Good morning.
Futures green across the board — SPX +0.6%, NDX +0.8%. Memory names leading the charge: SKHY ADR closed +13% on debut (opened at $170, printed $181). Nanya is the real story — REVENUE +684% Y/Y, GM 79.5%, ASP UP >500%. That's not a cycle, that's a super-cycle. DRAM contract prices now expected +13-18% QoQ in 3Q26, NAND +10-15%.
Asia backdrop: Korean and Taiwan memory names ripping. KOSPI +2.4% on SK Hynix and Samsung. Dollar a bit softer, yields steady. No macro drama, just pure AI demand printing.
Three themes framing the day:
1. MEMORY SUPER-CYCLE IS NOT A THEORY ANYMORE. Nanya's print confirms it — legacy DRAM tightness is structural. SKHY Chairman saying even doubled capacity in five years won't meet demand. This is a multi-year ASP expansion, not a cyclical blip. (MU, WDC, SKHY all direct plays. LRCX, AMAT, KLAC benefit from the capex push.)
2. META'S SILICON GAMBIT GETS REAL. Iris processor entering production in September. Doubling deploy compute from 7 GW to 14 GW by 2027. The fundamental story is credible — but the equity raise risk (30-50B?) is now a live debate. Markets will demand clarity on financing. (Positions NOW, META long, but watch the cap structure.)
3. OPTICS INFLECTION — COHERENT-LITE GOES VOLUME. Innolight shipping 2.4T coherent-lite transceivers for Google's TPUv9 scale-up network. Marvell doing the custom DSP. This is the transition from spec to volume that the bull case for optical networking has been waiting on. (LITE, MRVL, CRDO beneficiaries. Also positive for NVDA's networking ecosystem.)
4. NVIDIA' s MARGIN OPTIMIZATION TRADE. They're rationally starving gaming (GeForce) to feed Rubin datacenter wafers — DC margins ~3x gaming. This isn't a demand problem, it's a math problem. Net-net bullish NVDA DC revenue, but gaming supply constraints are real. (INTC foundry lobbying for NVDA or Apple as customers — long shot but worth tracking.)
We'll hit up SKHY and Nanya first, then get to META and the optical names.
Key incremental data point: Wolfe just lifted its 2027 capex estimate to $220B from a prior $15B capacity figure (yes, that’s a massive revision). Street was at $160B+. That’s a full standard deviation above consensus. Either they see something the rest don’t, or they’re front-running an inevitable capital raise.
Bull (BofA, Wolfe, Truist): Meta is building a custom silicon moat at a cost advantage that hyperscalers will envy. The Iris chip entering production cleanly in September changes the unit economics narrative. “Building capacity at below $30B/GW offers favorable economics compared to Amazon and Google’s $10-16B annual cloud revenues per GW” (BofA). With 82% gross margins and 26% revenue growth, Meta can absorb $145B in ’26 capex without diluting EPS — Citizens’ 24x ’27 GAAP EPS still leaves room. The Muse integration into Advantage+ is a direct ad revenue catalyst.
Bear (Benchmark, Citizens): The spending is staggering and the payoff is hypothetical. $220B in ’27 capex implies a TAM that requires AI revenue today — not 2028. Zuckerberg himself acknowledged AI agents are progressing slower than expected. “Meta will need to demonstrate incremental, durable non-advertising revenue sources and frame return on invested capital” (Wolfe — even the bulls are putting riders on the thesis). Citizens cut its PT by one turn on multiple because capex worries offset the growth narrative. Benchmark sits on a Hold, essentially saying “great tech, wrong risk/reward at 22x forward earnings with 40%+ capex growth.”
“META will need to demonstrate incremental, durable non-advertising revenue sources and frame return on invested capital for investors to gain comfort with elevated spending levels.” > — Wolfe Research
“Building capacity at below $30 billion per gigawatt could offer favorable economics compared to our estimates for Amazon and Google annual cloud revenues per gigawatt at $10 billion to $16 billion.” > — BofA Securities
“The memo disclosed multi-year supply agreements with Samsung for memory, Sandisk for flash, and Sumitomo for fiber optics. These deals were struck during a memory shortage severe enough to lift consumer hardware prices.” > — Benchmark (summarizing the internal memo)
Bottom line: The bull case is about cost efficiency and an accelerated timeline. The bear case is about ROI uncertainty. Both are right — but the rate of change favors the bulls if the Iris chip delivers and non-ad revenue materializes. Watch the capital raise narrative. Wolfe hinted at it. If Meta issues debt or equity at these levels, the PT math gets tricky. For now, the momentum is undeniably positive.
BUY-SIDE VERDICT: The street is going parabolic on WFE – and AMAT is the clearest structural beneficiary. Three firms raised PTs to $650-700 from a prior $530-540 cluster, all citing multi-year capacity buildouts in memory, leading-edge, and advanced packaging. The bull case is baked into $250B+ WFE by 2028. The bear case? Michael Burry just disclosed a short.
Memory fab ROI is >3x better this cycle. Demand locked via take-or-pay. That changes the risk profile – this isn’t 2022 inventory glut. AMAT gets the biggest slice of the fastest-growing pie: leading-edge front-end and DRAM. Mizuho specifically flags TSMC CoWoS hitting 140k wpm by YE26, >200k by YE27.
Blockquote of the day (TD Cowen):
“Under a $250 billion WFE scenario, Applied Materials’ earnings could approach $25, up 1.9 times from the firm’s calendar year 2026 EPS estimate of $13.35.”
