Good morning.
Memory complex ripping. DRAM +25% QoQ INTO 3Q, NAND +30-40% — HBM repricing +100-150% 1Q27. Binding constraint on AI itself. MU the clear leader, but SNDK, ASML also leveraged. AMD announcement this morning: Microsoft deploying Helios rackscale in 2H26 for frontier inference. That's the first real hyperscaler GPU commitment beyond NVDA — changes the competitive narrative. SemiAnalysis turning constructive. TSM printed $40.2B revenue, guiding 2026 growth SLIGHTLY ABOVE 40% YoY. CFO: “don’t plan to leave any food on the table.” MSFT disclosed $37B AI ARR growing triple digits — larger than OpenAI or AWS AI — and Yipit/M Science data accelerating into FY27. Defies the “excess compute” bear case. NOW beat and raised on pricing pull-forward, but AI monetization (Now Assist) still early — channel talking AICT, not usage. Asia macro: China open-source AI (Kimi K3, 3T-parameter sparse MoE) at frontier, Xi doubling down on open-source at WAIC. Kirin 9030 teardown shows 32.5nm metal pitch WITHOUT EUV — tighter than Intel 18A. That should spook the export-control crowd.
Three themes framing the day: 1) Memory is the new bottleneck — HBM pricing this extreme changes unit economics for everyone downstream (NVDA consumer GPU delayed on GDDR7 costs). 2) AMD’s Azure win is asymmetric — Helios rackscale is a real multi-generational commitment, not a pilot. Street will re-rate AMD’s AI share. 3) Data center leasing records keep coming — 9.6GW in 2Q, pipeline 12.5GW, OpenAI upsized to 30GW by 2030. Meta/BlackRock JV shows off-balance-sheet financing innovation. HUT Beacon Point fully commercialized at 1GW, $26.6B portfolio base-term. IREN cheapest neocloud at 0.18x EV/Planned Revenue. 4) China’s open-source AI + semicapability without EUV is a structural headwind for US export controls and a tailwind for memory/compute demand (more price-insensitive buyers).
We’ll hit up MU, AMD, and MSFT first, then get to the neocloud/data-center power complex (HUT, IREN, GEV).
Verdict: AAPL at $325, 3% off the 52-week high of $335. 59% gain in the last year – but the next 10 days are the test. Earnings July 30 is the immediate catalyst. The narrative is mixed: revenue beat looks baked in, but gross margin pressure, AI chip delays, and CEO succession hang over the print. DOJ settlement talks are a modest positive – but the scope of concessions matters. Net: positioning is long, crowded, and reliant on a clean print and a convincing AI story.
BofA has the highest PT on the street at $380. They see revenue of $109B (vs Street $108B) and EPS $1.89 (vs $1.87) – within AAPL’s own 14-17% growth guide. Product gross margins slipping to 36.8% in June and 34.1% in September before a recovery to 38.5% in December. Components are the culprit, plus the staggered iPhone launch (Pro/Max + foldable in Sep, base/Air in March at higher prices).
HSBC upgraded to Buy, PT $366. Jefferies still Hold at $300, citing China AI approval as already priced in. The Street is converging around the beat, but the real focus is guidance and any color on AI infrastructure spend.
Evercore kept Outperform, $365, after Bloomberg reported early settlement talks with the DOJ. Both sides are exchanging drafts – no trial date set. Key: the lawsuit wasn't about App Store fees or Google payments (those are separate). It’s about super apps, cloud streaming, iMessage, smartwatches. Evercore’s view: settlement avoids a multi-year litigation drag, but the ultimate impact depends on how many iOS concessions Apple has to make.
“Settlement could reduce regulatory overhang and potentially avoid a multi-year litigation path.” – Evercore
Not sure we can read too much into that yet – early stage, no guarantees. But the stock is already pricing in a favorable outcome.
Morgan Stanley reiterated Overweight, $360, but flagged the real risk: Apple’s M2 Ultra AI server chip is reportedly “struggling.” The Baltra ASIC is on schedule for small volumes in 1H27, with a second-gen in 2028 that could match merchant silicon. That’s a long wait. If Private Cloud Compute doesn’t have enough oomph, Siri AI either degrades or workloads get pushed to Google Cloud. Metrics to watch: SoIC bookings at TSMC, CoWoS bookings, second-gen ASIC tape-out, and Apple’s capex trajectory. Earnings call is the first real chance for Cook to address this head-on.
PMs are positioned for a beat. The real swing factor is whether Apple can convince the Street its AI infrastructure investments will bear fruit in 2027-2028, not just on-device gen AI. Any guidance light on gross margin or capex surprises will cause a rotation out of the name. Watch the TSMC and AI chip commentary closely.
VERDICT: ACQUISITION SPLITS THE STREET. Bulls see a $20B+ MRD TAM unlock. Bears see a distracted CEO throwing equity at a slowing core biz. Stock down 13.5% in a week to $48 — and the bull/bet spread is as wide as the bid-ask.
Tempus is buying the remaining 81% of Personalis (PSNL) for a net ~$1.5B in stock (cash opt-in on 50%). That's a 28% premium to PSNL's 30-day VWAP. Closing late '26 / early '27. NeXT Personal MRD test — tumor-informed, already reimbursed in four Medicare indications — is the prize.
Bulls (Needham, Canaccord) see this as the MRD piece Tempus was missing. Needham values the MRD market at $20B+ with <10% penetration today. Canaccord says it "optimizes" Tempus's internal tumor-naive efforts. Both reiterate Buy — $75 (Needham) and $80 (Canaccord) — well above current $49.
Jefferies is the lone holdout at $35 (Underperform). They flagged this exact deal at the May 29 analyst day — management said it would provide no net income benefit. Now they're doing it. Jefferies questions whether "the rationale for the acquisition is intended to offset a slowing core diagnostics business."
"We remain constructive on the science behind PSNL’s minimal residual disease test, but the timing and structure raise questions about capital allocation and near-term profitability." — Jefferies
BULLS AND BEARS ARE SHOOTING FOR TWO DIFFERENT PLANETS. DA Davidson cuts PT to $170 (Buy) on lower peer comps — still sees AI tailwinds. CLSA fires a bearish dart with an Underperform and a $72 target that implies 30%+ downside from here. Stock’s down 32.6% YTD, trading at ~18x CY26 FCF. The spread between these two targets ($170 vs $72) tells you what you already know: nobody agrees on where this thing settles. Earnings July 22 — options pricing in an 11% move.
