Good morning.
Semis ripping, software getting eviscerated. SOX at fresh highs — NINTH +5% SESSION IN 60 DAYS. MU prints ATH after the Anthropic deal. Alphabet -6% on AI talent exodus; SpaceX shed $800B in a week. CRM on its 15th straight red day. BofA now pencil in 3 Fed hikes. Quarter-end rebalancing massive — $165B flow out of equities.
Three themes framing today:
1. Memory bottleneck monetized. MU takes equity in Anthropic — not just cash — for supply priority. Memory players are extracting scarcity rents above the hyperscaler layer. Structural shift in allocation risk pricing. Watch SK Hynix, SNDK follow.
2. Grid is dead. Long live behind-the-meter. MSFT/Chevron sign 2.6 GW off-grid PPA, zero ERCOT draw. Stranded Permian gas turned into AI compute. This is the new blueprint. E&P majors (XOM, Coterra) will copy.
3. Compute efficiency destroys scarcity premium. GB200 serving costs down 2.5X in 70 days via software alone. CUDA eating its own scarcity narrative. H100 rental index rolling over — first canary. Hyperscalers win (usage monetization), hardware multiples vulnerable if rental rates break.
We'll hit up MU, MSFT, and CRM first, then get to semis & power infrastructure...
Memory is the new oil. Three shops just ripped price targets to $1,200–$1,550 from a prior ~$500 cluster. Needham leads at $1,550 (10x FY28 EPS of $155), Bernstein at $1,300 (switched to P/E from P/B, using trough multiple on peak earnings), Wedbush at $1,300 (9x FY27 EPS + net cash).
The collective thesis: this cycle is different — DRAM shortage more acute and longer than expected, LTAs providing multi-year demand visibility, and AI-driven HBM pricing staying elevated through calendar 2027. Bernstein now sees DRAM industry gross margins hitting 70% exiting CY28 (only 2018 was higher). Needham’s analyst says it best:
“Reflecting this multi-year visibility, together with the strategic importance of memory in AI, we believe investors should be comfortable paying a higher valuation multiple for memory stocks.”
Bull case: DRAM supply discipline + AI insatiability = 18+ months of pricing strength. The PEG ratio is 0.13. Earnings estimates are still going up — 12 upward revisions in the last month. Wedbush says “no reason to shift positive view” and expects estimates to keep moving higher.
Bear case: 820% YOY rally is pricing in perfection. Bernstein’s own model shows revenue declines in FY28 as clean rooms come online. Peak P/E multiples on peak earnings is a dangerous combo — ask anyone who owned memory stocks in 2018. The stock is 2% off its 52w high ($1,149) heading into earnings June 24. Derisking is not guaranteed.
Verdict: JBL is the AI manufacturing play that’s working — and the Street is scrambling to price it. Three firms raised PTs to $430-470 (from a prior $380-430 cluster), with Stifel and BofA both maintaining Buys, UBS staying Neutral. The thesis is simple: beats, raises, and a third hyperscaler win. The debate is whether 51x P/E is already pricing FY27’s $20B AI revenue or if acceleration in margin and cash flow deserves a multiple.
Revenue of $8.8B (2% above consensus), adjusted EPS of $3.16 (beat by $0.08), and operating margin of 5.8% (vs 5.7% expected). Management lifted FY26 guidance to ~$35B revenue, $12.70 EPS, and >$1.4B FCF — and revised FY26 AI revenue to $13.6B (+50% YoY). Early call for FY27 AI revenue: “similar percentage growth” → ~$20B. That’s the headline.
The third hyperscaler win (name not disclosed) is supposed to add “several hundred million” in FY27 and >$1B in FY28. Plus the Adani alliance for multi-gigawatt rack manufacturing in India starting FY28. Management is signaling a multi-year infrastructure cycle, not a one-quarter pop.
UBS flags that HII revenue growth decelerated to 21% (from 52%) against tough comps, with $200-300M shifting into Q4 on delivery timing. Normalized Q4 growth looks more like 23-26% — still strong, but the rate of change is slowing. UBS sticks with Neutral.
Bull case (Stifel, BofA): You’re buying a company with verified hyperscaler relationships, a 50%+ AI revenue CAGR, expanding margins (BofA sees +20-30bps annually), and a PEG of 0.62. The third hyperscaler de-risks dependency. Free cash flow is $1.4B+ and growing — capital returns are real.