That’s not pricing in the $400B NAND recovery scenario, which flips favor to Lam. But for now, AMAT is the 1x1 beneficiary.
Bull: Structural demand from AI memory, advanced packaging, and foundry is pulling WFE into a multi-year supercycle. AMAT’s installed base and product breadth let it capture disproportionate upside. P/E looks high (55x) only if you ignore the EPS ramp – forward P/E compresses fast.
Bear: Michael Burry is short – disclosed positions in AMAT, NVDA, TSLA, and SOXX. His thesis: overvaluation from AI hype. Also, if NAND stays weak, the $400B WFE scenario stays theoretical. Options flow shows heavy call activity but also 30k+ puts – someone’s hedging hard.
This is a high-conviction long for PMs with a 12-18 month horizon. The WFE narrative isn’t priced for $250B yet – consensus still lags. Key risk: if memory demand cracks or CoWoS buildout slows, AMAT is the most exposed. Pair with a short on a weaker semi equip name if you want to neutralize beta. Burry’s short is a headline risk, not a fundamental thesis killer.
KLAC is getting love from two shops this morning—both raising PTs to $260-270 from sub-$200, and the thesis is consistent. TD Cowen and Stifel both see the WFE supercycle extending through 2028+ with process control capture increasing, but they're not just extrapolating the current P/E. Stifel models 6.4% sequential process control revenue growth; TD Cowen pencils a $250B WFE scenario that delivers $9-10 in annual EPS (from $3.53 trailing). Stock is +80% YTD, already pricey at 63x -- but the rate of change in memory ROI is the driver.
TD Cowen: $260 target (from $200). WFE could hit $250B in 2028 and $400B in 2030, driven by memory -- where fabrication ROI is >3x prior cycles. Under the $250B case, KLA earns $9-10/sh. Under $400B, mid-teens. Stifel: $270 target (from $191). Sees semi process control system rev at $2.5B (+6.4% q/q). Notes the sales mix (currently 82% foundry/logic / 18% memory) normalizes to ~65-70% FD/LOG by late this year early next -- but both sides remain strong.
"The firm noted that demand is committed through 'take or pay' agreements in the memory sector."
(Both analysts echo the same message: visibility into 2027 is as good as it's ever been for WFE. The question is how much of that is in the stock at $230.)
THE VERDICT: WALL STREET DOUBLES DOWN ON THE AI INFRASTRUCTURE THESIS. Two top shops (MS, TD Cowen) re-upped Overweight/Buy calls with PTs in the $275-288 range after sitting down with Jensen and Co. The message is consistent: compute is still in shortage, demand is broadening beyond hyperscalers, and the market is framing the opportunity too narrowly. This isn't a "one-trick AI chip" story anymore – it's an infrastructure platform play.
Both MS and TD Cowen came away from recent NDR/lunch events with the same conviction. Morgan Stanley reiterated Overweight, PT $288, calling NVDA a top pick. They see an "accelerating and diversifying growth story" that should appeal to both value and growth investors (PEG of 0.28 is a data point they flagged). TD Cowen (Josh Buchalter) kept Buy, PT $275, also a Top Pick. He emphasized Nvidia's "co-design across hardware offerings and vertical-specific software" as a durable competitive moat.
The collective thesis: demand is not slowing. Compute remains in shortage. TD Cowen cited rising legacy GPU rental prices, widening AI adoption, and recent cloud deals at premium prices. MS added that space, power, and geopolitics are now central to decision-making – sovereign and neocloud customers could outpace hyperscalers over time. The Vera CPU ramp ($20B in sales by year-end) is a concrete expansion into Intel/AMD turf.
Bull: NVDA is supply-constrained, not demand-limited. The TAM is expanding into enterprise, sovereign, and neocloud. The Vera ramp and nuclear-powered datacenter deal (Valar) show execution breadth. At a PEG of 0.28, value investors are starting to bite.
Bear: Michael Burry is short. Zuckerberg admitted AI agent development is slower than expected. The AI bubble narrative lingers. But BofA dismissed the recent sector pullback as "temporary adjustment, not fundamental shift" – and still sees NVDA leading near-term.
TD Cowen, after lunch with Jensen Huang, CFO Colette Kress, and IR Toshiya Hari:
"Nvidia’s technology stack and opportunities are being framed too narrowly by the market. The company’s hardware and software are positioned to serve as infrastructure for AI across industries and regions."
Bottom line: Two top-tier shops just doubled down. The rate of change is still accelerating (revenue +71% LTM, $20B Vera guide). The bear case is noise until we see actual order cancellations.
Verdict: Engagement slippage + churn risk = hard to see near-term catalysts. The structural moat (scale, distribution, recommendation engine) is being tested. Stock at $75.47, 6% off 52-week low ($70.86). Earnings July 16.
"The firm said rising churn threatens Netflix’s largest moat, which is its scale and distribution... if engagement slows and churn begins to rise, the core of this structural advantage begins to erode." (Citizens)
Bottom line: Not screaming buy yet. The narrative shift from “streaming king” to “scaling with churn” is real. If subs beat, you get a relief rally. If engagement data disappoints, $70.86 gets tested. I’m watching the July 16 print for a rate-of-change signal – not for a value trap.
VERDICT: BMO STARTS MARKET PERFORM ON AI CONCERNS – BUT THIS IS A “SHOW ME” CALL, NOT A BREAK-UP STORY. $393 PT implies ~10% upside from $355. Simply put: patience is the ask. The bull case ($500 from Raymond James) exists, but BMO’s hesitation is about rate of change in the model, not current fundamentals.