DA Davidson cuts to $170 from $190 — maintaining Buy. The trigger: lower SaaS growth peer multiples. New target uses 30x CY26 FCF, which they call “more in line” with the group. No change to the underlying FCF estimates. They acknowledge the fear that AI CapEx is crowding out other IT spend, but push back hard:
“DA Davidson said it remains confident in ServiceNow’s opportunity to benefit from AI tailwinds across its entire business.”
CLSA initiates at Underperform, $72. Analyst Bhavtosh Vajpayee values NOW at 25x FY27 GAAP P/E — the same multiple they use for Microsoft. Notes the stock trades at mid-40s GAAP next-twelve-months P/E and mid-20s non-GAAP. Their view: markets have already declared NOW a winner in enterprise AI, but the premium prices in perfection. At $72, that's a ~50% haircut from here. Aggressive? Sure. But directionally, they’re questioning the current valuation relative to growth deceleration.
Earnings July 22 is the next catalyst. Options market says 11% swing either way. DA Davidson’s $170 is based on an unchanged 30x FCF multiple — so if the macro multiple compression continues, that target slides further. CLSA’s $72 is a structural valuation call, not a near-term tactical one. Between now and earnings, the tape will tell you which side is winning. But with the stock down 50% from highs and still trading at a premium to Salesforce and Microsoft on growth-adjusted basis, the burden of proof is on the bulls for the next print.
The verdict: Memory cycle is still accelerating and MU is the cleanest way to play it. KeyBanc and UBS both came out swinging this morning, but from different angles — pricing momentum vs. cash return optionality. Two different bull cases, same conclusion: buy the dip (if there is one). MU at ~$990B market cap, sitting just below $1T. That psychological level is noise — the math works either way.
KeyBanc reiterates Overweight, $1,750 PT — that's ~75% upside from here. Their thesis is simple: DRAM and NAND pricing hasn't peaked yet. DRAM up 45-55% QoQ in FQ2, then another 15-20% in FQ3 and 15% in FQ4. NAND even crazier — 75% in FQ2, then 30-40% in FQ3. They don't see supply loosening until at least 2028 because clean room expansions don't start contributing until H2 2027.
UBS takes a different angle — free cash flow. Tim Arcuri sees >$400B in cumulative FCF through CY2028. MU can't buy back stock until Dec 9, 2026 (restriction from the CHIPS Act grants). After that? UBS says at today's price, buybacks could retire 40%+ of the float by end of 2028. That's a massive delevering event.
KeyBanc's estimates are well above consensus: FQ4 revenue $52.1B vs $50.5B consensus, EPS $32.36 vs $31.16. FQ1 (next year) $60.2B rev vs $56.3B, EPS $38.03 vs $34.76.
Bull: Memory is structurally undersupplied for the next 18 months, HBM4/HBM3E pricing resets are an accelerant (HBM4 up ~100% YoY, HBM3E up 150%), and MU's 16 LTAs with hyperscalers represent >$100B in minimum contracted revenue with price floors. This isn't a spot-market boom — it's a contracted super-cycle. The FCF story is the kicker: once buybacks start, the stock compounds faster than earnings.
Bear: Semi cycles don't die of old age, they die of oversupply. Every memory upcycle has ended with capacity coming online just as demand plateaued. KeyBanc says 2028 — but what if it's 2027? If DRAM pricing rolls over, the FCF math breaks, and 40% buyback float reduction is an illusion at lower prices. Also, $1T market cap for a memory company? Multiple compression risk is real.
From KeyBanc (the more detailed of the two):
"The firm does not expect industry conditions to loosen until at least 2028, as clean room expansions will not begin contributing capacity until likely the second half of 2027... HBM4 pricing projected to increase approximately 100% year-over-year and HBM3E up 150% year-over-year."
This is the key tension: supply latency vs. demand persistence. If they're right, the bull case is conservative. If they're wrong by even one quarter, you get a crater.
MU is a rate-of-change play. Pricing momentum is peaking (you can see the sequential deceleration from +45-55% to +15-20%), but absolute levels are still rising. The FCF story gives downside optionality via buybacks — but that's a 2027 story, not today's catalyst. For those with 12-18 month horizon, the r/r is still favorable. For quick-twitch PMs, watch DRAM spot prices and HBM qualification newsflow. MU at 19x trailing P/E with 72% gross margins and 47% ROA? That's not a value trap — but it's pricing in a lot of perfection.
GOOGL SETS UP WELL INTO EARNINGS WEDNESDAY. First mover among the hyperscalers (reports July 22, 4:30pm ET) and the set-up is actually clean. Wolfe keeping Outperform, $460 PT. Monness Crespi Hardt more balanced but constructive. The real signal is in the data points between the two notes.
Ad resilience is the headline. WPP raised its 2026 ad forecast ex-US political to 8.9% YoY from 7.1%. That's a meaningful acceleration. Google Search demand stable. Mid-terms plus World Cup provide a backstop for 2H. Wolfe explicitly says OpenAI's search threat "poses no threat to current estimates." Bold, but they're right for this quarter.
Cloud is the swing factor. BofA (noted in the second article) bumped cloud growth estimate to 70% with a backlog implying $230B in revenue over eight quarters. Demand still outpaces supply. This is the wedge that separates GOOGL from the other mega-cap internet names near-term.
Anthropic is now a real financial asset. Series H at $965B valuation (up from $380B in Series G). Wolfe expects other income could contribute >$50B from this. That's not nothing for a $4.3T market cap — it's a free option on the AI infrastructure theme that's increasingly monetizing.
Bull case: Ad cycle resilient through macro noise, Cloud inflecting hard on AI workloads, and the Anthropic carry creates a $50B+ income tailwind that's not in estimates. PEG of 0.57 says you're not paying for the growth.
Bear case: Valuation is stretched (26.86x P/E on a business that's still mostly ads), capex is unsustainable at current levels, and the AI competition dynamic is real — OpenAI just raised $65B for a reason. Monness Crespi Hardt sums it:
"Competition in the AI era is dynamic, valuation is stretched, outsized capex spending is unsustainable, and the macro environment is treacherous."