Bear case (UBS, and the 51x P/E): The easy comps are behind. Revenue growth in Q4 is projected to decelerate to ~13% normalized even with the shift. Without AI, the rest of the business is growing mid-single digits. At $374, you’re paying for FY27 perfection — any miss on the $20B AI target and the multiple compresses hard.
Stifel, in raising to $460:
“Jabil continues to execute at a high level, leveraging its position in AI infrastructure to drive above-consensus results. The third hyperscaler win and FY27 AI outlook reinforce our view that the secular tailwind is accelerating, not peaking.”
Our take: JBL is a high-conviction AI industrial name, but the risk/reward is narrowing. The bull case requires FY27 AI to hit $20B and margins to hold 6%+ — both plausible, but priced in. We’d look to add on pullbacks, not chase the 83% Y/Y return. (We own it in the long book, trimming into strength.)
THE TAKE: THE SHAZEER HIT IS REAL, BUT THE MARKET IS OVERWEIGHTING IT. Two analyst notes this morning both flag the same risk — Noam Shazeer jumping to OpenAI — but neither is cutting numbers or downgrading. The stock has already repriced (down 8% from highs), and the bull case still rests on a distribution moat + Cloud acceleration that no single exec departure derails. That said, the narrative risk is sticky.
Citizens (Outperform, $515 PT) and Jefferies (Buy, $445 PT) both addressed the Shazeer departure. Citizens frames it as a return to the bear case from 2-3 years ago: talent drain to OpenAI/Anthropic. But they point out that Alphabet has delivered 121% RETURN over the last year despite those fears. Jefferies goes further — calls the talent churn "expected to continue" given AI scarcity — and says the company's "deep bench" and "long AI heritage" offset it.
"The bear case on Google from two to three years ago centered on concerns the company would lose talent to OpenAI and Anthropic and fail to catch up. We continue to monitor executive movement for that reason." — Citizens
BULL: Massive distribution (search, YouTube, GCP), Cloud rev accelerating, TPU vertical integration. PEG ratio of 0.6x makes this cheap if growth holds. Options flow yesterday hit 928K CONTRACTS (2.2:1 call/put ratio) — smart money leaning bullish.
BEAR: Talent flight is a slow bleed. Shazeer co-led Gemini. Acqui-hire risk (paid $2.7B for Character.ai license) now looks like a short-term fix. Multiple normalizing from premium (Jefferies PT at 20x NTM EBITDA vs. historical 22-24x). Magnificent Seven rotation ahead of frontier-lab IPOs means this group is an ATM.
Waymo recalled another 3,900 ROBOTAXIS for a software issue (construction zone mapping). Second recall in six weeks — minor operational hiccup, but optics matter when autonomous narrative is hot/cold. UK's CMA slapped new conduct requirements on Google Search (data portability, transparency). No near-term rev impact, but regulatory tail never ends.
Meanwhile, HSBC signed a multi-year Google Cloud deal using Gemini. And Google is taking a $75M STAKE in movie studio A24 for an AI research partnership — small check, but signals enterprise + creative vertical expansion.
Jefferies holds at $445, Citizens at $515. The spread tells you the range of conviction. I lean toward the higher end given the Cloud ramp and the fact that 121% returns tend to follow talent scares, not preempt them.
Verdict: still a story stock, but the optical thesis is getting real enough that 13 analysts have juiced estimates. Stifel just ripped their PT to $350 from $250 after a two-day European roadshow with management. Evercore initiated at $325 with Outperform. The stock is up 218% in twelve months and trading at $302. That’s priced for perfection, but the roadmap is starting to look like it can deliver.
Stifel’s read from the roadshow (CEO Brennan, CFO Fleming, VP O’Neil): three products on track to be $1B+ revenue businesses long-term. The optical DSPs cross $100M in FY27. Same for silicon photonics PICs. ZF AOCs hit $300-400M. Management expects to double optics capacity exiting FY27, triple it exiting FY28. That’s not a slide deck promise — they’re building it now.
Evercore sees the existing $5-10B TAM expanding 10-20x via optical. They model EPS >$13 in 2028, which is 40% above consensus. That’s a 3-year earnings CAGR >70%. At $325 (30x 2028E discounted back), the PEG is 0.4x. Growth at a reasonable price — if you buy the numbers.
Blockquote from Evercore:
"Credo created the AEC interconnect standard and delivers complete cable and chip systems rather than chips alone. We expect its product roadmap to expand its existing $5-10B TAM by 10-20 times."
BULL: Credo owns the AEC standard. Neoclouds are a new 20% mix vector over time. Optical DSP + SiPho + ZF AOC = three $1B+ legs. Edge inference is the killer app for OmniConnect. This is a 3-5 year compounder at >70% EPS CAGR.