Daniel Jester put the call out this morning – Market Perform, $393 target. This is not a bearish thesis on the business quality. ROP is a 29-business compounder of niche vertical software + tech products, 75% software mix, mid-single-digit organic growth, ~$9B in acquisitions since FY23 with a “forever ownership” mentality. Gross margins 69.4%, 12 years of dividend increases. Solid.
The issue: AI disruption risk to the long-term durability of the financial model. Specifically, the acquisition strategy – already shifting toward earlier-stage and bolt-on assets – now faces more complexity and risk given tech shifts in application software. BMO wants to see evidence that management’s internal AI initiatives work and that future M&A direction is clear before getting more constructive.
“Patience is warranted as management executes on its internal AI initiatives and evidence emerges regarding the direction of future acquisitions.” – Daniel Jester, BMO
The $393 PT is not a conviction sell. It’s a “wait and see” on a stock already down 35% Y/Y. They’re not calling for a breakdown; they’re saying the r/r is balanced until the AI narrative clarifies.
BULL: Raymond James reiterated Strong Buy at $500 post-1Q beat (revenue $2.10B vs $2.06B est, organic +6% vs 3.9% est). Margins held at 38.1%. $1.5B buyback authorized. The collection of niche verticals is inherently defensive – small, predictable end markets that don’t get vaporized by AI overnight. The bolt-on M&A pipeline remains active (Illumia, Aderant AI launches show they’re not sleeping). At 22x P/E with mid-single-digit organic growth plus M&A, the floor is real.
BEAR: BMO’s caution is the right hedge. ROP’s model depends on buying stable, non-disrupted vertical software businesses. If AI starts commoditizing or compressing margins in those niches – think ERP, medical devices, compliance tools – the compounding math breaks. The “forever ownership” mindset becomes a liability if the assets age faster than expected. And $9B of acquisitions in ~3 years means a lot of integration risk in a shifting tech landscape.
This is a classic “earnings quality is fine, narrative is foggy” setup. ROP isn’t broken – it’s a 22x P/E compounder with a buyback and a beat. But BMO is right that the market needs to see how AI changes the trajectory before paying up. For PMs: this is a hold, not a hero trade. Let the evidence bake.
Verdict: The 11% pullback from the 52-week high is an entry, not a trap. Both DA Davidson ($74 PT) and Rosenblatt ($63 PT, raised from $52) are leaning hard into the semi analytics story post-offering dip. Consensus sees the 20% CAGR target through 2030 as achievable and "potentially conservative." But at 310x trailing P/E you're paying for a lot of future — this is a high-conviction, high-premium name.
THE ANALYST VIEW: Two bullish calls, same core thesis — PDFS is a unique asset in the semiconductor supply chain, combining data, analytics, and orchestration capabilities no one else has. DA Davidson doubles down post-pullback, citing "various levers" for upside to 2026-27 consensus. Rosenblatt’s upgrade cycle continues: new PT $63 on a blended 2027-28 valuation, positive read from semi capex vendor commentary this week, and a key competitor exiting the manufacturing analytics market. Both see the recent equity financing (3.8M shares, 500k primary) as enabling DFI market development, not a desperate capital raise.
"We view the recent pullback as a buying opportunity and believe the company's target of a 20% revenue CAGR through 2030 is achievable and potentially conservative." — DA Davidson
BULL CASE: PDFS is a secular winner in semiconductor complexity — Exensio, secureWISE, and DFI all gaining traction. Q1 beat (EPS $0.31 vs $0.21 est, rev $60.1M vs $59.7M) shows momentum. Backlog stable at $246M. H2 should accelerate on new product launches and renewals. Competitor withdrawal removes noise. 20%+ top-line growth with expanding margins is the path to re-rating.
BEAR CASE: The valuation is egregious — 310x P/E, 35x 2027 EBITDA. The equity offering (6.9% drop) signals a need for cash even if framed as growth investment. Semi capex is notoriously lumpy; a macro slowdown would crush the premium multiple. Stock is up 137% in a year — mean reversion is real. Not sure we can read too much into a competitor exit without knowing actual competitive overlap.
THE NUMBERS: DA Davidson PT $74 (35x 2027 EBITDA). Rosenblatt PT $63 (blended 27-28). Current: ~$57. Key levels: $52 (Rosenblatt's prior), $56 (DA's prior before this note), $71.69 (52-week high). Q2 print expected early August — platform revenue growth is the swing factor.
Verdict: Equipment cycle is still accelerating, not peaking. Analysts are layering in 2028-29 WFE estimates that imply a multi-year capex supercycle—AI memory + leading edge logic is sticky. LRCX sitting at $353, price targets clustering at $400-425 (Cantor outlier at $500). The bull case is that we're only in inning 3 of this up-cycle. The bear case: semi equipment is structurally lumpy and consensus is already pricing perfection.
Mizuho raised PT to $400 (from $380), Stifel to $425 (from $325). Both maintain Buy/Outperform. The real signal is in the WFE forecast escalation:
BULL: AI-driven memory (HBM) capex is accelerating into 2027-28, not slowing. Foundry/logic spending remains durable as leading-edge nodes ramp. LRCX has structural share gains in etch and deposition. Gross margins get a tailwind from better ASP environment H2. "Multi-year ramps" is the phrase—not a one-off.