That's the honest counterweight. But rates of change matter more than levels here — and the rate of change in Cloud and Other Bets is improving faster than ad deceleration.
Options market pricing 5.1% move. GOOGL has historically exceeded implied moves in prior prints. If Cloud prints 70%+ growth and Search holds, that's a beat-and-raise setup. Risk is if macro cracks ad spend in the forward guidance.
One thing to watch: BofA's revenue estimate of $102.1B vs Street $101.0B. EPS divergence is weird — BofA at $8.38 vs Street $2.90 (must be an op income or tax timing thing). Worth checking the comps Wednesday.
Positioning feels neutral-to-light heading into the print. 88% trailing return creates some "good news priced in" risk, but the Cloud acceleration narrative is still building, not peaking.
BAND IS THE POSTER CHILD for infrastructure software's escape velocity in a horrid tape. Up 365% YTD, the stock has detached from the median software name (down 25%). Two firms just boosted PTs to an $85-86 cluster from a prior $55-70 range (Citizens and B.Riley). The narrative is simple: AI voice demand + platform expansion = accelerating usage. But the stock now trades at a valuation that screams "priced for perfection" with earnings due in 9 days.
Infrastructure software is the only game in town. 8 of 15 infrastructure names in Citizens' coverage are up YTD; the application software bucket is a bloodbath. BAND sits at the intersection of communications APIs and AI agents — their new Bandwidth Build platform lets AI agents autonomously provision voice and messaging services. That's the hook. B.Riley explicitly cited "advances in AI voice technology as a driver for increased demand." The convertible note offering ($275M at 0%) gives them cheap capital to lean into this.
"The firm pointed out that software consists of two subsectors: infrastructure and applications. Infrastructure is in a far better position, with 8 of the 15 infrastructure companies in its coverage having traded up."
Bull case: BAND is the purest play on AI voice infrastructure. The 365% move still has room if they can convert the Bandwidth Build pipeline into billings acceleration. Two analysts just raised numbers, profitability is in sight (analysts expect profitable this year), and the convertible note structure avoids dilution risk at current levels.
Bear case: That 365% is already pricing years of success. InvestingPro flags the stock on its "Most Overvalued" list. Earnings in 9 days are a binary event — miss and the air comes out fast. The 0% convert is a timing signal: management chose debt over equity, but $275M in leverage adds balance sheet risk if growth slows. One bad print and the momentum crowd heads for the exits.
Earnings next week. Watch for Bandwidth Build early adoption metrics and any guidance raise. The convertible note over-allotment option (13 days) means they'll likely upsized into the $316M area. PMs should treat this as a momentum name with razor-thin error bars — either it breaks higher on the print or the infrastructure trade gets a severe repricing.
Verdict: Cloudflare is the right kind of software stock right now. Infrastructure > applications, and NET is squarely in the former. Citizens just lifted PT to $330 from $270, Market Outperform, and the stock is already up 41% YTD at $277, sniffing its $291 52-week high. That’s the setup.
Citizens’ note is a sector-level call disguised as a single-stock upgrade. Their coverage universe median is down 25% YTD (IGV -10%, S&P +9%). But the 15 infrastructure names? Eight of them are up. They like DigitalOcean, Snowflake, Twilio, and NET heading into earnings. The pick-and-shovel theme is alive.
Two other analyst moves on NET this week: TD Cowen raised to $300, Buy, citing ~30% revenue growth and a top AI/cybersecurity pick. Scotiabank upgraded to Sector Outperform from Perform, also $300, with the Workers platform as the catalyst.
“Software consists of two subsectors, infrastructure and applications. Infrastructure is in a far better position.” — Citizens
Bull case: NET is an infrastructure compounder riding AI traffic growth (OpenAI research pilot on network signals for search indexing), a sticky developer ecosystem (Workers), and a cybersecurity tailwind that’s only getting hotter. TD Cowen called it a top AI/cyb pick. 30% revenue growth is the bogey — if they hold that, multiples expand.
Bear case: The entire software tape is still sick — 33 of 45 Citizens names down YTD. NET at 41% YTD is already pricing in a lot. Any guide-down on macro hesitation or Azure/AWS margin squeeze in the CDN space could snap the rally. The $277 level is near a 52wk high — not much r/r left for the next 30 days unless they absolutely crush the quarter.
Verdict: This is a positioning trade, not a catalyst play. MSFT is flat YTD vs SPX +28% — DB calls it “lagging” and that’s generous. Two buys this morning (DB, CLSA) both pointing at a stock that’s priced for failure at ~21x NTM GAAP P/E, 0.79 PEG. For a company compounding EPS mid-teens plus, that’s a setup that usually ends well. Earnings Wednesday is the pin.
Both shops land on the same vector: Azure at ~40% growth, Copilot adoption still early (20M paid vs 450M Office base), and the rest of the business compounding at 10%. DB’s math: strip out OpenAI stake and you’re at 18.5x NTM — a 3-year trough for a name that historically trades closer to 29x. CLSA uses a five-year mean (includes 2022 correction) and still gets to a target P/E of 25x Dec ’27 earnings. That’s not aggressive.
The bear case DB explicitly calls “overblown”: rising memory costs pushing capex higher / FCF lower (true in the quarter but structural returns on AI infra aren’t visible yet), dependency on OpenAI / Anthropic for models (CLSA flags that MSFT “lacks a self-grown AI model”), and concentrated backlog exposure to OpenAI. Fair concerns, but DB says they’re “more than fully reflected” in the multiple.
“The bear case centers on rising component prices pushing capex forecasts higher and margins and free cash flow down; persistent doubt about underlying returns on AI investments; and concentrated backlog exposure to OpenAI. We believe fears around each of these areas have become overblown and are more than fully reflected in shares now trading at 20x next-twelve-months earnings, or an estimated 18.5x subtracting their stake in OpenAI.” — Deutsche Bank
That’s the whole trade in a paragraph. MSFT is priced for a cloudy recession, not a secular shift. Wednesday’s print just needs to not be a disaster — and “Azure +41%” (Wells Fargo) would do the trick.
THE STORY IN ONE LINE: Mainframe pre-announcement blew up in their face — stock down 27% in a week, now trading at a discount to SOTP — but both JPM and Stifel are sticking with their Buy/Overweight ratings. They see value, but visibility into the recovery is garbage.
