Good morning.
NVDA +3% PRE-MARKET on China approving 200K H200s for DeepSeek, ByteDance, Alibaba — the real story isn't the approval, it's that Beijing signed off on training chips for inference-first players. ANET closing in on ATHs, MU and SNDK green on the memory cycle re-read. S&P and NDX flat to slightly positive. AVGO +5% AH after Apple's >$30B multiyear commitment through 2031 — locks the ASIC franchise and pulls the rug on MRVL in custom silicon land. Asia mixed: China semis up on the NVDA news, KOSPI semicaps getting smoked (5-15x 2027E vs. ARM/ALAB at 90x).
Three themes driving the tape today.
One: Memory BOM squeeze hits the consumer supply chain. PC OEMs warning of 15-20% DRAM/SSD hikes right when AI PCs need 32GB+ configs. Android OEMs (Xiaomi, Transsion) eating the margin pain — Apple insulated via procurement scale. MU/SNDK structural tightness thesis intact, but the derivative pain in HPQ/DELL and smartphone names is real.
Two: Custom ASIC consolidation favors AVGO over MRVL. Apple deal de-risks Broadcom's AI revenue through 2031 (already $10.8B Q2, guiding to $16B Q3). Marvell gets the relative multiple compression — market now pricing AVGO as the default ASIC partner for hyperscaler scale, MRVL as the second call.
Three: AI infra valuation divergence is absurd. Korean semicaps and AI plays at 5-15x 2027E, posting 30%+ growth, while ARM and ALAB trade 90x. That gap closes eventually — question is whether the degen wipeout in Hynix/Samsung flows reverses before multiples compress on the US names.
We'll hit up NVDA, AVGO, and MU first, then get to ANET, INTC, and the consumer supply chain pain trade in DELL/QCOM/SWKS.
MTZ is a buy. The Superior acquisition is exactly the kind of bolt-on that gets a PM’s attention — accretive, strategically coherent, and priced right. Stock’s off 8% in the last week (from a 113% run) so you’ve got a bit of entry pullback to work with.
$1.65B total consideration. $1.175B cash + $475M stock. Plus a 36-month earnout. Implies 6.9x 2026 EBITDA — cheap for a business where 90% of revenue is data center (70% hyperscaler). Superior is a self-performing IBEW electrical contractor with ~3,000 employees, a $1.4B backlog, and a 300k sq ft prefab facility. Revenue growing >40% in 2027 at midpoint, with low-teens EBITDA margins (14-15%). This folds into MTZ’s Power Delivery segment and should lift segment margins from ~9% to low double digits.
Three firms weighed in, all positive. The narrative is unified: this completes the data center framework MTZ telegraphed at its May Analyst Day. Cantor has the highest PT at $545 (Overweight), Mizuho raised to $502 (Outperform), Stifel holds at $455 (Buy). The cluster is $455-545 — wide spread but all above current price (~$365? need to check, but they're all above). No one is bearish post-deal.
"The acquisition completes the multidimensional data center framework MasTec outlined at its May 12 analyst day." — Mizuho
That’s the key line. MTZ is executing on its own roadmap, not chasing.
Bull case: MTZ is riding a generational infrastructure cycle — AI power demand, grid modernization, data center buildout. Superior adds immediate scale in the hottest end-market (data center electrical) with a self-performing model that drives margin. Revenue growth in the 20%+ range, backlog up 28% YoY, and this deal is immediately accretive. The 6.9x multiple is cheap for a 40%+ grower.
Bear case: Integration risk is real — folding a 3,000-person shop into a legacy union contractor isn't frictionless. The earnout could create misaligned incentives. And 8% decline in a week suggests the market is already pricing in perfection. If AI capex pauses or hyperscaler spending slows, MTZ’s data center exposure becomes a liability, not a moat.
Buy the pullback. The deal makes MTZ the go-to electrical platform for hyperscalers in the Midwest/Mid-Atlantic, and the valuation is reasonable for the growth profile. Keep an eye on Q2 print for margin trajectory in Power Delivery, and the earnout structure (36-month window) to see if management hits the high-end of the revenue guidance.
Verdict: Rocket Lab is executing a land grab. The Iridium deal is the thesis — vertical integration, recurring cash flow, and a credible path toward SpaceX-lite economics. The market is still digesting the premium paid (16.8x 2025 EV/EBITDA) and the dilution mechanics, but the long arc here is about owning the stack: launch, satellites, and now connectivity. Two analyst confirmations this morning reinforce the story.