BEAR: Semi equipment is historically volatile. Consensus is already embedding 25%+ WFE growth for two consecutive years. Any macro slowdown or memory oversupply (remember 2023?) crushes estimates. Stock up 250%+ in a year—that's a lot of good news priced in. The Cantor $500 PT feels like a "buy the stock, not the story" trap.
From Mizuho's note on AI chip demand:
"The firm sees upside to wafer fabrication equipment estimates in 2027-29 as AI demand supports durable ramps for leading edge logic and memory... cites a supply-demand gap continuing to grow given a global memory shortage."
That's the core thesis: supply-demand gap is widening, not narrowing. If you believe memory shortages persist through 2028, LRCX prints money. If you think HBM gets overbuilt by 2027, you sell into the strength.
I lean constructive. The WFE estimates feel aggressive but not insane—AI infrastructure buildout has a multi-year tail, and memory is the tightest it's been in a decade. LRCX's equipment is mission-critical for both HBM and leading-edge NAND. The $400-425 targets imply about 15-20% upside from here, which is reasonable for a name that could compound earnings at 20%+ for two more years. Not chasing it into the $500 Cantor dreamland, but I'd add on dips below $340.
Stifel raises PT to $120 from $75 (Hold) – stock at $112.54, so not much juice left near-term. HSBC went $200 (Buy) the same day. Clear divergence on where the turnaround lands. Stifel sees in-line to modest upside prints, but warns CPU/GPU expectations are already elevated. The real driver now is end-demand commentary on server CPU ASPs/volumes and GPU yields – not the quarterly beat.
"The longer-term story matters more for Intel at this stage of its transformation."
That's the Stifel hook. Margins and EPS face less downside pressure over the next couple quarters, but the stock is pricing a lot of success already. HSBC's $200 target implies Foundry traction is more advanced than Stifel's cautious Hold suggests. PMs should watch supply-side GPU commentary – that's the swing factor for the narrative, not the P&L.
BTIG kicks off coverage with Buy and $1,210 target — not exactly contrarian given the +36% YTD run to $1,035, but the thesis has teeth. They’re calling AI a general-purpose technology on par with electrification. The numbers are eye-popping: global data center capacity growing 86% to 262 GW by 2030, implying $5.1 TRILLION in cumulative spend. Hyperscaler capex projected at $663B for 2026 — that’s 4X since the AI boom started in 2023. Rent growth 63% over five years and still accelerating because supply can’t keep up.
“AI workloads require 10 times the power of traditional software applications, and legacy infrastructure struggles to support these demands.”
EQIX sits at the interconnection layer — purpose-built for denser racks and direct substation ties. The Cisco/NVIDIA partnership announced this week adds a tactical catalyst: standardized AI factory blueprints that make enterprise adoption easier. Stifel reiterated Buy at $1,250, citing increased enterprise activity moving from pilots to distributed architectures. William Blair dialed their data center power index to 75 (from higher), flagging local opposition and power constraints — a real headwind, but one that benefits existing operators with locked-in capacity like EQIX. Not a clean bull case (the power supply deficit is real), but the rate of change in demand still dwarfs the impediments.
KeyBanc bumps GOOGL PT to $445 from $425 (Overweight) — not a massive move, but the thesis is clean: cloud + search durability underpinning 20%+ revenue growth through 2028. Stock at $356.66, up 102% in the last year. Patterson’s revenue forecast sits 5% above consensus for ’27/’28. He no longer expects the torrid quarter-over-quarter backlog growth in Cloud, but argues the Street is underestimating the staying power of both Cloud and Search.
"Consensus underestimates the durability of Google Cloud and Search growth."
At 27.3x P/E with a PEG of 0.59, growth-adjusted valuation is not pricing in much deceleration. KeyBanc’s $445 target implies ~25% upside from here — on a 25x P/E (2028E), supported by sum-of-the-parts. The non-analyst headlines (€1B Austrian data center, AlphaEvolve for Cloud, ad transparency rules) are background noise. The signal is that KeyBanc sees a multi-year compounder trading at a reasonable multiple relative to its own growth trajectory.
Verdict: Still in the structural shortage camp. TD Cowen re-ups Buy after investor meetings with Mehrotra & CFO Murphy. Physical supply constraints persist beyond 2027. Customers with capacity agreements still can’t get full allocation — that’s a pricing put.
KEY NUMBERS: Stock at $975 (market cap ~$1.1T, up 700% in 12 months). PT unchanged at $1,600. DDR ASPs expected to print >15% Q/Q in CQ3, then mid-single-digit gains through H1 2027. That alone implies upside to Street’s $160 EPS for CY2027. Twenty-seven analysts have revised EPS up recently — momentum is real.
“Nearly 50% of total revenues are expected to be covered by supply and capacity agreements long term, spanning hyperscalers, enterprises, and automakers.”
The bull case: Supply locked in, demand AI-driven, ASP trajectory straight up. TD Cowen sees the shortage extending past 2027 — that’s a multi-year compounding story. The $250B U.S. capex plan and Clay, NY facility pour ahead of schedule reinforce that Micron is playing offense.
The bear whisper: Samsung’s earnings miss hit the sector last week (MU sold off). If DRAM/NAND contract prices soften on macro, the 700% run gets sticky. But Bernstein’s modest Q2 price uptick and TD Cowen’s >15% Q3 print argue the opposite.
Bottom line: This is a high-conviction structural shortage name. The only question is multiple — and with EPS estimates still moving up, the r/r favors longs.