BEAR: 218% in a year — this thing is on InvestingPro’s Most Overvalued list. Rosenblatt is Neutral at $215. Optical volume doesn’t really ramp until late FY26/FY27. Execution risk on tripling capacity. Multiple is already discounting the FY28 story.
CRDO is not a value play. It’s a conviction bet on optical AI connectivity scaling faster than the market expects. The analyst cluster (13 firms revising up, PTs $250-350) is unusually tight for a stock that’s already up 3x. That’s the rate of change you want to see. Watch the optical revenue split next quarter — that’s the tell.
DOUBLE UPGRADE DAY. Barclays and JPMorgan both pulled the trigger this morning — Overweight/Outperform from Equalweight/Neutral. Combined price target range of $145-$165, implying 27-45% upside from $114. That’s a coordinated message: the auto supplier grind is over.
JPMorgan's Rajat Gupta:
“Visteon offers higher visibility and limited execution risk compared to new products and new end-markets at peers.”
Bull case: VC is a structural compounder hiding in plain sight. Toyota launches, Honda two-wheelers, TRATON, China AI-HPC wins, India penetration — the win pipeline supports $5B revenue / $700M EBITDA by 2029 (JPM). SmartCore HPC and D6Sigma edge AI (Qualcomm collab) open TAM in industrial. Cash-rich balance sheet ($3.04B market cap, >$400M EBITDA) means they can self-fund.
Bear case: Auto end-market is still a zero-sum game. European OEMs losing share in China, tariffs, EV adoption slowing. VC’s 20% YTD run already prices in some of the turn. If growth disappoints (say 4-5% not 7%), the multiple stays stuck. Trailing EBITDA of $414M is real, and 2026 estimates still show no growth vs. TTM. We need deliveries, not hopes.
I like the risk/reward here more than most auto suppliers. The multiple is cheap because the market has been punishing negative organic growth for three years. That’s an anchor that lifts if the narrative shifts to structural outgrowth — and both Barclays and JPM are betting on that shift. Investor day on the 25th is the logical trigger for a squeeze into year-end. Not a high-conviction alpha name, but a decent pair against weak-handed auto shorts.
THE STOCK IS $222 — AND NOBODY'S PT GETS THERE. BofA at $195 (Underperform), Cantor at $200 (Neutral), even the bullish Bernstein upgrade to $210. The market is pricing in data center moonshots that analysts see as plausible but far from guaranteed. Investor Day Wednesday is the catalyst — either the narrative gets a concrete roadmap, or the stock gives back some of that 50%+ YOY gain.
Three firms moved PTs this week — BofA to $195 (from $165), Cantor to $200 (from $150), Bernstein to $210 (upgrade to Outperform). All cluster below the tape. The collective thesis: QCOM is diversifying from handsets into a broader AI compute platform — edge, auto, IoT, and now data center. BofA sees a $2-5B addressable opportunity by FY27/28. Cantor math: if data center revs hit $30B, EPS could hit $25, implying a $300 stock. But that's a huge if.
The bear case is simple: QCOM is re-entering a data center market it has failed at before, against Nvidia, Broadcom, AMD, Marvell, Cerebras, ARM AGI CPU, plus custom chips from Amazon and Google. BofA calls it "hyper-competitive." Cantor is skeptical of CPU traction outside the server head node. The stock's current valuation is already discounting some success — but not the full moon.
Bull case: AI feature cycle in smartphones (Bernstein upgrade) is real and sticky. Data center optionality is asymmetric — if the Tenstorrent acquisition ($8-10B) closes and QCOM delivers a credible server CPU, the multiple re-rates. Physical AI / edge compute (Neura Robotics partnership) adds longer-dated upside.
Bear case: The PTs are all below the stock. That's rare and telling. Even bullish PTs require data center revenue that competitors will fight to block. Handset growth is mature. At 15x forward EPS, the floor is $165-180 — a 20% downside from here. The risk/reward vs. the incumbents (NVDA, AVGO) is unfavorable on relative multiples.
"We remain skeptical of Qualcomm’s CPU success outside of the server Head Node, where share will be limited, and expect XPU and Custom ASICs to take time for sustained success." — Cantor Fitzgerald
Wednesday's Investor Day is the only game in town. PMs should watch for: (1) a quantified data center revenue target (anything above $5B by FY28 is bullish), (2) Tenstorrent confirm/deny, (3) any guidance on non-handset revenue mix approaching 50%. If the narrative is big but vague, the stock drifts. If they deliver concrete numbers, the short-term momentum could carry it to $240-250 — but that's already pricing in a lot of execution trust the incumbents don't give them.