Rocket Lab is buying Iridium at $54/share in cash + stock, ~$8B enterprise value. Jefferies notes that’s a "modest premium" to Iridium’s $51.09 pre-close (the stock was already up 178% in six months on speculation). The combo gives RKLB:
Morgan Stanley reiterated Overweight / $105 PT but raised its bull case to $293 from $185 (a 58% jump) to reflect launch + connectivity upside. They see a positive risk/reward skew of 4.8x. Jefferies doesn’t formally price the stock in the excerpt, but the tone is constructive — detailing the strategic merits without a downgrade.
Both are converging on the same narrative: RKLB is replicating SpaceX’s model with a smaller balance sheet and a different starting point. MS put it best:
"While SpaceX has significant scale and cost advantages over Rocket Lab, the firm sees Rocket Lab’s commercial progress and Iridium acquisition as moving the company toward a SpaceX model."
That’s the long-term bull case. The near-term tension is valuation. MS’s $105 PT implies downside from current $83.41 (stock fell 16.7% in the past week, likely on dilution fears). The bull case at $293 is a multi-year view.
BULL: RKLB is building the only publicly traded vertically integrated space platform with recurring revenue. Iridium’s cash flows de-risk the path to profitability. Launch cadence is increasing (12 missions in 2026, NASA contracts, VICTUS HAZE demo). At 4.8x risk/reward, the asymmetry favors longs for investors with a 2-3 year horizon.
BEAR: You’re paying 16.8x EBITDA for a satcom asset that already ran up 178% pre-deal. Integration risk is real — RKLB has never managed a constellation of this scale. SpaceX’s Starlink is a decade ahead on cost and coverage. The $8B stock consideration dilutes existing holders meaningfully. The 16.7% weekly drop suggests the market is pricing in execution complexity, not potential.
This is a show-me story with a 2027 close date. The strategic logic is sound — buying spectrum, subs, and cash flow to plug into a launch/mfg engine. But the stock’s near-term trajectory depends on messaging around dilution and margin visibility. MS’s $105 PT is the floor for now. The bull case at $293 is aspirational, not actionable today. Watch the next quarterly call for Iridium revenue contribution guidance.
This is an AI infrastructure name that's re-rated hard, and the analysts are chasing. Both Citizens and Stifel raised PTs this morning after the Q3 beat-and-raise. The stock is at $62.71 but getting pulled along by the narrative. The question is whether the multiple has room to run.
Q3 was a clean beat. PENG printed $0.84 EPS on $479M revenue vs consensus of $0.56 and $421M. That's a 48% YoY revenue growth rate. They raised full-year guidance and gave a preliminary F2027 outlook calling for 30% growth in both revenue and non-GAAP EPS.
Two analyst moves:
Bull case: This is a pure-play AI infrastructure compounder trading at 11x EBITDA with a PEG of 0.16 and 30% top-line growth incoming. The memory cycle tailwind plus AI factory platform strategy gives multi-year visibility. Stifel's quote sums it up:
"The results support Penguin Solutions' improved revenue outlook for fiscal 2026, with expected growth of 20% to 24% year-over-year, and fourth-quarter revenue projected to grow in the mid-single digits or higher sequentially."
Bear case: Stock is up 196% in the past year and down 9% in the last week — some profit-taking or positioning unwind happening. InvestingPro's Fair Value flags overvaluation at current levels. The 14-22x forward multiple the analysts are projecting assumes sustained triple-digit AI growth, which is a high bar if the memory cycle turns or hyperscaler capex normalizes.
Bottom line: PENG is now priced like a grower, not a value trap. The data supports the re-rating through F2027, but the rate of change matters more than the absolute level. PMs should watch the memory and AI factory pipeline prints for the next leg.
Verdict: Wells Fargo starts VZ at Equal Weight, $43 target — essentially a call that the stock is fairly valued here with a Starlink option embedded. The thesis is bimodal: VZ has the most to lose AND the most to gain from an MVNO partnership with Starlink. The new CEO may be more open to the deal than prior management.
"Verizon has the most to lose & the most to gain from an MVNO with Starlink, & the new CEO may be looking at things differently."
Wells applies a 40% probability to that outcome. The rest of the setup is pedestrian: 2026 self-help improves subs and churn, but 2027-28 sees industry FWA and postpaid headwinds. EBITDA estimates ~1% below Street, EPS 3-6% below in 2027-28. Leverage stuck around 3x, buybacks lower than cons. The 6.64% dividend is the floor, but share price upside is capped absent the Starlink catalyst. The JV with BT (loss on sale of $700-800M) is a non-event for the wireless thesis. Net net: a hold with a lottery ticket.