IBM dropped a Q2 pre-release last week that was horrid — $17.2B revenue vs. $17.86B consensus, EPS of $2.93 vs. $3.02. The culprit? Mainframe and related software took a late-quarter hit as clients shoved capex into servers, storage, and memory ahead of anticipated price increases. JPMorgan’s Brian Essex cut his PT to $250 from $291 (still Overweight), calling the mainframe weakness "disappointing" but noting the stock now trades at a discount to SOTP. Stifel went to $235 from $290 (Buy), flagging the same capex shift toward AI compute.
Both firms acknowledged the stock is near its 52-week low of $204.44 (currently around $212-214). Nine analysts have revised Q2 estimates down since the pre-announcement.
Value is there, even if the narrative is broken. JPMorgan’s SOTP math still shows upside — at $214, you're getting IBM for less than the sum of its parts. Stifel ran base/bull/bear scenarios for 2026-2027 and found a relatively narrow range of outcomes, which limits downside if mainframe and software recover. Break-up speculation is back, though Stifel concedes "no obvious path to adding immediate value" — but the optionality is real. Q4 is seasonally strong, and if the capex shift is just a timing issue, January 2027 could be a very different conversation.
Visibility is a black hole. Stifel explicitly says Q3 is seasonally weak, Q4 is strong, and you won't get any real read-through until late January 2027. That's a long time to sit on a name that just cratered 27%. Multiple analysts cut PTs (JPM to $250, BofA to $280, Argus to $280, Bernstein at $280 Market Perform, Oppenheimer downgraded to Perform). The pre-announcement feels structural — if mainframe is truly losing share to AI compute, that's not a one-quarter problem. Oppenheimer's downgrade to Perform is a direct vote of no confidence in the near-term setup.
"Various factors lead to a relatively narrow range, though the multiple range may be wider with downside risk near-term depending on how software and mainframe growth recover." — Stifel, July 20
This is the key uncertainty. The range is tight on earnings, wide on multiple. If software doesn't snap back, the stock could easily revisit the $204 low — or break it.
You're buying a broken narrative at a discount. The r/r is decent if you have a 6-12 month horizon and believe the capex shift is transitory. But near-term catalysts are nonexistent, and the Q3 seasonal hole will test everyone's patience. Not a position I'd want to size up into earnings season uncertainty — but if you're a value guy with a long leash, this is where you start nibbling. Just don't expect a quick bounce.
Verdict: Two firms going long, but with wildly different PTs ($200 vs $410) tells you everything about the range of outcomes here. The bullish case is a multi-bagger on AIaaS market share; the bear case is competition and execution risk before revenue even inflects. We're buyers on pullbacks but sizing matters.
### THE CALLS
Northland slaps a $410 PT (Outperform) — that's 122% upside from $184. Freedom Capital upgrades to Buy with a more conservative $200 target (up from $159). Gap between them is a feature, not a bug.
Both are reaching the same narrative from different angles: NBIS is playing the AI infrastructure game. Northland uses a DCF that pencils in ~14% share of an $800B AIaaS market by terminal year. That's aggressive. They model a 30% terminal FCF margin — below hyperscaler 40%+ but in line with CRWV. The $410 includes ~$29/share from non-core businesses.
Freedom Capital goes simpler: 4-6x 2028 EV/EBITDA (same as IREN, CRWV). They forecast 41x revenue growth from 2025 to 2028. That's not a typo. 41x. They're reluctant to slap on higher multiples despite that explosive trajectory — "too soon to account for the explosive revenue growth projections."
"The asset-light model will enable Nebius to bring incremental power to market relative to its levered financial capacity." — Northland
### BULL VS BEAR
Bull case (Northland lens): NBIS has a scalable non-dilutive model after proving DDTL 1.0 back in July '23. The new asset-light infrastructure partnership model lets partners finance and operate data centers while NBIS provides the software stack and architecture — near-zero capex for NBIS, unlimited scaling potential. Recent $775M debt facility secured against deployed GPUs and contracted cash flows from an investment-grade customer. The Reflection AI deal ($1B+ compute through 2029) is a multi-year revenue visibility bomb. At 14% share of an $800B market, terminal value is enormous.
Bear case (Freedom's implicit caution): Revenue growth is heroic but underpinned by a single large customer (Reflection AI) and a nascent business model. Meta reportedly entering cloud computing — that's a direct threat if NBIS is piggybacking on the same Nvidia GPU pool. Freedom values at only $200, implying the stock is already pricing in a lot of the upside. Also, 4-6x 2028 EBITDA on what could be peak-cycle earnings is risky if hyperscalers overbuild. And beta of 1.4 — this thing flies both ways.
### POSITIONING TAKE
$184 after a 15.6% pullback in a week. That's a shakeout, not a breakdown — the stock is up 237% in a year. We've seen this movie in TMTB: parabolic names that find a bid on the first 20% dip if the narrative is intact. It is. But the PT dispersion means PMs need to decide which time horizon they're playing. Short-term, the $200 Freedom target is already above current price — upside limited. Long-term, the Northland $410 is a pipe dream until we see the next catalyst (likely next quarter's revenue print or another Reflection AI-type deal). We'd wait for either a further pullback to ~$150 (where the 50-day might be) or a catalyst that validates the 41x revenue trajectory. Not chasing here, but watching.
KeyBanc's tower preview sets the table for a moot quarter. Firm sees US leasing roughly flat QoQ for Q2 and the rest of the year — no catalyst, but no disaster either. The bar is low: in-line results and maintained guidance are enough. KeyBanc thinks 2026 marks the bottom for the sector, but that's cautious optimism, not a re-rate trigger.
"We anticipate guidance will be reiterated and forward commentary on bookings will reflect cautious optimism that 2026 represents the bottom for the tower sector."
Cross-check the comps: towers at 18x EV/EBITDA and 15x P/AFFO vs. data centers at 18-21x on the same metrics. CCI isn't getting love relative to DLR/EQIX — but that’s the setup. No news is good news here. If CCI prints in-line and holds guidance, the stock should hold. If they guide up? That's the real signal.