THE AI TAILWIND IS REAL — AND BTIG IS ALL IN. PT raised to $37 from $30 (Buy maintained), with the stock already sitting at $33.26, up 87% YTD and kissing 52-week highs. BTIG's conviction is building after a direct chat with VP Corp Dev David Schroeder pre-quiet period. The TrojAI acquisition is the catalyst — it broadens the security roadmap and gives the AI firewall story legitimate teeth.
"BTIG came away from the discussion more constructive on the stock" — specifically citing AI as a structural tailwind for next-gen network/security products.
Customer concentration (31% of TTM revenue) is the obvious bogey, but BTIG says dynamics are healthy and sees improving trends across the remaining enterprise/SP base. Confidence in that 12%+ annual revenue growth target is actually improving — and the LTM numbers (revenue +12%, gross margins 79%) back it up.
Not alone in the bull camp either. BWS Financial went to $45 from $30, calling for accelerated revenue and FCF from AI demand. ATEN is getting multiple upgrades in a tight window — that tells you the rate of change is real, not just noise.
Risk? The 52-week high is $33.63 — we're 1% away. Positioned for a breakout, but at 87% YTD, a lot of good news is priced in. That said, if the TrojAI integration and the enterprise base re-acceleration holds, the next leg is $37+ and maybe that $45 target becomes the floor.
Oppenheimer raises PT to $450 (from $370), Outperform. The call is all about the Dragonfly G5 AI platform – demand trending stronger than expected with no split orders at key 2.5D logic layers. The stock trades at 31x 2027E P/E, a massive discount to KLA/Cameca at 48x (normally at parity). That gap is the whole r/r.
"Recent meetings reinforced that Onto Innovation is entering the 'good part of product cycle' with company-specific growth opportunities layered on accelerating industry growth."
Consensus is building. Since this article, seven analysts revised EPS upward. Freedom Broker, Morgan Stanley, Deutsche Bank, and Stifel all initiated or reiterated bullish stances with PTs from $350-$370 (Oppenheimer now the high at $450). The $1.3B convertible note (upsized from $1.1B, conversion ~$381.80) adds a small overhang but signals management confidence in the equity. At 31x forward vs. 48x for peers, this is still a catch-up trade – not a peak narrative.
Verdict: Best SMID idea at TD Cowen, but a 119% YTD run means you’re paying for perfection. The call is loud and clear: data center leasing hit ~9.4GW in Q1, lead times elongating, and FPS is the picks-and-shovels play. TD Cowen bumps PT to $73 from $63 — that’s ~18% upside from here. The blockquote:
"Well positioned to capture NT upside."
The thesis is pure rate-of-change on data center buildout. Revenue growth forecast at 85% for 2026, 6 analyst estimate revisions up. Both TD Cowen and Jefferies (PT to $56, Buy) are leaning into the same narrative: backlog duration extending, market share gains, sustained demand.
But let’s be real about the setup. The stock printed $2B in a secondary at $47 (32.8M shares), largely from Neos Partners cashing out. That’s a massive overhang being digested. At $61.95, it’s already close to the $65.56 high. The risk/reward here is binary — either you believe the data center demand wave is still accelerating and FPS is under-owned (fair), or you think the secondary + YTD run have front-loaded returns. I lean the latter for the next few weeks, but the structural story is legit. Watch the next lease print.
Verdict: Buy. Channel check with a digital agency exec reinforces the product cycle thesis. Rosenblatt’s Catherine Trebnick hosted a call with Matthew Weinberg of Happy Cog (85 employees, serves startups to Fortune 1000). He’s the primary tech vendor decision-maker and has hands-on production experience with Twilio’s auth and comms platforms. Takeaway: the new AI-powered tools — Conversation Orchestrator, Memory, Relay, Voice AI — are real, not vaporware.
Trebnick described Weinberg as “balanced and credible, neither overselling nor underselling.”
That’s the kind of check that moves the needle for PMs trying to separate hype from adoption. Rosenblatt stays at Buy, $230 target (22% upside from $188.11). Not a huge number, but the broader sentiment has been shifting — 18 analysts revised EPS upward recently, and a cluster of firms (Tigress $255, Needham $250, Oppenheimer $235) raised PTs on the back of AI product momentum and the 40% headcount reduction showing cost discipline. Stock’s already up 61% in the past year, so the easy money is gone — but the r/r still favors the long side if product attach rates accelerate.