Neutral with a $110 PT, but the stock is in the penalty box post-SYNA grab. Cantor's call: the setup is better than it looks, but the market needs to see the acquisition thesis actually land before giving credit. They model 42% gross margin exiting the year (already at 42.67% LTM, so maybe conservative) and a $6 EPS stretch for 2028 supporting the pullback. 27 analysts revised EPS up — so the fundamental narrative isn't broken. But the SYNA deal threw cold water on the consensus long that was building. Now you've got a messy mix:
"Cantor Fitzgerald is more positive on the setup... but ON is potentially in the penalty box following the acquisition announcement."
The bull case (Needham $130): SYNA gives them system-level breadth, margin upside from Fab Right, and a cleaner long-term story. The bear case (UBS $95, Stifel $107): Integration risk, no preannouncement, and the DC power opportunity is still mostly narrative. ON sits at $91.10 — that's 21% upside to Cantor's PT if they deliver, but the stock needs to earn that back. Show-me territory, not a conviction long right now.
Cantor Fitzgerald doubles down with a $550 PT, the Street high. That’s 45% upside from $379 — and they’re not just slapping a number on it. The thesis is sharp: under-appreciated power opportunity (Empower Semi acquisition now closed), strong AI spending leverage, and a valuation that’s still cheap at 19x CY28 EPS. This thing has already ripped 57% in the past year, but Cantor thinks estimates still have room to run.
"The firm sees meaningful upside potential to current estimates for the semiconductor company."
The Empower deal isn’t just bolt-on M&A. It gives ADI a direct line into AI processor power delivery — a market that’s about to explode as datacenter power demands go vertical. BofA just called ADI a top pick in AI power semi alongside TXN and ON. Argus raised to $460 on broad-based revenue growth, but Cantor is swinging bigger. The bet: power content per AI rack multiplies, and ADI’s analog muscle finally gets the growth premium it’s been missing.
Multiple analysts are now pulling in the same direction — 24 EPS revisions higher in the last period. The bull case is cyclical tailwinds (industrial, auto bottoming) layered with secular AI power demand. The bear case? Execution on the Empower integration and whether 19x forward is actually cheap if the cycle turns again. Right now the r/r tilts bull.
HOLD AT $59. Freedom Capital fires the opening salvo on TEM with a Hold and a $57-handle price target (stock at $57.13, $10.8B mcap). Thesis is simple: healthcare AI is real, growing like a weed (revenue +70% LTM), but the public market isn't pricing it right yet. Too much of the generative AI spend is flowing to startups, not incumbents like TEM. The whole healthcare IT group is trailing the S&P 500.
Freedom leans into the structural story — $5T in healthcare spend (18% of GDP), 25-30% going to admin, workforce shortages squeezing everywhere. Physician AI use has more than doubled since 2023. 66M Americans have used AI to search health info. The pieces are there.
But the r/r is neutral here. Not a short, not a slam dunk long. You're paying for optionality on the 37-43% CAGR healthcare AI market, with TEM as one of the few public pure plays. That deserves a valuation premium — just not a massive one until we see sustained margin progression or a catalyst that breaks the "startups win" dynamic.
“The majority of generative AI spend in healthcare is flowing to startups rather than large incumbent healthcare IT companies.”
That's the tension. TEM has the data moat (600 de-identified angiosarcoma records, FDA-cleared AFib software, the digital pathology consortium with MSK and Yale). But public healthcare IT underperforming the SPX suggests the market still sees more fastball risk than upside. Canaccord still Buy at $80, FWIW. For now, Hold feels right — wait for a better entry or more proof of unit economics.
Cantor reiterates Overweight / $125 PT on MCHP – but the real question is what breaks the current narrative. Stock trades at $84.15 (+33.5% YTD) after being the worst analog performer. Investor positioning is still underweight, and the Aug 4 earnings setup is “not demanding.” Cantor likes the beat-and-raise odds but acknowledges the market needs more than a routine print to re-rate this name. Potential catalysts: better data center revenue clarity, a major PCIe customer win, or clearer DC power exposure. EPS upside tracking to $6 CY2028. The catch? Data center is only ~7% of total revenue right now (J Capital flagged this yesterday), though product ramps start H2.
“What would change the broader narrative for the company?” – Cantor Fitzgerald, noting that a beat/raise alone may not sustain momentum without a structural shift in how the Street views MCHP’s industrial/auto-heavy mix.
Quick reminder: MCHP also made its MPLAB Pro compilers and ML dev suite free (previously paid licenses) – a smart move to lock in embedded developer mindshare. And a new rad-tolerant clock generator for space/aero keeps the defense angle alive. But the real debate is whether the next leg comes from DC traction or just another “good enough” print that fades.