BofA is leaning hard into the AI narrative ahead of META’s July 29 print. They’re sticking with Buy and $835 PT, but the real signal is in the estimate revisions — Q2 rev of $60.6B (Street $60.2B) and EPS $7.50 (Street $7.18) already baked in. The firm is now explicitly adding AI capacity revenue to the model: 2027 rev bumped 1% to $316B, but the big move is 2028 (up $11B). The capex guide is getting a haircut too — they expect an update to $135-150B from $125-145B. That’s a lot of spend, but BofA sees the payoff coming faster than the market prices.
The headcount cuts (10% workforce reduction in May, job postings down 49% QoQ) are lowering the SBC drag. Gross margins sit at 82% — that’s pricing power in digital ads, not just AI hype.
"We are adding $5B in revenues for AI capacity benefits following reports of a potential compute deal with Anthropic."
The key tension: capex is going up, but so is the revenue story. BofA is basically saying the Anthropic deal (if real) turns infrastructure spend into a direct revenue line. That’s a narrative shift — from cost center to profit center. PMs should watch the July 29 guide for explicit AI revenue disclosure. If it’s there, the $835 PT starts looking conservative.
KeyBanc is sticking with Overweight and a $725 PT, and they're the most bullish thing I've seen on AMD in weeks. Thesis: agentic AI is driving outsized demand, and AMD is a direct beneficiary on the server CPU side — hyperscaler refreshes plus inferencing workloads that need higher CPU intensity. They're explicitly calling for better June quarter results and a raise in guidance.
The supply chain math is getting better for AMD, too. CoWoS allocation is moving from 80k in 2025 to 130k by 2027, and the MI455 volume ramp is on track for H2. KeyBanc's Q2 revenue estimate sits at $11.6B vs consensus $11.3B — that's the kind of beat-and-raise setup PMs want to hear. (MTK order cuts on N4 smartphone frees up capacity for AMD; 15-20% unit growth in 2026 looks achievable.)
"KeyBanc said agentic AI is driving outsized demand with AMD well positioned as a beneficiary."
Not sure the $808B market cap leaves much room for error, but the rate of change on supply and workload mix is moving in the right direction. Worth a look against NVDA if you're playing the AI infrastructure rotation.
Verdict: Bullish signal remains intact. Jefferies bumps PT to $320 (from $290) on AGI CPU demand acceleration — but the real story is the street's collective vision of a $475-500 stock if the CPU-inference thesis plays out. Not reacting violently today (flat-ish pre-market), but the set-up is getting cleaner.
Key action: Jefferies now sees Arm's AGI CPU revenues hitting $18B by FY31, up from Arm's own guidance of $15B. That's a 20% upside to the company's number. The catalyst? Agentic AI workloads moving to CPUs — Oracle and ByteDance are now customers. And Softbank's potential GPU launch next year (using Arm's design services) adds a royalty kicker.
The broader analyst consensus is way more aggressive: six firms have jabbed PTs to $475-500 in recent weeks, anchored by the same thesis — CPU will capture a meaningful share of inference spend, and Arm's architecture is the default for power-constrained data center fleets.
"Agentic AI driving increased demand for CPUs" — Jefferies, not exactly a hot take but the revenue conviction is new.
The bullish case is a structural narrative shift (GPU → CPU for inference) plus a product expansion (Softbank GPU). The bear case: Arm trades at ~58x forward EBITDA even after the pullback from highs. The TAM expansion is real but the market is already pricing in perfection. For PMs sizing it: this is a high-conviction growth name but the risk/reward tightens above $300. Keep it as a beta-to-AI theme trade, not a fundamental long unless you get a better entry.
Jefferies sees Databricks' $188B mega-round as a direct comp read-through. They bump SNOW PT to $310 from $300, steady Buy. The logic: Databricks implies 17x FY28 revenue at a 65% CAGR assumption; Snowflake trades at 13x with 31% reported growth. Even a discount 15x multiple gets you to the new PT and a ~$115B EV. Stock is at $274, kissing its 52-week high — not much room to the upside from here, but the relative valuation gap gives the narrative legs.
"Assuming a 65% revenue compound annual growth rate from fiscal year 2026 to fiscal year 2028, the Databricks valuation implies approximately 17 times fiscal year 2028 revenue. Snowflake currently trades at approximately 13 times fiscal year 2028 revenue on a $100 billion enterprise value."
Other noise worth flagging: UBS already at $370 (Buy) and the board just tied a $448M CEO comp package to hitting $184B market cap. Databricks' data warehousing run rate hit $1.5B (competing directly with SNOW's core). The bull case rests on AI workload displacement and the data cloud secular story — Databricks is the wind, not the storm.
CLSA throws cold water on the AI narrative. Initiated Hold with a $165 PT (15x GAAP FY28 P/E) — that's a big discount to MSFT's 25x and implies zero upside from here. The firm isn't questioning the moat (150k customers, 77% gross margins) but says organic growth is still slowing to ~6% YoY once you strip out acquisition tailwinds and Agentforce revenue. They're willing to wait for proof.
"Beneath the tailwinds from acquisitions and AI revenue, organic growth is still slowing and is likely at just 6% year-over-year."
The Street is split on the timing. Evercore lowered PT to $250 but stays Outperform (headwinds in Tableau/Commerce Cloud). Goldman reiterates Buy at $242, banking on organic acceleration in H2. KeyBanc downgraded to Sector Weight — they're skeptical on Agentforce driving growth anytime soon. Guggenheim upgraded to Buy at $228, calling current valuation an attractive entry. Net net: bull case rests on AI monetization, bear case is slowing core and a PEG of 0.5 screaming value trap. CLSA sits in the middle — patient, not convicted.
Verdict: Infrastructure tailwind is real. TWLO’s +45% YTD is being rewarded with a fresh wave of PT hikes — Goldman at $300 is the outlier, but the consensus is tightening around $240-260. The stock is now in the “good company, bad sector” bucket — software applications getting annihilated (median down 25%), but infrastructure names like TWLO are the exception. Citizens bumps to $250 (from $210), Mizuho to $240, Stifel upgrade to Buy/$260, and Goldman initiation at Buy/$300. The bullish throughline: AI developer activity driving Twilio’s Conversations and data tools. 50% of Forbes 50 AI startups were Twilio customers as of Sept 2024.