Verdict: Monness finally steps in with a Buy upgrade at $155, calling the valuation compelling after a 41% YTD drubbing. The stock is 58% off its late-2024 highs and trades at 18.1x P/E with a 0.47 PEG — cheap for a name with 77.6% gross margins and a 12% FCF yield. The upgrade is a valuation call, not a catalyst call.
"Salesforce has fallen 58% from its all-time high... the stock now trades just 1% above its 52-week low."
Monness cites margin profile, cash flow, and the buyback as supports. Also nods to the agentic enterprise pivot. The Fin acquisition (Intercom, ~$3.6B all-cash) is a separate positive — Jefferies, Canaccord, Stifel, Wolfe all maintained Buy ratings with PTs of $220-250. UBS sits Neutral at $185, calling the AI spend uncertain.
The setup is classic deep value: 18x earnings, 0.47 PEG, 12% FCF yield, and a stock that’s been left for dead. The question is whether AI disruption fears have permanently re-rated the software multiple or this is just a cyclical trough. Monness is betting on the latter — but they’re early, as usual. Worth watching for a bottom-fishing entry if the macro cooperates.
Still a show-me story. Mizuho lifts PT to $135 from $128, but keeps Neutral — not exactly a barn-burner endorsement. Stock sits at $133.99, basically kissing the new target. The thesis hinges on advanced packaging (EMIB-T, glass substrates) as a real moat versus TSMC, but yields remain the bogey.
"EMIB-T to be cheaper but notes the need for improving yields to 99%."
Glass substrate and 3D stacking = the next leg? Mizuho’s expert call flags EMIB-T and CoWoS-L as the current winners, with glass substrates bringing better thermal/density characteristics. Equipment makers (LRCX, MKSI, AMAT) get a nod as incremental beneficiaries. But Intel’s path to capturing that value is still fuzzy — 536% return over the past year already prices in a lot of foundry hopium. With 18A-P now in risk production and talks with Google/Nvidia for backup foundry work, the narrative has more legs, but r/r at $134 on a $135 PT? PMs should wait for a better entry or a catalyst that actually moves the yield curve.
Verdict: Analysts are circling wagons — higher PTs from Truist ($300) and Benchmark ($290) post-Summit, UBS holds $370 with a Buy. The real debate isn't Snowflake's growth (31% LTM revenue, mid-high 30% forward), it's how AI shows up in the P&L. Databricks disclosed pass-through AI/model resell rev accelerating to ~80% growth. Snowflake confirmed to UBS those revenues are immaterial. The firm reads that as a net positive — Snowflake isn't diluting margin or identity with low-margin pass-through. The AI lift is still early and will come from native workloads, not reselling compute.
"The firm views Snowflake's multiples of 14 times and 26x calendar year 2026 estimated revenues and 59 times FCF as defensible given mid-high 30% growth."
UBS's $370 PT implies 17.5x CY27 EV/Sales. At $231, that's a ~60% implied upside. The bull case rests on AI being a pure platform adoption catalyst (Data Clean Rooms, Cortex, Snowpark) rather than a revenue share model. Bear risk: if Databricks's data warehouse $1.5B run rate starts eating Snowflake's core (still TBD — Databricks IPO delayed, no public disclosures). For now, consensus is a 1.49 Buy — but the narrative is tightening around how AI monetizes, not if.
Verdict: The story just got a lot bigger. Needham rips its PT to $400 (from $350) after attending Vicor’s annual meeting — but the headline isn’t the number, it’s the long-term model rewrite. New targets: $2.5B revenue (up from $1.0B prior), 70% gross margin, 40% operating income. That’s a 5x revenue leap from current $427M. The stock has already delivered a 652% return over the past year — but management is telling you the growth runway just expanded dramatically.
The catalyst pipeline is real. Vicor will sample an improved second-generation VPD solution to Cerebras in Q1’27. It may also have secured second-gen VPD design wins with existing licensees that ramp in late 2027. Licensing revenue is also flagged as having upside to the $200-300M annual target. And they’re making progress on a second ChiP fab — capacity expansion is the key constraint here.
“The company raised its long-term financial model, now targeting revenue of $2.5 billion compared to a prior target of $1.0 billion, gross margin of 70%, and operating income of 40%.”