Truist bumps PT to $130 from $120, stays Buy. Verdict: enterprise AI adoption in H1 2026 is real, and NOW’s positioned as a context provider that actually gets used. Not just a platform trade — it’s the “claude code eats the world” thesis for ITSM.
“The runway for continued enterprise adoption outweighs potential optimization impacts on overall spending in the second half of the year.”
That’s Miller Jump’s core take. He’s betting the H2 bifurcation (winners vs losers in infra software) favors NOW and Snowflake — broad platform presence + enterprise context moat. The talent war is heating up, and NOW’s ability to align pricing to outcomes (not seats) matters as CFOs get pickier.
Stock at $107.88 ( $111B market cap ) with 76.6% gross margins and 21.7% revenue growth. Not screaming cheap, but the narrative shift from “terminal value risk” to “AI adoption flywheel” is the real delta here. H2 earnings will tell us if this is a re-rate or just another positioning trade.
BTIG kicks off coverage with a Buy and $215 PT — the AI infrastructure thesis in its purest form. The firm argues the demand curve is structural, not cyclical: AI workloads require 10x the power of traditional apps, and the existing footprint can't handle it. Global data center capacity needs to hit 262GW by 2030 (up 86%), implying $5.1 trillion in cumulative spending. The kicker? Real demand could support 376GW if power, equipment, and capital weren't bottlenecked. That’s a supply shortage BTIG calls structural — $10B-$13B per GW of new buildout.
Hyperscaler capex hits $663B in 2026, up 4x since the AI boom began. Rent growth is 63% over the past five years. DLR delivers a solid 17.5% YTD and a 2.72% yield paid for 23 straight years. The bull case is straightforward: pricing power + secular demand + constrained supply. The bear case? The stock may already reflect it — InvestingPro flags overvaluation at current levels, and consensus PTs range from $180 to $250 with the stock at $179.
"AI workloads require 10 times the power of traditional software applications. Legacy infrastructure struggles to support these demands, prompting purpose-built data centers with denser rack deployments and direct connections to utility substations."
The macro tailwind is real. The micro setup has two near-term overhangs: the $7.8B Blackstone deal closing (DLR issuing $2.3B in shares + $1.2B cash for 64% of three NOVA data centers) and the associated $2.35B secondary offering (non-voting stock). That's dilution and an overhang, but PMs should note the equity-linked nature of the financing — Blackstone is taking shares, not dumping. The acquisition adds high-quality NOVA assets (the premier data center market) and signals continued hyperscaler demand. For a 20-year TMT covered-call book, DLR is a core infrastructure hedge with asymmetric upside if the AI capex cycle extends.
Stifel rips PT to $70 from $50, maintaining Buy. The call is straightforward: AI test revenue hits top-three segment for COHU in CY26, layered on top of a cyclical recovery in auto/industrial/mobility. At $55 and change, the stock is already up 168% in the past year — but Stifel sees room to re-rate. They apply 20x on CY28 EPS of $3.50, a 30-35% discount to test peers. That discount narrows if AI demand sustains AND traditional end markets inflect.
“The combination of AI-related growth and recovery in traditional markets should drive a higher valuation multiple.”
Other analysts are piling in, too. Needham and TD Cowen both raised targets recently (Cowen to $80, Needham to $54), Baird launched Outperform at $65. The consensus narrative is the same: Eclipse handlers and HBM inspection systems are the AI vectors, plus the DiamondX platform just scored a ~$5M order for GaN power devices in AI data center power. Q1 revenue beat, Q2 guide up 15% sequential. Rate of change is accelerating, and the PM crowd is rotating into test equipment as a late-cycle AI beneficiary. Not a huge name, but the setup is clean — cheap vs. peers, top-line inflection, AI tailwind hitting P&L in real time.
Stifel upgrades TWLO to Buy this morning, PT to $260 from $175. The stock is at $218.60 ($33.2B market cap, up 86% in the last year). The call is pure AI infrastructure narrative — Stifel sees Twilio as key plumbing for agentic voice agents, with its Conversations portfolio and data offerings catching the next wave. The restructuring is done (headcount -40% from peak), R&D re-focused on core + AI. Multiple other firms have piled on recently: Goldman at $300, Tigress at $255, Rosenblatt at $230, TD Cowen at $210. All cite the same constellation — accelerating revenue trajectory, margin expansion, and half of the Forbes 50 AI startups as paying customers.
"Agentic AI investments by Twilio’s customer base will generate meaningful results as voice agents become more widespread." — Stifel
The bull case writes itself. Twilio is the developer layer for the next cycle of voice-first AI apps. The 40% headcount reduction plus new leadership (CEO switch?) means cost discipline is now institutionalized, not cyclical. Margins have room to run. The bear retort: most of this is already in the stock at 86% Y/Y gains. The PT range ($210-$300) implies upside is mixed — even the Street isn't unified on a single number. And agentic adoption is still early; revenue acceleration needs to prove itself in actual billings growth, not just narrative. Verdict: the upgrade is a timing catalyst, not a new thesis. Stifel was late to the party, but the party might still have legs if Q2 prints strong.
Verdict: Buy the AI narrative, but don’t chase the near-term comp. Needham raised its PT to $430 (from $400) on a Buy, echoing a Street that just pushed targets to a $390-515 cluster post-print. Atlas grew 29% (3% beat), revenue beat by $23M, and full-year guidance got a bump. The question isn’t if AI spending hits MDB — it’s when the external-facing apps reach mass scale.