“Infrastructure is in a far better position, with 8 of the 15 infrastructure companies in Citizens’ coverage having traded up.”
Not all roses. The infrastructure vs. applications gap is widening. TWLO is riding that wave, but the broader software tape is horrid — IGV -10% vs. S&P +9%. If rate of change on macro spending slows, this stock is not immune. Positioning is likely getting crowded after a 45% YTD rip. But for now, the narrative is clean: public cloud + AI developer flows = revenue acceleration.
Key level to watch: $240-250 area is where the new PTs cluster. Stock at $207 has room if Q2 print (next week?) shows guidance beat. Watch for any deceleration in messaging revenue.
Still stuck in the mud. BofA cuts PT to $105 (from $110) on lower Data Center and Cloud revenue forecasts — now sees FY27 revenue of $7.17B vs. Street $7.37B, just 11% y/y growth. Stock already down 52% in the past year, trading at 11x EV/CY27 FCF. That multiple is cheap for a reason: the Data Center transition is a multi-quarter headwind and the Cloud ramp hasn’t proven it can fill the gap fast enough.
The broader analyst consensus is a mixed bag, not a clear signal. BofA and Citizens stay on the sidelines (Neutral / Market Perform). Bernstein, Truist, and Piper all maintain Overweight/Buy — but they’re leaning on future catalysts (AI monetization, Team 26 event, improved H2 visibility) rather than current data. No one is pounding the table right now.
“A guide above consensus will likely serve as a catalyst.” — BofA
The near-term r/r hinges entirely on that FY27 guide. Above = squeeze candidate. At or below = more pain ahead given weak software sentiment. For now, TEAM is a show-me story with a low bar — but the trajectory is still unclear.
AMZN = the cloud reacceleration trade is alive and well. Truist reiterates Buy with $320 PT, while KeyBanc pushed to $335 and Citizens sees Q2 rev hitting the high end of guidance. The single narrative: AWS growth inflecting from 28% in Q1 to 30% in Q2, and sustaining 30%+ for the rest of 2026. The catalysts are real — AI workload ramp from OpenAI/Anthropic partnerships plus a slug of data center capacity coming online in H2.
Truist’s proprietary card data shows NA marketplace sales tracking +13% YoY, beating the e-commerce industry print. Resilient consumer even with higher fuel costs and macro noise. They raised Q2 estimates accordingly.
“Truist projects sustained momentum of 30% or higher throughout fiscal year 2026 driven by faster adoption of AI workloads from partnerships including OpenAI and Anthropic, and significant data center capacity coming online in the second half of 2026.”
The beat-and-raise playbook is set for July 30 print. PMs are long this name for a reason — narrative, rate of change, and positioning all align. Nothing here to disrupt the bull case.
TXN is the consensus long — and that’s exactly why Cantor stays Neutral at $340. Bulls have their narrative locked: data center power exposure, capacity leverage, FCF inflection, H2 pricing uptick. A beat-and-raise is the base case. The problem? Everyone already owns it. The question Cantor is asking isn’t whether TXN prints a good quarter — it’s whether “seasonal-plus” is enough to drive real alpha from here.
“Primary pushback to the bullish thesis centers on the sustainability of 90% year-over-year data center growth seen in the first quarter. Questions also remain about the magnitude and breadth of pricing increases in the second half.”
Valuation doesn’t help the r/r. 48.4x P/E, 4.48x PEG — that’s a lot of blue sky already priced in for a semi cycle that’s getting long in the tooth. The Morgan Stanley and Stifel upgrades give cover, but the positioning is crowded and the setup is binary. Earnings in 2 days. Either the DC growth story extends or the multiple compresses fast.
Verdict: Data center hyperscaler win is the story, but the AI trade cooling is real. TD Cowen stays Buy at $285, raising 2027 estimates after a June award that adds ~$600M to the committed backlog. The stock’s down 5% on the week but up 33% in six months — the market is pricing in a second customer, and Cowen sees that as high probability within weeks.
The thesis is simple: Generac is transitioning from a residential generator play to an AI infrastructure beneficiary. That shift is real — the hyperscaler award is tangible, and the 26kW HSB platform has less tariff exposure. But the P/E at 67x is pricing in a lot of perfection. The real catalyst isn’t the award itself, it’s the second one. Cowen thinks it’s imminent.
"TD Cowen sees a high probability that a second customer will be added in the coming weeks, which could further impact estimates."
Other noise in the tape: Barclays and UBS both bumped PTs (to $285 and $335 respectively), facility acquisition in Belvidere for large-megawatt packaging, and a new international EVP. All incremental positives, but the core debate remains: how fast does the hyperscaler book expand, and does the residential business stabilize?
Needham stays in the camp. Reiterates Buy and $55 PT on RNG (current $41.32, +44% YTD, +60% over six months). They see a modest Q2 beat despite enterprise pricing headwinds that persist through early 2027. SMB pricing is steady, AI add-ons are driving incremental ARPU. The real story: growth re-acceleration in 2027 as enterprise headwinds fully lap and AI product adoption ramps.
"Pricing remains the key consideration for enterprise contract renewals, but pricing in the small and medium business segment remains steady and favorable when considering the increasing adoption of artificial intelligence add-ons."
Valuation is still depressed — 13x something (Needham doesn’t specify the metric, but FCF yield is 17%). That’s cheap for a biz that should grow again in ’27. Multiple expansion expected through the year. Not a sexy name, but the r/r works if you believe the narrative holds. Moody’s just affirmed Ba2, downgraded unsecured notes to B1 — not a credit panic, they noted progress on profitability and deleveraging.
AI agent expansion on RingCX (announced at CCW) is a nice narrative hook but not a near-term catalyst. The stock dipped after that news — probably just noise from other macro rotation. We’d hold. No need to overthink a steady grind with a clean FCF backstop.
SNAP remains a show-me story and the setup is getting worse, not better. Cantor cut PT to $5 from $6 (Neutral) — stock at $4.53, so new target isn't exactly a call to arms. Q2 is fine: revenue +15% YoY to $1.55B, near the high end of guidance, EBITDA margins expanding 960bps to 12.7%. But it's the forward look that matters, and Q3 models deceleration to +11% on a 1pp tougher comp. Rate of change is negative.