The bull case is an AI power inflection. Vicor is the highest-performance power delivery solution for the most advanced ASICs (Cerebras, others). The long-term model implies a massive scale-up in both product and licensing. The PT of $400 is 50x Needham’s 2028 EPS estimate of $8.00 — so you’re paying for a 2028 story today. But with a $15.1B market cap and a $2.5B revenue target, that’s only ~6x sales on the LT model. Not cheap, but not insane if they execute.
The bear case is always the same with VICR: it’s a one-customer story (Cerebras) with a premium valuation and lumpy licensing revenue. The new VPD design wins are still “potential” and second-gen sampling doesn’t start until next year. If Cerebras slows or shifts architecture, the entire long-term model crumbles. Also, 50x 2028 EPS — that’s a lot of patience required in a multi-manager book that hates waiting for the third out.
The rate of change is undeniable. This wasn’t a modest PT bump. It was a fundamental re-rating of the addressable market. Needham went from $260 in May to $350 in June to $400 now — momentum in analyst thinking is clear. For momentum PMs, this is a signal that the narrative is still accelerating. For value PMs, it’s priced for perfection. Pick your poison.
FERC just handed BE a fresh catalyst — faster grid hookups for data centers. The utility commission agreed to let big energy users (read: hyperscalers) bypass interconnection bottlenecks if they pay for upgrades. That’s an accelerant for on-site generation like BE’s fuel cells, which skip the grid waiting room entirely. Stock already ripped to $341.94, blowing past UBS’s $322 PT — but the regulatory direction is still a tailwind, not a sell signal.
“The key goal is to shorten the timeline and remove bottlenecks in grid interconnection for very large loads.”
UBS’s Manav Gupta kept Buy on the news. The broader setup: DOE pushing to outpace China on AI power demand, and BE sits right in that crosshairs. Other analysts are mixed — Bernstein just initiated at Market Perform ($276), BMO Outperform ($279) — but the rate of change on the regulatory side is what matters here. Stock’s had a 1,412% run. That’s repriced a lot of good news, but the narrative just got another gear.
The data center thesis is getting realer. Modine’s not just riding the wave — they’re locking in contractual floors. Davidson reiterated Buy / $330 after a call with CEO & CFO, and the key takeaway is a long-term agreement with ONE hyperscaler that provides a $4B MINIMUM CHILLER DEMAND THROUGH 2029. That’s a revenue visibility floor for almost a decade, not a hope-and-pray TAM.
"Modine’s data center visibility now extends through the decade, supported by a recent long-term agreement providing a floor of $4 billion in chiller demand for one hyperscaler customer through 2029."
Stock’s up 207% over the past year — you’re late if you think this is a discovery moment. But the rate of change in the backlog narrative is accelerating. The Q4 beat (EPS $1.71 vs $1.57, rev $954M vs $921M) already triggered a wave of PT bumps: UBS to $355, GLJ to $428, Davidson to $330 (from $265). That’s a $330-428 cluster now, meaning the stock at $291 still has headroom if they execute.
Bull case: $4B floor is just one hyperscaler. Multiple pathways to grow beyond that. Supplier risk is being managed — Davidson notes “multiple pathways to protect the business.” Inorganic optionality post-spin-off of Performance Technologies is another lever.
Bear caveat: 28% revenue growth baked into FY27 forecasts. Execution risk on supply chain and hyperscaler concentration. But the floor changes the r/r — you’re not betting on upside alone anymore, you’re betting on a guided path. PMs who missed the first 207% can still own the compounder phase.
Verdict: Stifel's PT bump to $330 (from $280) is the headline, but the real signal is the margin inflection thesis gaining traction. This isn't just another EMS name riding the cycle — PLXS has genuine operating leverage beyond 6% OPM, and mgmt's recent investor meetings confirmed the path is near-term achievable. The stock is at $300, already up 129% in the past year, so the easy money is gone. What matters now is whether the data center power angle and the $6B revenue capacity give the next leg.
"Stifel identified data center power as a potential longer-term revenue driver for the electronics manufacturing services provider. The company’s engineering-focused approach creates customer stickiness and competitive advantages in winning new business."
Bottom line: PLXS is execution-dependent from here. Needham also chimed in — $310 PT, Buy — reinforcing the same narrative. At 10% gross margins, the operating margin expansion story is credible but not discounted. If mgmt can deliver >6% OPM in the next 2-3 quarters, the stock re-rates further. If not, the 129% run already prices in a lot of perfection.
BMO goes to C$68 (from C$53) after the Blue Canyon acquisition — and that 103% YTD rally still hasn't fully repriced the opportunity set. The $620M all-cash deal opens up a much larger TAM in US gov/defense, supplementing an already robust pipeline in Canadian/international defense and commercial LEO. Stock's still cheap on relative basis given the breadth — P/E of 59x is optically high, but you're paying for a space-adjacent compounder with real moats.
