINESS. Hybrid bonding tools transitioning from roadmap to PO reality. Plus the memory WFE lag read-through.
KLAC — Named Micron capex beneficiary, and stochastic defectivity flagged even at 24nm CD holes — dose as the yield governor at leading edge means metrology/inspection intensity keeps rising. Core franchise, same lag discipline.
WDC / STX — KV-cache offload expanding SSD TAM is a MODERATE NEGATIVE long-duration read-through for performance-sensitive HDD tiers. Explicitly not a near-term earnings downgrade. Relative trade, not a short thesis. Watch nearline commentary over the next two prints for displacement evidence.
SIMO — AI boot drives GREW 1600% LAST QUARTER, sampling with Google, going into DC racks. Raised a $1B 0% CONVERTIBLE to buy NAND — locking prices ahead of the +15-20% Q4 contract move. A company using its balance sheet to arbitrage the shortage it sells into. Smart.
FTAI — Initiation frames Power as a free option: reverse SOTP implies the market pays NOTHING for the aeroderivative turbine business against a ~15GW 2027 NA power shortfall. GEV booked to 2032 with cash commitments; grid power takes ~5 years. Base case ~10 MOD-1 deliveries in 26Q4, ~80 in 2027 vs ~70 consensus, worth ~$82/share. $1.465B Power order already announced; ~10.5x forward EV/EBITDA with no re-rating yet.
SMCI — Lost a large enterprise Vera Rubin deployment to a channel systems integrator — "big loss for SMCI." Consistent with rack-scale value migrating toward integrators and neocloud-aligned channels as deployments get more complex. The integration value-add is leaving the box-maker.
QCOM — Announced FlexCache for Snapdragon 8 Elite Gen 6 a month ago; Apple already shipped the same concept in M6. Being second to ship your own announcement is a small but real cadence signal. Bigger overhang: memory shortage squeezing smartphone units (30-40% industry downside risk flagged).
MRVL — Demonstrated 2nm chips for both 1.6T C/L-band ZR and O-band coherent-lite at ECOC — "AI optics is not a CPO-only story." Positioned on multiple architecture paths while the slow-and-wide vs fast-and-narrow debate plays out. Optics demand evidence keeps stacking (substrate prepays, PCB export prices).
TSEM — Optics exposure via SiGe BiCMOS laser drivers/receivers (long-dominant merchant share) plus mature high-yield hybrid bonding already at volume in image sensors — a close technical parallel to where optics packaging heads. Risk only if DSP absorbs into FinFET.
TSLA — Sell ratings are SHRINKING after the 2026 drawdown — analysts won't say sell anymore. That's sell-side capitulation to the Musk AI narrative, not fundamentals; historically that投降 shows up near bottoms, not after them. Sentiment state marker, not direction. France Sept registrations +61.86% is the only hard data — one market, one month.
NFLX — Sarandos: the company "ISN'T GROWING AS FAST AS I WANT." Rare public dissatisfaction. Live programming eats 5% of the $20B content budget, drives 1% of viewing. Management talking down its own growth is a yellow flag for a stock priced on engagement durability.
RKLB — Largest commercial Electron deal by mission count: 20 LAUNCHES FOR SYNSPECTIVE (2028-2031), Synspective to 47 contracted, total backlog above 100 MISSIONS. Value undisclosed, but backlog depth at this cadence is the revenue-visibility story small-launch investors wanted.
OXY — Goldman to Buy, PT $63 TO $69, structural improvements plus a $4B cash flow improvement target by 2030. Brent back above $100 on genuine supply shocks (China export pause, Russia diesel ban) is the macro tailwind. Oil's second inflation leg is the same force pressuring long bonds — energy is the hedge side of that trade.
DELL — Supplies rack-scale AI infra for the $15B+, 400MW Chiba AI DC near Tokyo — Apollo-financed, JERA behind-the-meter gas, phased 2028 to full 400MW by 2029. Part of Japan's ~$140B five-year AI ambition. Government-backed capex is filling the financing gap banks are leaving — Dell is the box beneficiary.
DASH — Agent-mediated ordering math now quantified: $0.02-$0.20 compute per order vs $1.70 contribution profit — the 13.5% take-rate model absorbs it, but ~$0.30/ORDER AD REVENUE (~$1B ANNUALIZED, 7% OF REVENUE) never shows up in an agent feed. Watch net revenue margin, not order growth, as agents scale.
NET — OpenAI's Dots cloud agents use Cloudflare for egress, same as Grok Bot. Small detail, real read: agent architectures generate compounding growth in traditional compute/networking/storage, and egress is NET's meter. New workload class landing on the network.
ANET — Feed cites Arista expected to "double revenue by 2028 again." One line, no sourcing — treat as sentiment confirmation within the networking-strength complex. Directionally consistent with everything else in the AI networking stack this window.
ARM — NVDA's new agent-safety platform pairs OpenShell controls with BlueField hardware monitoring, optimized for Vera — an ARM-architecture platform. Agent governance is becoming a product category, and the plumbing runs on ARM. Footprint expands by default.
GFS — Got used as the cautionary exhibit: "what happens without strong industrial policy and the risks of splitting design and manufacturing." That's an INTC-bull framing, not a GFS data point — but being the cautionary tale in someone else's thesis is not where you want to be.
ADSK, AKAM, APPF, APD, BE, BRZE, BURL, COP, GNRC, HII, JCI, GLNG, LOAR, MOD, NICE, NTAP, OOMA, PSQL, SEDG, SNOW, SPCX, TSN, WDAY — No signal in the feed this window. Nothing to act on.