Cost saves are flowing (margins expand another 730bps in Q3 to 19.4%), but that's table-stakes execution. The top-line narrative is what matters, and the read-through is that advertising recovery isn't sticking for SNAP like it is for broader digital (Evercore noted robust Amazon/Google spend). Engagement in North America is a concern (DA Davidson Neutral, $5; Benchmark/Truist both Hold). The AR Specs launch at $2,195 is a nice PR moment, not a revenue driver.
CLSA comes out of the gates with an Outperform and $300 PT — 15x FY27 GAAP earnings, benchmarked to CRM’s multiple. That’s a 26% upside from here. The firm argues the AI disruption narrative is overdone, and that Adobe’s slowdown predates the AI wave (just a post-Covid pullback). At 13.6x P/E, it’s sitting at a steep discount to software peers — CLSA calls it one of the platform’s most undervalued names. HSBC also upgraded to Buy recently, highlighting resilience vs. AI competitors. Piper Sandler stays Neutral at $240, so not everyone’s onboard — but the rate of change in sentiment is clearly tilting bullish.
“The narrative of AI disruption affecting Adobe is overdone, and metrics support this view.” — CLSA analyst Bhavtosh Vajpayee
Quick backdrop: Q2 revenue grew 12.7% YoY, FY26 guidance at +11.8%. The Topaz Labs acquisition (AI video/image enhancement) plugs into Firefly and Creative Cloud. For PMs: the r/r here is about whether Adobe can hold the line on margins (89.4% gross margin is elite) while AI-native tools eat at the edges. Bulls say the moat widens with data flywheels; bears say vector is eroding. Right now the tape is buying the CLSA/HSBC thesis.
BofA sticks to Buy ($140 PT) despite the 52% drubbing. Stock at $73, down 12% last week on Meta/SpaceX competitive noise. BofA isn't buying the structural competition argument – demand still crushes supply.
Firm raised FY26 capex estimate to $34B (from $29B) reflecting hardware buildout dynamics. Profitability focus is intensifying: Q2 operating margin expected at 2.4% (Street 2.8%), but they see a ramp to 14.6% by Q4. Loss of –$3.81/share for FY26 is the bogey.
"The firm does not view the compute market as structurally competitive, stating that demand for AI compute far exceeds supply and access to capacity is the primary gating factor rather than provider selection."
That's the core thesis. Meta offering compute capacity? Real, but not a fatal threat. The incremental active power story drives revenue recognition here – operating leverage is the 2H catalyst. Not a name for the faint of heart at these levels, but the r/r flips if the margin ramp materializes.
CLSA initiates with a hard Underperform and $92 target — basically telling the Street the emperor has no clothes on GAAP earnings. Analyst Bhavtosh Vajpayee flags slowing growth, rising M&A appetite, and unclear product moats. The founder's return is acknowledged as a narrative shift, but they're not buying the pivot yet. The GAAP vs. non-GAAP gap is the widest in CLSA's coverage universe, and they prefer GAAP. That $92 target is 15x FY28 GAAP P/E — a massive discount to the current 45x trailing non-GAAP multiple.
"Workday is optically inexpensive on non-GAAP earnings but fails most tests of safety for the investor."
Counterweight: Monness upgraded to Buy at $150, Cantor at $160, Oppenheimer at $165 — all post-product meetings and despite WDAY being the worst performer in Monness' coverage this year. That chasm between the bears and bulls is the setup. CLSA is early or wrong? The PEG of 0.54 suggests the growth-adjusted valuation isn't crazy, but revenue productivity per employee is middling and GAAP losses still sting. Legal overhang from AI-screening discrimination claims adds no visibility.
Cantor bumped PT to $200 (from $180) but stays Neutral – that's the cautious end of the spectrum. The ad story is real: Q2 revenue mid-40s%, Q3 guide ~40% y/y. The DPA product is the unlock for incremental budgets through 2027. Multiple firms (Jefferies, B.Riley, Citizens) all pointing to strong ad momentum, AI tools, and market share gains. Earnings in 10 days. The real watch item is data licensing renewals with AI labs expiring 1H27 – that's the next narrative catalyst.
"The new DPA product should continue to unlock incremental advertiser budgets through the second half of 2026 and fiscal year 2027."
NEEDHAM BULLISH ON NVDA OPTION EXERCISE. Raised PT to $145 from $128, Buy maintained. The trigger is straightforward: Nvidia exercised its option on the remaining capacity at Hut’s Beacon Point facility, locking in Phase 2 (352 IT MW) starting Q2 2028. Total signed capacity there is now 704 IT MW. That crystallizes $1.86M per MW in NOI on a 15-year triple-net lease — terms that match CIFR/WULF and beat APLD/CLSK on a per-MW basis.
Needham’s math: 2028 rev estimate goes to $1.70B (from $1.44B), EBITDA to $1.32B (from $1.07B). The new PT is 19x discounted 2028 EV/EBITDA. For context, stock trades at $101 — up 323% in the last year — with a beta of 6.07. That’s a volatile name, but the thesis is about conversion of MW into contracted cash flows. Benchmark also raised to $165, and Lucid initiated at $226 (yes, 2.2x upside from here), citing $9.8B lease value across the campus.
The bear case: insider sales (CLEO sold $1.25M) spooked the tape briefly. But the core story is that Nvidia is effectively backstopping the buildout. If you believe AI infrastructure demand stays structural, HUT offers a triple-net lease model with a 1,000 MW total utility capacity and only 704 MW signed — upside to the remaining 296 MW.
"The economics for both Phase 1 and Phase 2 at the Beacon Point site remain unchanged, representing $1.86 million per megawatt in net operating income over an average 15-year triple-net lease." — Needham
Verdict: The Beacon Point lease is the key catalyst. It derisks the build and gives multi-year revenue visibility. The stock is pricing in some of that, but the remainder capacity and potential for similar deals elsewhere (HUT has other sites) keeps the r/r interesting. Watch the Q2 2028 ramp timeline and any insider selling patterns.
Verdict: CLSA drops a Hold initiation with a $145 PT, flagging ORCL’s debt-fueled AI capex spiral as a high-beta trap. Piper Sandler still calls Overweight on OCI ramp — but the bear case has teeth.