"Acquisition should significantly expand MDA’s total addressable market by opening the door to a broader slate of U.S. government and defense opportunities."
Q1 was a beat (revenue +32% YoY, EPS CAD0.38) and the new Montreal facility doubles their manufacturing floor space — they are scaling aggressively. 49North contract renewals (up to $43M through 2031) add further visible revenue. The r/r still looks attractive even after the run, assuming Blue Canyon integration goes smoothly.
Verdict: GLW is the purest optical AI infrastructure play in the TMT universe, but at 96x trailing P/E (46x 2027, 36x 2028) the tape is pricing in a lot. Multiple firms raised PTs overnight — Truist to $205 (Hold), UBS to $223 (Buy), Wolfe to $230 (Buy) — but the collective view is that the easy money has been made (+291% over the past year) and the next leg needs delivery on the Springboard plan ($20B sales by YE26, $35B by 2030). Hold-rated Truist sees limited absolute upside vs consensus; the bull case is that the NVIDIA optical connectivity partnership and Solar growth drive 27% EPS CAGR through 2030, which justifies the multiple if execution holds.
"The Optical segment is integral for artificial intelligence infrastructure builds."
Truist’s line sums it up. The 17% sales CAGR from Optical/Solar (vs 15% prior) is the accelerant, and the three new U.S. manufacturing facilities with NVIDIA keep the domestic sourcing narrative alive. But at these levels, the risk/reward is fine — not screaming. PMs already own a lot of this name. Watch the print for the optical backlog inflection.
TD Cowen comes out swinging — PT to $70 from $55, Buy reiterated. The call is simple: destocking is done, secular growth in auto and AI data centers is real, and the market still under-appreciates Allegro's moat in magnetic sensing and power. Stock's already UP 124% YTD, but the firm sees more room — calls it a top SMID pick with favorable r/r.
"Allegro’s leadership position in magnetic sensing and power technologies positions it for profitable mid-teens growth over the long term."
Caveat: the Q4 beat was met with a sell-off (investor concerns elsewhere), so the tape isn't falling in love yet. That could be a setup if TD Cowen is right about the narrative shift.
Verdict: Evercore pulls the ripcord on valuation. Downgrade to In Line from Outperform, PT slashed to $230 from $260. They're not bearish on the story — they're saying the risk/reward is now balanced.
TEL trades at ~17.3x forward P/E (vs 5-year avg of ~18x). That's not a screaming value, but it's not cheap either given near-term uncertainty. The key tension: AI data center is real ($1.4B in 2025, targeting >$3B), but it's only 12-13% of revenue and has gone "flat for a couple quarters." The math on the ramp from $480M March quarter to $700M+ per quarter exiting FY26 is the swing factor.
"Current valuation… adequately reflects long-term potential against near-term uncertainty."
That's the whole thesis in one line. Evercore's $230 PT = 18x on ~$12.60 FY27 EPS. That's roughly where the stock sits today. No catalyst to push multiple higher until AI re-accelerates or the auto cycle turns. Richards acquisition is running ahead of plan (high-30s% EBITDA margins on $400M sales), and North American utility capex is a multi-year tailwind, but those are already in the price.
Bottom line: TEL is a good business at a fair price. Not a short, not a long — a park-it-until-you-see-the-next-leg-higher name. PMs should wait for a better entry or a clearer AI reacceleration signal.
Verdict: BofA isn't sweating China's export control list. They see this as a strategic endorsement, not a headwind. Lawson Winder reiterated Buy and $85 PT (stock at $61.14, +61% YoY). Thesis: MP has already stripped out Chinese dependencies from its own supply chain. The listing actually reinforces the company's role as a national security asset.
"BofA Securities views the export control listing as reinforcing MP Materials’ strategic value as a U.S. national asset."
Winder calls MP the "only vertically integrated rare earths magnet manufacturer outside China" — that's the whole bull case in a sentence. No impact expected from the new restrictions (MP doesn't rely on Chinese inputs). DA Davidson also in the bull camp with an $82 PT, but the BofA note is the clean read: this isn't a risk event, it's a positioning event.
FROG KEEPS PRINTING. TD Cowen bumps PT to $100 from $80, adds to Best SmidCap Ideas list. Stock at $79.54 — up 96% in the last year, 30% YTD. The thesis is clean: coding agents are exploding the number of software packages, and JFrog is the supply chain layer that manages them. That’s not a one-quarter trade.