“The comparison between MongoDB and other consumption-based infrastructure software vendors is unfair because each company operates at different layers of the technology stack with different monetization models.”
Multiple firms (Stifel $435, BofA $390, Tigress $515, Cantor $416) all leaned in after the quarter. The bull case hinges on Atlas as the AI data bedrock — the bear case? Production migration timelines stay lumpy, and the 24% revenue growth needs to accelerate to justify the 73% 1Y return. Right now, r/r is OK, not great. Wait for the next catalyst (probably a big customer win).
Wells Fargo just threw another log on the fire. PT to $730 from $575, Overweight maintained. The firm is now 6.5% above consensus for CY27 and 16% above for CY28 on revenues of $20.9B and $27.4B respectively. That implies a 60.4% gross margin in CY27 and 65.1% in CY28 — levels that make the legacy HDD narrative look stale.
"Customer engagements are extending into calendar year 2030 and beyond, with nearline hard disk drive production capacity allocation through calendar year 2027 and extending visibility into calendar year 2028."
That's not just a bullish data point — it's a structural repricing of the HDD cycle. Major CSPs are locking in supply six years out. The bull case: WDC and Seagate are becoming annuity-like capacity plays, not cyclical component vendors. The bear case? That visibility is already priced in at $199B market cap and a 791% YoY move. But Wells is still calling for another +26% to $730. Hard to argue with order books that stretch to 2030.
Other noise: post-quantum crypto integration on Ultrastar drives (security sell to hyperscalers), plus a Sandisk share swap settlement in May. Evercore also raised to $575 (now below Wells). Lynx called NVDA a better buy than SpaceX — irrelevant here but shows the rotation appetite.
Verdict: The demand visibility argument is steel-pipe now. PMs should watch if Seagate gets the same treatment (Wells raised Seagate too). Risk/reward still favors the long side given the contract backlog, but the entry is rich.
ICHR IS A 4-BAGGER ON THE YEAR FOR A REASON. Stifel rips its PT to $115 from $76 (Buy) on a growing conviction that the vertical integration/proprietary product strategy is finally hitting an inflection point. Stock at $96.44 — up 423% YTD, ~307% one-year — so the easy money is gone, but the thesis is getting crisper, not fuzzier.
The old ICHR was a gas panel integrator for etch/deposition tools — grew with customers but captured limited value. New CEO Phil Barros (promoted Nov '25, former CTO) is shifting the model to own more of the stack. Stifel frames 2026 as a "transitional period" but says they're now MORE convinced about both revenue growth and structural margin expansion.
The gross margin is 12% — horrid for a semi supplier — but Stifel sees that as the swing factor. If they execute, that number doubles. If it doesn't, the stock is pricing in an outcome that hasn't happened yet.
"The firm believes the stock warrants trading more in line with semiconductor capital equipment and diversified supply chain peers."
(Translation: There's multiple expansion left if they deliver the margins.)
The counterpoint: 423% YTD is pricing a lot of good news. Needham also raised to $72 (from $48) post the Q1 beat — revenue $256M vs $236M consensus, EPS $0.15 vs $0.06 forecast — but $72 is still below current price. The bull case needs margin execution, not just revenue growth. At 12% gross margins, one hiccup in the supply chain or a capex cycle slowdown and this thing gets cut in half.
Bottom line: High conviction, high execution risk. Worth watching for the 2027 setup, not chasing the 2026 momentum. PMs should let the Q2 print (due early Aug) confirm the margin trajectory before adding.
Verdict: KeyBanc is the loudest bull — $175 PT from $130, Overweight — and they’re not alone. The consensus story is AI identity plumbing, and Okta is the lead pipefitter. Stock already up 72% YTD, trading at $148.90 (near 52-wk high $153.20). The question is whether Mizuho’s downgrade to Neutral on valuation is just noise or an early warning.
The KeyBanc upgrade is the headline. They’re leaning hard into Okta’s AI agent security products (Okta for AI Agents, Auth0 for AI Agents, token vault, ID-JAG/XAA, A2A) and calling its go-to-market execution “early-mover.” Gross margin at 77% helps the math.
“Okta’s access management capabilities and market leadership, combined with its development of the ID-JAG standard and alignment with Anthropic, position it as the critical layer for securing AI agent interactions.”
The broader analyst picture is mixed but leaning constructive. Cantor Fitzgerald and UBS both raised PTs (to $125 and $150 respectively) citing strong Q1 beat and cRPO growth of 12% YoY. Scotiabank went to $135, explicitly calling out “securing AI agents” as the next leg. Mizuho downgraded to Neutral from Outperform — not on fundamentals, just that the stock has run too far too fast (up 72% YTD). That’s a valuation call, not a thesis breaker.
Bottom line: The bull case is about identity becoming the trust layer for AI agent-to-agent and agent-to-resource interactions. Okta has the installed base and the product roadmap. The bear case is that the 72% YTD rally already prices that narrative in, and Mizuho is the first to blink. For PMs: if you believe AI agent adoption accelerates, Okta is a direct play. If you think multiples matter, wait for a pullback. The 1.7% revenue beat and 12% cRPO growth aren’t screaming acceleration — they’re steady execution. The AI premium is real but fragile.
Verdict: Mizuho bumping PT to $415 from $400 with an Outperform is a modest reaffirmation, not a fresh catalyst. The real story is the WFE spending trajectory they’re embedding — 2027 jumps to $192B (+25% y/y), with 2028/2029 at $221B and $214B. That's a long-duration AI capex buildout thesis, and MKSI sits right in the funnel (critical subsystems, vacuum, photonics). But the stock already up 232% in a year — the easy multiple expansion is behind us. This is about earnings delivery vs. expectations now.