Bullish signal from Evercore, but it’s a 5-year structural call, not a Q2 catalyst. Reiterated Outperform, $365 PT. The headline is the expanded Broadcom deal — a multi-year commitment exceeding $30B for custom silicon and wireless components. That includes a $1.5B capex boost for Broadcom’s Colorado fab and runs through 2031. Evercore sees this as Apple locking up strategic supply chain capacity before the next wave of component inflation hits. The market barely reacted (AAPL flat-ish pre-market), which tells you this is already priced into the supplier narrative — more AVGO than AAPL.
"The agreement is another example of Apple proactively locking up strategic parts of the bill of materials at a time when broader component availability and cost inflation remain a concern."
Still, the deeper read: Apple’s move in-house on connectivity silicon (N1 chip, modem roadmap) doesn’t preclude Broadcom from being a core partner. The deal specifically covers RF, FBAR filters, and ASIC for multiple product generations — think custom silicon for future iPhone/iPad/Watch radios. Evercore’s take is correct: it’s a risk-reduction move on cost and supply, not a margin giveaway. Long-term structural positive, but don’t expect any earnings sensitivity from this for at least 4-5 quarters.
Initiating at Equal Weight with $170 target — and the stock is already at $185. Wells Fargo isn’t bearish, but they’re outright saying the easy share gains are over. The two headwinds they flag: Verizon stops bleeding market share, and Starlink eats into incremental industry growth.
TMUS still has the best earnings trajectory in wireless — just not the same rate of change that got it here.
“T-Mobile should remain the strongest earnings grower in its sector but with less upside compared to its historical performance.”
That’s the whole debate in a sentence. The stock’s 20x earnings, $200B market cap. The multiple expansion phase is done unless something structural changes (spectrum, fixed wireless, AI edge). For now, they’re a steady compounder with a ceiling, not a re-rate story.
Verdict: Neutral initiations and a stock down 45% in six months tell you everything. DA Davidson starts at Neutral with a $5 PT (current $4.65, near the 52-week low of $3.81). The thesis here is grim but straightforward: engagement headwinds in North America, ad ARPU expansion failing, and a margin lag that keeps them unprofitable ($0.24 net loss per share). Benchmark, Truist, and Piper Sandler all kept their Hold/Neutral ratings post-Specs launch — nobody is pounding the table.
The bull case is thin but real: revenue grew 10% LTM to $6.1B, they're monetizing users via subscriptions, and Specs is a legitimate AR bet (100k units at $2,195 a pop targeting developers). But the bear case is louder — engagement problems don't fix quickly, ad budgets are flowing to Amazon/Google/TikTok, not Snap. The digital ad backdrop is healthy (Evercore ISI noted strong Q2), but Snap isn't capturing it.
"Ongoing engagement headwinds in North America and continued challenges in driving advertising average revenue per user expansion will weigh on the company in coming quarters." — DA Davidson
Don't fight the tape. This stock needs to show user engagement stabilizing and margin improvement before it earns a multiple. Specs is a long-duration option that won't print for quarters.
Verdict: Initiation from DA Davidson adds another buyer to the wall of worry. $26 PT, 16% upside from here. That’s not heroic, but it’s consistent with what we’re hearing — engagement compounding (10 straight quarters of YoY user growth), ARPU creeping higher mid-to-high single digits, and the stock getting absolutely clubbed on multiple compression. The thesis is straightforward: the operating metrics are fine, the valuation is not, and eventually the narrative catches up.
“Pinterest’s valuation has declined despite consistent engagement growth and ARPU expansion — the market is pricing in competitive risk that may be overdone.”
Consensus is fragmenting though. The analyst PT range is $21 to $42 — that’s a huge spread for a $22 stock. UBS at $30, Guggenheim at $24, TD Cowen at $38. The bull case (TD Cowen) sees Pinterest as a “leading mid-cap idea” driven by AI-powered ad improvements and user momentum. The bear case? Competition for ad dollars and user attention from Meta/TikTok — but those headwinds have been priced in for 18 months. If PINS can keep printing 16% revenue growth (last twelve months) and expand margins, the re-rating is the trade. Not sure we need to get aggressive yet, but the risk/reward tilts long if they don’t screw up the next print.
Wolfe upgrades AMT to Outperform from Peerperform, PT $188 (current $165, ~27% upside). The call is pure thesis shift: U.S. tower market has finally digested carrier consolidation — no more structural overhang. They see a return to predictable organic growth.
The valuation hook is the PEG ratio at 0.45 — dirt cheap for a large-cap REIT if growth stabilizes. Wolfe isn't alone either; this adds to a cluster of recent upgrades (RBC, Bernstein, Truist all turned bullish). The consensus is congealing: concerns on DISH churn and satellite disruption are overdone.
"The shift sets the stage for more predictable organic growth for the company." — Andrew Rosivach, Wolfe Research
Not pricing in perfection yet, but the r/r is finally leaning long.