"Coding agents are causing a significant increase in the amount of software packages."
The durability call matters. 77% gross margin, 25% revenue growth, cloud estimates have upside, and the security side (Curation product) is gaining traction. TD Cowen thinks the premium multiple sticks as long as growth stays above 20%. Cantor and Stifel also chimed in recently — $80 and $85 PTs, respectively — so the street is converging higher, not just one house.
The Anthropic Claude Code plugin launched this month. That’s a concrete AI-native use case: developers scanning artifacts inside the coding agent loop. Inclusion in the Russell 3000 later this month adds index demand. Nothing to dislike at the current r/r if you buy the AI workflow thesis.
Bottom line: FROG is the pick-and-shovel for AI-generated code bloat. Growth durability is the debate — so far the numbers back it up.
Verdict: Truist stays Buy at $440 despite the stock getting absolutely destroyed — down 51% in the past year and trading just above its 52-week low. The bull case rests on a raised long-term SaaS growth target (~20%) and transaction revenue upside. Valuation is the hook: below 20x FCF (inclusive of SBC) for a company with a perfect Piotroski Score. Not exactly screaming value, but the r/r flips if execution holds.
Street consensus is broadly constructive — Citizens ($500), DA Davidson ($460), and Stifel ($400) all stay Buy. The real question: can management deliver on that 20% SaaS growth when macro headwinds are flattening state/local budgets? AI in permitting and application review is a new angle, but early innings.
No blockquote worth pulling — the Truist note is mostly reaffirmation after investor meetings. The $6B liquidity pile and $150M buyback (part of $1B authorization) are incremental positives, but the stock needs proof of re-acceleration, not just guidance tweaks. Monitor transaction revenue prints for the real leading indicator.
MSFT – THE CHEVRON DEAL IS A STRUCTURAL SHIFT. 2.67 GW behind-the-meter, zero grid draw, 20-year PPA. Bypassing ERCOT entirely. This is the new blueprint for hyperscaler power. Stock at 20x fwd earnings, cheapest since 2016 — either a value trap or the best setup in tech. The market is ignoring the physical track advantage while obsessing over Alphabet’s talent exits. (Depreciation time bomb narrative is unproven.)
NVDA – RENTAL INDEX ROLLING OVER. H100 spot softening, but GB300 premium holds. The 2.5x serving cost drop on GB200 in 70 days via software is expanding effective compute supply faster than hardware can depreciate. If scarcity compresses, the capex thesis gets challenged. Still, NVDA is converting inference memory into a moat via BlueField-4. CUDA efficiency loop is real — but so is the ASIC commoditization threat.
AMZN – DOWN 4.5% ON NO COMPANY NEWS — pure hyperscaler contagion. The Maverick Capital “air pocket” risk is live: can AWS revenue keep pace with capex? Expect Amazon to follow MSFT with direct E&P PPAs (Exxon, Coterra are likely). Custom Trainium/Inferentia keeps silicon costs low, but no software moat means they’re a consumer of compute, not a creator of scarcity. Long-term, that’s fine — but near-term sentiment is brutal.
META – -14% YTD, no specific catalyst. The market is pricing in AI ROI skepticism. Meta’s open-source (Llama) strategy means they benefit from model commoditization — a structural advantage if the narrative flips. The $900M-$1B spend on CRED/Kunal Shah signals free cash flow deployment outside AI. Not a bad thing, but adds noise to the margin story.
ORCL – CUT 21,000 JOBS (13% OF WORKFORCE) in 12 months, explicitly citing AI. This is the first major enterprise software company to show real AI headcount replacement. Margin expansion thesis is real — if FY27 margins beat, ORCL is the canary for the whole sector. OCI is niche but the cost structure is getting right.
SMCI – UP TODAY ON COLOSSUS 2 ORDERS (3,100 racks). Upgraded to buy. But the beta to NVDA is extreme — any GPU demand miss hits SMCI disproportionately. The SpaceX 90-day out clauses on compute deals show uncertainty in long-term demand. Short-term order flow is positive, but the stock remains a high-beta proxy.
AVGO – LOST GOOGLE TPU V9 TO MEDIATEK on 448G vs 336G SerDes. This is a signal that ASIC competition is shifting from compute to interconnect. Broadcom’s moat is narrower than investors think. But the AMD employee interview suggests hyperscalers will come back for photonics and chip complexity. 2028 is the realistic scale timeline. Near-term loss hurts, but relationship intact.