Mizuho's move is a downstream read-through of memory and advanced packaging tightness. They're explicitly calling a growing supply-demand gap in global memory and HBM capex accelerating into 2027-2028. On the packaging side: TSMC CoWoS capacity hitting ~140k wpm by YE 2026 and 200k+ wpm by YE 2027. MKSI is levered to both (ionization, power, gas delivery for etch/dep, plus Atotech for packaging chemicals/equipment).
"Mizuho raised its estimates for global wafer fabrication equipment spending in 2027 to $192 billion, up 25% year-over-year following an expected 23% increase in 2026."
That $25M Guangzhou expansion (doubling capacity by Q4 2027) confirms they're betting on the physical buildout, not just hoping. BMO also initiating with Outperform is noise — it's a one-off initiation, not a cluster — but it reinforces the positive skew.
The risk the article doesn't hammer: MKSI's optics/spectroscopy and laser segments are more cyclical than the semi capital equipment narrative implies. If AI front-end spending peaks before the packaging ramp fully kicks in, there's a gap. But right now the tape is buying the WFE super-cycle story. Rate of change is positive. Positioning is crowded. Watch for the next print.
UBS RAISES PT TO $35 FROM $32, STAYS NEUTRAL — but the tone shift matters more than the bogey. Hosted an investor lunch with CEO/CFO/IR. Came away feeling better about the near-term demand backdrop. Key areas of conservatism from the last print (macro, price-sensitive cohorts, DAP contribution, tech layoffs, reorg disruption) all "generally sounded stable to better" vs. where they were a month ago.
"Management commentary around strength in seat growth, described as 'very healthy,' stood out as the most positive datapoint."
That's the headline. Seat growth has been the variable everyone's been sweating. If that's stabilizing — or accelerating — the bear case around GTLB being a code-spend rationalization victim gets harder to make.
DAP still a FY28 story per UBS. Largest customer they spoke with is planning a pilot early C4Q after upgrading to 18.8 in late C3Q. Nothing imminent. FY27 guidance looks "increasingly conservative" but at ~3.7x CY27 revenue, r/r is balanced from here. Not enough to get constructive without more proof on seat growth stabilization and DAP traction. Fair.
OPPENHEIMER RAISED PT TO $255 FROM $225 — still a 23% bogey from here, but the real story is the data center ramp. The new CEO (Aamir Paul, ex-Schneider North America) + $735mm in E-Pod wins from late 2H25 = low double-digit organic tailwind in 2027. That’s not a 2026 story yet — shipments mostly hit next year, possibly spilling into 2028. DA Davidson and Morgan Stanley both chimed in with Buy / upside calls, so the street is coalescing around the roll-forward thesis.
“Regal Rexnord’s organic growth is expected to receive approximately a low double-digit percentage boost in 2027 from $735 million in E-Pod wins secured in late second half 2025.” — Oppenheimer
The switchgear datacenter book is already visible: $180mm in 2026 (+50% YoY), stepping to $240mm in 2027. Short-cycle prospects also improving — that’s a two-engine setup. Stock up 49% YTD (still only $13.9B market cap, 49x P/E), so the multiple is already pricing in some of this. Risk? If 2026 E-Pod shipments slip further into 2028, the narrative gap widens. For now, the rate of change on datacenter exposure is accelerating, and the CEO swap gives credibility on execution.
Verdict: Street is getting more comfortable with the identity security narrative – but the real math is about the subscription ramp. BMO raised PT to $98 (from $87), Outperform, citing durable growth from Identity Recovery/Resilience and potential RAC contribution in FY28. That’s on top of a revenue print that beat consensus by 5.7% – SUBSCRIPTION ARR GROWING 32%, net new ARR of $103M vs $85.5M bogey. Revenue surged 46% LTM to $1.42B. Profitable this year is the baseline expectation.
The identity thesis is the linchpin. Two products: Identity Recovery (six quarters live) restores AD/Entra/Okta; Identity Resilience (three quarters) adds posture monitoring. BMO thinks Rubrik is displacing Semperis here. That’s a real wedge, not just bolt-on.
“Rubrik’s identity recovery solutions are largely complementary to identity security providers’ capabilities at this time, though the firm believes Rubrik is replacing Semperis.”
Sell-side consensus is tightening: Baird Fresh Pick at $110, Cantor OW at $95, Truist/DA Davidson both Buy at $90. Not a crowded callback but the narrative is coherent – durable growth, expanding TAM, and a clear path to profitability. Risk? Multiple is still pricy at these rev levels – but the rate of change is in their favor.
BERNSTEIN STICKS TO TOP PICK CALL INTO EARNINGS. PT up a measly $2 to $220 (stock at $160, up 107% in the past year). Not exactly a barnburner, but the conviction is the signal – they see positive surprise potential as liquid cooling capacity ramps. The big question: can NVT keep the 107% momentum?
Bernstein acknowledges margins have been a worry, but they expect operating leverage in the SP segment and pricing/productivity fixes in EC to close the gap. Uneven orders remain a watchpoint, especially in SP.
"Expectations are elevated among investors who follow the stock but identified potential for positive surprises, particularly as liquid cooling capacity increases."