Bernstein still sees a path to $100 despite cutting their PT from $110. They're not waving the white flag — maintaining Outperform — but they're calling out 2026 subscriber pressure HARD. They slashed 3M subs from their 2026 estimate. Stock at $76.18, down 40% in a year, down 19% YTD. The bull case rests on 2027, where they expect a reacceleration of +4M subs.
"Bernstein sees a reacceleration in subscriber growth in 2027 driven by ad-tier expansion into 15 new markets."
Multiple compression to 26x from 29x is the price of admission here — they're baking in real near-term pain. Not alone in that view: Wells Fargo (Equal Weight, $105) flags revenue growth deceleration on slowing engagement. Citizens (Market Perform) says current estimates already price in a 2027 price hike — limited upside. The broader narrative keeps getting muddier: lost the Roku bid to Fox ($20B), rumored to be sniffing around Lionsgate, but no formal offer. M&A activity increasing but no wins yet. The signal here is that NFLX is being forced to pivot — buying scale or partnering (TF1 deal) — because organic sub growth is hitting a wall. Bernstein's 2027 call is a bet on ad-tier scaling, not a guarantee. PMs should watch engagement data and ad-tier adoption as the rate-of-change needle movers, not just sub adds.
Cantor Fitzgerald is holding firm on Overweight / $400 on NXPI, arguing it’s the single best value in analog semis — a 17x NTM P/E vs. a peer average of 28x. They see numbers grinding higher post-earnings, with CY27 EPS tracking north of $18 (consensus $17.66) and CY28 closer to $20.50. The $400 target = 21x discounted back CY28 numbers.
The car exposure narrative is a headwind, but Cantor thinks the through-cycle resiliency and secular growth (SDV, radar, connectivity) are still underappreciated. Multiple other firms (Truist, Barclays, Cantor itself earlier) lifted PTs to $310-340 range, so Cantor at $400 is the outlier bull.
“The company's through-cycle resiliency and secular growth opportunities remain under-appreciated by investors.”
Separately, NXPI declared a $1.014 interim dividend (payable July 9) and shuffled IR — Mike Lucarelli taking over for retiring Jeff Palmer. Nothing moves the tape, but signals stability.
NORTHLAND STAYS LONG THE AI-AT-THE-EDGE STORY, RAISING PT TO $86 (from undisclosed prior) on 12x CY27 revenue. Stock at $75.48, down 14.6% on the week but still +165% YTD — the multiple reflects the premium for a company that’s proving out design wins in watches, fitness trackers, rings, eyewear. The firm sees revenue from medical/industrial/smart home segments more than doubling in 2026.
"Ambiq’s revenue from medical, industrial, and smart homes and buildings markets is expected to more than double in 2026."
The thesis is straightforward: ultra-low power is the enabler for AI inference at the edge, and AMBQ is winning sockets across an expanding list of wearable form factors. R&D spend is ramping aggressively to extend the moat. At 12x forward revenue, you’re paying for the growth to materialize — so execution on the 68% FY26 growth bogey is non-negotiable. The recent equity offering ($179.4M gross) adds dilution but also balance-sheet runway for the investment cycle.
VERI IS A MONSTER r/r BUT THIS STOCK HAS LEARNED TO HURT YOU. Needham reiterates Buy with a $10 target (stock at $1.25, screaming distance from the $1.22 52-week low). That's an 8x implied return on a name trading for less than a cup of coffee. The catalyst? Product updates, an Oracle partnership, and the VDR revenue ramp supposedly landing in H2 2026.
"Needham expects the company to ramp VDR revenues in the second half of 2026 as anticipated."
The bull case rests on two things: (1) Veritone has closed Google and Nvidia as clients — real validation, not vapor. (2) The business reorganization is supposed to align the cost structure with this revenue event. If VDR hits, the stock is repricing to something that doesn't look like a distressed balance sheet.
But the bear case is just as obvious. Revenue came in at $20.26M vs a $29.31M bogey — that's a 31% miss. Negative $51.8M in levered FCF over the last twelve months. UBS started coverage at Neutral with a $2.50 target (down from $6), explicitly citing "complexity of forecasting deals" as a path-to-profitability concern.
Bull: This is a deep value AI play with blue-chip partners and a product refresh. The VDR ramp is real, the Oracle deal adds distribution, and the Russell inclusion (effective today) forces passive buying from funds that couldn't touch this before.
Bear: Missed the bogey by 31% and we're supposed to believe H2 2026 is different? Cash burn is still horrid. The reorganization hasn't produced results yet. Hyperscalers as partners sounds nice until they decide to compete.