CLSA’s Bhavtosh Vajpayee estimates Oracle may need up to $500 BILLION in capital to hit management’s 2030 targets, with internal cash covering only 20%. The balance sheet already carries $167.4 BILLION in total debt and a debt-to-equity of 4.46x. That makes the stock a levered bet on AI cloud demand materializing faster than the interest bill.
“Financial estimates are highly dependent on minor variations of input variables” — a polite way of saying the model has no margin for error.
The $145 PT implies 18x FY28 GAAP EPS, a 20% premium to enterprise SaaS peers. Not cheap given the execution risk.
On the other side, Piper Sandler reiterates Overweight, pointing to ~$2.2 BILLION in potential OCI revenue add by FY27 from capacity expansion. That’s the bull case — if ORCL can deliver on the cloud narrative without tripping over its own debt stack.
Bottom line: 63% off the high ($126 now) looks tempting, but the capital structure math is ugly. This is a high-conviction binary, not a value play.
Deal done, downgrade follows. Deutsche Bank cuts GHH to Hold from Buy after the Arlington Capital Partners cash offer at 1234.9p (1230p cash + 4.9p interim dividend). The new PT of 1230p effectively pins the stock to the bid – upside is the dividend stub. No point arguing with a 45%+ premium to where the stock traded pre-announcement, but the fundamental story (strong defense/semi photonics, industrial orders +49% YoY in H1, A&D cash conversion CAGR of 21%) is now a merger arb footnote.
"The portfolio value has improved materially under the company’s reset strategy of recent years. The outlook for photonics markets remains robust, particularly in defense and semiconductors."
Completion expected Q4 2026. If you were long the turnaround, take the money. If you're late, the r/r is boxed in.
Cantor sticks to $28 target on a de-risked guide. The stock’s at $17.86 – that’s a 57% implied upside to Cantor’s target and north of both Stifel ($24, Buy) and Goldman ($26, Buy). The key twist: management is deliberately sandbagging less. Q1 beat was only 2.7% vs the historical 4-4.5%, and Q2 guide of $359-363M (23-24% growth) looks lowballed again. Cantor pencils in a similar 2.5% beat → ~$371M actual, or 26.5% growth. The question is whether that $9M overage gets folded into the full-year number (implying ~23.7% growth) or saved for another beat. Either way, the rate of change in guidance philosophy is the story – lower beats are the new normal, but the underlying growth trajectory hasn’t collapsed.
All three brokers see upside, but the consensus PT cluster ($24-$28) implies the stock is pricing in a lot of fear. With the new CFO (Erica Smith, September) and the Klaviyo Social Marketing launch adding a CRM-social integration angle, the narrative is shifting from “will growth slow?” to “how sticky is the platform as SMBs tighten?”. Not sure we can read too much into the product launch yet, but it’s a logical extension.
“Cantor Fitzgerald estimates that assuming a 2.5% beat for the second quarter, similar to the first-quarter beat, actual second-quarter revenue would reach $371 million, or 26.5% growth. The firm suspects management will flow through the approximately $9 million second-quarter beat to the 2026 guidance, implying 23.7% expected growth versus the current 23%.”
Net: cheap for double-digit growers (26% rev growth on a beat-adjusted basis), but the compression in beat rate is a real headwind to sentiment. PMs should watch the Q2 print – if they miss that 2.5% beat threshold, the multiple re-rating stalls.
Bernstein is comfortable staying long. Reiterated Outperform and $150 PT (about 20% upside). The call is not about a growth acceleration—it's about pricing power becoming visible through the P&L. Their proprietary basket tracker shows realized prices are rising even as headline sticker prices hold flat. The mechanism: platform voucher spend is coming down, so consumers pay closer to retail. That's a direct margin tailwind.
"Shopee has kept headline prices largely unchanged while pulling back on coupon subsidies, allowing realized selling prices to climb."
This is the right framing. Forget GMV growth debates for a second—the margin story is improving without needing aggressive price hikes. TikTok Shop is also easing promo intensity, though their delivery cost absorption is muting the net benefit. Shopee's relative pricing gap vs TikTok is narrow and stable. No share shift alarm bells.
The Q1 print muddies the picture ($7.1B revenue beat, $0.67 EPS miss on operating spend). PMs will focus on the revenue trajectory, not a one-quarter EPS hiccup. 0.51 PEG ratio keeps the valuation anchored if execution holds. Risk: if Shopee has to match TikTok's delivery subsidies again, the pricing tailwind reverses. Not priced for that scenario.
RBC pounding the table on SHOP — $170 PT, 36% upside from $124.89. They’re leaning on third-party data and think Q2 results AND Q3 guidance both beat consensus. Revenue +32% LTM, but that P/E of 122x — you’re paying for growth, no question.
The broader analyst community has been turning more positive: Jefferies and Stifel both upgraded to Buy (PTs $160 and $150), Citizens reiterated Market Outperform. UBS remains Neutral at $130. The narrative is shifting toward agentic commerce — Manus integration, Spring 2026 Editions with Catalog and Universal Commerce Protocol. It’s a platform story, not just GMV growth.
“RBC Capital stated that it believes Shopify’s second-quarter results and third-quarter guidance may exceed consensus expectations based on the available data points.”
Biggest risk: software sector sentiment. This name trades on narrative and multiple expansion — interim data points and macro wobbles will whipsaw the stock. But the position here is fundamentally constructive. Long-term compounding story if you can stomach the multiple.
BUY. Stifel staying with $125 on the Lumibird Medical deal. €300M purchase (15.9x FY25 EBITDA, plus a €50M earnout) triples the medical segment to ~$204M pro forma. Takes Advanced Solutions from 16% to 26% of sales. Smart use of that $800M cash pile doing nothing.
"The deal aligns with management's strategy of pursuing complementary bolt-on acquisitions in the $50M to $200M revenue range, with medical as a priority vertical."
Needham upgraded to Buy at $110 last week on the revenue turn. Q1 earnings beat by $0.02 (big whoop) but the top-line beat is more notable — $265.5M vs $256.4M consensus. Stock at $103, so Stifel's target implies 21% upside.
CFO Mammen selling ~$1.2M at $119-120 — not reading into that. Diversification at multi-year highs. The 39% 1-year rally has the tape a little stretched (PMs hate buying after that kind of move) but the medical anchor changes the r/r equation. New TAM expansion of ~$1B is not priced in.