That’s the entire bull case in one line – liquid cooling is the catalyst, and NVT is the plumbing play. (UBS reiterated Buy alongside this, expecting >25% organic sales growth from data center demand.) The modest PT lift suggests not much upside to the $220 level from here, but the top pick designation is the real takeaway – they think NVT beats the crowd into the 2Q print (due in ~21 days).
Reiterate Buy from UBS at $2125 PT on Comfort Systems. This thing has returned 216% in the past year — and the datacenter thesis is still intact, not exhausted. UBS came out of management meetings pointing to multi-year hyperscaler buildouts, strong pricing conditions, and scale advantages that let FIX bid bigger projects at better margins while offering competitive wages to grow labor. That's the kind of structural moat PMs should love.
The catch: book-to-bill will likely print at normal seasonality of 0.9-1.0x. For a stock priced for perfection (it is), that might spook the momentum crowd who want 1.2x every quarter. UBS expects Q2 upside vs Street on the income statement and post-earnings commentary to stay bullish. So the fundamental path is higher — just don't expect a blowout bookings number.
"Post-earnings commentary is expected to remain bullish, with estimates moving higher to support investor conviction."
We already saw a massive Q1 beat ($10.51 EPS vs $6.78 consensus). The AI infrastructure buildout is real, and FIX is still the best pure-play labor-constrained beneficiary in the space. Goldman and Oppenheimer just started coverage with Buy/Outperform, adding incremental fuel. The stock needs to show it can scale margins as revenue grows — that's the next catalyst.
KVYO IS GETTING MARKED DOWN FOR GUIDANCE PRECISION, not for weak demand. Stifel cuts PT to $24 from $28, still Buy. Stock at $17.16 — that's a 40%+ implied upside even at the lower target. The problem is the beat pattern.
The company told Stifel Q1's 2.9% revenue beat is "a good proxy" going forward. That's a step down from the 4.0-5.4% range in 2024/2025. Management's own line: "the tighter beat reflects that improved ability to forecast the business." Stifel now flags a 3% beat as the new benchmark. The implied deceleration is real — partially from lapping profile enforcement benefits.
Multiple firms have trimmed targets (Goldman $26, Canaccord $32, Piper $26) while maintaining bullish ratings. The base case: 30% revenue growth, turning profitable this year, but the multiple is compressing as the beat shrinks. Not a broken story — just a reset on expectations for how much they'll clear the bar.
Bottom line: KVYO is a r/r play on whether the tighter guidance is a one-time transition or a permanent lower beat rate. If growth holds and margins inflect, $24+ looks cheap. But the stock needs to prove it can beat by more than 2-3% to re-rate.
CEVA getting a lot of love this morning — and for good reason. Stifel just bumped its PT to $50 from $42, Buy maintained, bringing the street consensus higher alongside Needham ($55), UBS ($48), and TD Cowen ($45). That’s a tight cluster in the $45-55 range, and the stock is already at $46.82 after a 102% run over the past year. Not a ton of upside left from here on pure multiple expansion, but the thesis is about a revenue inflection, not today's print.
Stifel laid out the three-part bull case cleanly: (1) licensing shifting to higher-value system-level engagements, (2) a multi-year royalty ramp in Wi-Fi/combo chip/cellular IoT that's still early innings, and (3) an AI/NPU royalty stream that goes from interesting to MEANINGFUL IN 2027. The full-year 2026 is a transition year — licensing wins from the past 12-18 months convert to royalties starting next year. The PT is based on 8.5x EV/Sales on FY2027 revenue, which compares to a 5-year median of 4.4x. So they’re paying up for that inflection, but the range history goes to 13.8x, so it’s not insane.
"2027 setting up as a potential inflection in growth as prior licensing wins convert to royalties." — Stifel
The narrative got a big assist from the recent announcement of a licensing deal with a MAJOR US SOFTWARE AND AI PLATFORM COMPANY (customer undisclosed) for a custom AI silicon program using Ceva’s NeuPro-M NPU IP. That’s the kind of high-value, integrated engagement Stifel is talking about. And the Q1 print validated it: $27M revenue vs $26.1M consensus, licensing revenue of $17.8M (best in three years), adjusted gross margin 87%.
Balance sheet is a fortress. Current ratio 10.26, more cash than debt. Optionality in robotics and physical AI is a nice kicker but not priced in yet. The risk? Edge AI proliferation is real, but Ceva's IP model means they get a smaller piece of a growing pie — revenue scale matters. At $50, it’s not cheap on current numbers, but the trajectory is clean if the royalty conversion materializes. PMs positioning for 2027 should be watching royalty revenue prints in the next two quarters for confirmation.
SKHY – ADR DEBUT PRICED AT $149, OPENED $170-181, CLOSED 13% HIGHER. $26.5B raise – largest foreign IPO ever – 7x oversubscribed. Pure HBM exposure with Chairman saying even doubled capacity over 5 years won't meet demand. The key debate: cyclical peak multiple or structural re-rating. Bulls point to LTA structures and multi-year prepayments; bears say 80%+ gross margins mean-revert. What this changes: ADR creates second pricing anchor for HBM, compresses MU discount, forces re-rating of the memory complex. Rate of change accelerating – demand outrunning supply faster than consensus models.
NANYA – REVENUE UP 684% Y/Y, GM 79.5%, ASP UP >500% Y/Y. This is staggering pricing power in legacy DRAM. Clean-room capacity coming 2028-2030 doesn't address near-term scarcity. The read-through: tightness extends beyond HBM into commodity DRAM. Rate of change accelerating – ASP up >60% q/q. This legitimizes the super-cycle narrative across the memory food chain.