Verdict: Hard pass at 10x multiple confirmation. Let the VDR revenue print before you catch this knife. The r/r is optically attractive but the execution gap is massive. Either Needham is right about H2 or this thing tests $1.00. Not buying the thesis without a quarter of proof.
(Charter: this works or it doesn't — no in-between.)
Verdict: UBS is leaning into the Google Cloud partnership as the narrative catalyst — and they’re not wrong. ACN at 11x earnings after a 52% haircut is pricing in a lot of pain. This deal opens up midmarket TAM ($240B growing high single digits) that the Street has largely ignored while fixating on large-enterprise slowdown. The setup is asymmetric.
UBS reiterates Buy, $275 PT (vs. $141 close). That’s ~95% upside. The math works if the AI deployment wave actually scales — and the partnership gives ACN a differentiated product bundle (Gemini, Agent Platform, Data Cloud, Threat Defense) that targets the $300M–$3B revenue segment through their new Edge unit. It’s a classic “sell shovels to the gold rush” thesis, but applied to a market that has historically been under-served by Big Consulting.
“The announcement is evidence that AI deployment is moving beyond experimentation toward scaled enterprise adoption.”
The thesis isn’t complicated: ACN’s core large-enterprise business is mature, but midmarket is underpenetrated. The Google Cloud hook gives them a pre-built, industry-specific suite that lowers the friction for smaller firms to hire Accenture. If they execute, this isn’t just a stock that re-rates — it’s a business that re-accelerates. The 4.6% dividend and 11x earnings provide a floor. The risk is execution (midmarket clients are less sticky) and macro (they still need to spend). But the r/r is the best it’s been in years.
Verdict: PANW is the platform consolidation story that keeps delivering, and the channel checks are getting louder. Evercore just ripped its PT to $415 from $320 (52x CY27 FCF) based on initial partner feedback showing sustained platform momentum, stronger identity conversations with CYBR, and continued traction on SASE and Cortex. The firm sees PANW sitting at the intersection of identity, observability, and AI-powered security ops — and thinks vendor consolidation + platform expansion drives durable growth and FCF expansion.
"Partner feedback was broadly constructive for Palo Alto Networks… highlighting sustained platform momentum and stronger identity conversations with CYBR."
The broader analyst crowd is piling on: Needham went to $425 (highest), FBN to $330, Cantor to $340, and William Blair bumped FY26 FCF to $4.2B on improved cash conversion. That’s 13+ firms raising into a stock already up 83% YTD — and 41 analyst EPS revisions higher. The bull case is straight-forward: the platform flywheel is accelerating, and the premium multiple is earned. Bears will argue the stock is pricing in perfection at ~52x FCF, but the rate of change in partner talk suggests there’s still room to run. (Not sure we can read much into a single check, but the trajectory is unmistakable.)
EVERCORE ISI JACKS PT TO $500 — STOCK AT $417, UP 234% YTD, AND THEY THINK THERE'S MORE. Demand for AI infra is still outstripping supply, and Dell’s management sees supply constraints getting more acute in CY27 (DRAM/NAND primary, but CPUs, HDDs, optics becoming new bottlenecks). The firm now has 5,000+ enterprise AI factories in production — up 50%+ in six months — with repeat purchases accelerating.
"Dell executives said supply constraints are expected to become more acute in calendar year 2027 compared to 2026, with DRAM and NAND as primary bottlenecks."
Storage is the key margin lever: Dell is pivoting away from third-party storage to its own IP, and management explicitly says storage is the most profitable way to deploy NAND allocation. On-prem inference thesis is gaining traction — lower cost per token when workloads stay close to data, less reliance on frontier models.
Not all rosy — GF Securities downgraded to Hold on valuation after this rally. Fair point — 234% YTD is a lot of good news already in the price. But with supply constraints worsening and enterprise AI spend sticking, Evercore sees the run continuing.
BABA is a contrarian bet at $98 — down 36% in 6 months, within a hair of its 52-week low. Jefferies keeps Buy with $185 target (nearly 2x upside). The thesis rests on AI-driven cloud acceleration as the real catalyst, not the stagnant core e-commerce. AliCloud revenue is already RMB9B (30% of external cloud per Bernstein), and Jefferies expects Y/Y acceleration plus sequential margin improvement. QC fundamentals came in better than feared. The macro noise (flat EBITA for China e-commerce / AIDC) is discounted. Multiple analysts raised PTs to $195–220 cluster.
"For AliCloud, the firm expects it to accelerate year-over-year and perform better than expectations on strengths in AI demand and MaaS."
Balance sheet is fortress-grade (more cash than debt, per InvestingPro). If the AI narrative gains traction, this is a $1T+ market cap story trading at 20x earnings. PMs should watch the cloud margins print next quarter — that’s the rate-of-change signal.
Canaccord sees 100%+ upside here, bumping PT to $11 from $10.50 (still Buy). Stock already ripped 76.9% in the past year but sits at $5.33 — so the target implies another double. Thesis is pure tailwind: the drone & counter-UAS market gets a $74B shot in the arm from planned reconciliation legislation, plus $70B passed last month for DHS. Canaccord slaps a 3.1x EV/Rev on FY27 revenue of $142M (includes Vecima). That gets them to $11. They think enterprise IoT comps trade at low double-digit to mid-teens EBITDA (Itron at 11.6x is the bogey).
"The Vecima industrial IoT acquisition provides an opportunity to sell Lantronix modems into the acquired customer base with expected margin improvement."
Bear case? The public offering of 4.17M shares at $7.20 (gross $30M) dilutes existing holders, and the $5.33 price is still well below that — not a great look for momentum. Execution on cross-sell is unproven. But the funding wave is real, and LTRX is putting hardware in the path of that spending. R/r favors bulls as long as the drone narrative stays hot.
Amazon is taxing its balance sheet to build AI infrastructure because demand is "continuing to outstrip supply" — and Needham sees that as the cleanest read-through for the stock. They're sticking with Buy / $300 PT (current $246, Street range $207-$370). The $25B debt raise tells you AWS compute demand is real, not aspirational.
The more provocative edge: Needham wants Amazon to stop funding competing projects and funnel ALL capital into AWS. Their logic — AWS has the highest ROIC by a mile, and falling behind in AI compute is the biggest downside risk in the story.
"Amazon should cut competing projects that use capital in order to allocate capital solely to AWS capabilities and products. These have the highest return on invested capital and the greatest downside valuation risk if Amazon falls behind."
Worth noting: AWS just instructed supply chain partners to boost Q3 shipments 30% off Trainium 3 demand — that's Taiwan suppliers gearing up. The narrative isn't a spot-check, it's a rate-of-change acceleration on the hardware side. Balance sheet can handle it (13% ROIC, 0.53 D/E) but every dollar spent on non-AWS projects is a dollar not compounding in the highest-return bucket.
Cantor says Overweight, $177 target — but the story is bigger. DOCN pre-announced Q2 and it's a blowout. RPO >$800M (10x YoY), revenue growth accelerating to ~29% (vs 14% last year), and they signed multiple nine-figure inference + cloud commitments. The kicker: 20MW incremental capacity for late 2027/early 2028 takes total to ~155MW. This is the AI inference pivot finally showing up in the numbers.
Stock's already up 365% in a year to $137. So the r/r is about how much of this acceleration is already priced. Cantor's $177 target implies ~30% upside, which feels achievable if that exit growth rate north of 30% materializes. But PMs should ask: is this a 10x revenue multiple on 2027 numbers becoming a consensus story, or is there still room?
"Management attributed the momentum to multiple nine-figure annual customer commitments for inference and cloud products signed in the quarter."
The pre-announce removes downside risk ahead of formal earnings. Biggest binary: can DOCN sustain this pace of multi-tenant capacity adds without diluting margins? Cantor raised 2027 estimates and introduced 2028 above Street. Worth watching the next quarter for churn on the legacy SMB base — that's the offset nobody talks about.
NVDA — Still the core AI bet, but the signal mix is messy. China H200 approval for DeepSeek, ByteDance, Alibaba adds near-term demand (200K units) but forces inference onto domestic processors — regulatory bifurcation is the new reality. CPO for scale-out in production, Kyber/Rubin Ultra roadmap confirmed intact per IR. Trading below 20x forward — cheapest since early 2019. Macro and positioning flows dominate, but demand signals remain robust. Key debate: is this a 2019-style trough or structural de-rating?
MU — Memory supercycle thesis intact despite the 22% drawdown. Contract DRAM up 20-30% QoQ, NAND +35-40%. HBM4 revenue already >$1B. The gap between contract and spot (especially HBM) still leaves catch-up room. Positioning unwind from Korea retail leverage and Hynix ADR overhang is real, but fundamentals are improving. Key debate: healthy correction or cycle peak?
AVGO — Apple’s >$30B multiyear deal locks RF/custom ASIC through 2031 — a step-function de-risking event. AI semi revenue accelerating (+143% YoY, Q3 guided +200%). Broadcom is consolidating the highest-value custom programs (negative read for MRVL). Fort Collins expansion is small for WFE but signals domestic supply chain deepening. Key debate: does this lock AVGO as #2 semi franchise behind NVDA?
INTC — Foundry customer wins are real: Apple M/A series, SpaceX AI ASIC, Marvell network chip, Google TPU (EMIB). Capex inflection to $25-30B by FY27 implies serious buildout — consensus is dramatically light. Xeon pricing power (+12% hike to $13,955) shows DC CPU strength. But TSMC still got Apple’s AI server chip (Baltra on N3P). Key debate: viable #2 or niche foundry?
ANET — Ethernet AI networking thesis reasserting near ATHs. EOS software creates a durable moat in campus and DC switching. CPO scale-out from NVDA and TSMC PIC ramp are tailwinds. Market cap $228B; long-dated calls in play. But macro risk-off could drag even strong names. Key debate: is the Ethernet share gain thesis fully priced?
META — Massive 1GW Canada DC ($9-10B) underscores AI capex commitment. Foundational model ownership integrated with Advantage+ creates a durable ad-tech advantage. Renting excess compute to neoclouds (200MW Anthropic deal) is optionality, not overcapacity. Key debate: rational capex or overbuild?
CRM — Anthropic emerging as a direct competitive threat in enterprise deals. Sellers flagging it alongside AWS, Microsoft, Databricks. Agentforce adoption not yet broad-based per sentiment proxy. Key debate: losing the AI agent war?
GOOG — Electricity use +37% Y/Y validates AI infra scaling. Spinning out DeepMind narrative persists (retain 49%, bring external capital). Google AI Studio’s GitHub integration improves developer accessibility. Model competition remains intense. Key debate: can Google regain AI leadership momentum?
AMD — Morgan Stanley models 2.4M GPUs this year, 2.7M next. Hot Aisle enables low-risk GPU eval at ~$600/week — developer flywheel building. CPO roadmap inclusion with TSMC COUPE. Software gap with NVIDIA remains the key debate.
MRVL — Broadcom is consolidating highest-value custom ASIC programs (Apple chose AVGO over MRVL for Baltra). Negative read for relative multiple. CXL memory expansion cards gaining traction in enterprise (Penguin Solutions wins) could offset. Key debate: find other marquee wins?
QCOM — Apple C1 custom modem accelerating share loss; Qualcomm’s modem share could drop from 100% to 20%. Growth must rely on Android, auto, IoT. Key debate: can they replace lost Apple revenue?
SWKS — Apple deepening RF relationship with Broadcom (Fort Collins expansion covers filters). 67-69% Apple revenue exposure — erodes incremental socket opportunity and pricing power. Key debate: diversify away from Apple?
QRVO — Same dynamic as SWKS. 46-47% Apple exposure. Broadcom’s Fort Collins expansion increases competition in RF. Key debate: maintain Apple socket?
ALAB — CXL memory controller gaining traction in enterprise (Penguin Solutions highlight). Trades at ~90x 2027 EPS — extreme valuation dispersion vs Korean semicaps at 5-15x. Key debate: premium warranted?
RMBS — CXL memory controller IP gaining traction; Penguin using RMBS in MemoryAI KV servers. Key debate: competing with ALAB in CXL controllers?
WDC — NAND contract prices up 35-40% Q3, but stock down 28% from highs. Contract still below spot. Key debate: buying opportunity or mirror of MU dynamics?
STX — Similar NAND/HDD exposure; down 24% from highs. Key debate: AI storage demand or legacy leverage?
LRCX — Most central stock in 30x30 correlation matrix (avg corr 0.593). WFE triangle tight. ASML EUV supply constraint may cap 2027 WFE. Key debate: market proxy or sector leader?
AMAT — Broadcom Fort Collins expansion immaterial to large WFE, but Apple’s American Manufacturing program includes AMAT. TSMC N3P and PIC capacity expansion benefit. Key debate: does US chip localization move the needle?
KLAC — Part of WFE triangle; high correlation with AMAT/LRCX. EUV supply constraint may limit 2027 WFE. Key debate: inspection spending grow faster than litho?
WULF — CFO aggressively pushed back against Jim Chanos short thesis (Chanos modeled $1-1.5M/MW maintenance capex, $8B D&A on $4.4B build). Stock +11%. Key debate: flawed short thesis or structural overbuild risk?
BE — Hunterbrook short report alleging China scandium dependence. Bloom’s 5GW goal would need ~220 tons/year (3.7x global market). Q1 2026 related-party revenue $373M of $751M total. Key debate: real risk or manageable supply chain?
LITE — Optical supply chain for AI networking. TSMC PIC capacity reaching 25K wafers/month by 2028, CPO adoption tailwind. Key debate: key beneficiary or peripherally exposed?
SNDK — NAND price strength but stock down from highs. Target >$2,700 from some analysts. CXL/threat to NAND demand growth is the medium-term risk but near-term pricing momentum strong. Key debate: selloff overdone?
ATCOB — No material new signals. Passing.