Good morning.
Semi tape is bid pre-open — ASML +4% on the INTC raise read-through, cap equipment generally stronger. NVDA pushing higher on the $500B financing vehicle news — CDO-like structure says they've maxed out direct balance-sheet exposure, but the credit moat gets wider anyway.
INTC is the real story: $20B UPSIZED AT $95, DREW $100B+ DEMAND. Oversubscribed foundry war chest (Ohio fabs, Terafab optionality). Simple read-through — INTC spends, ASML/KLAC collect.
LITE flagged InP substrate tightness on the call, chuckled about needing AXTI's help. Optics bottleneck shifts from lasers to substrates.
Memory tight through 2027. SK Hynix's $26.5B ADR debut, Kioxia at 40-50% fulfillment. HBM downspec unlocks 30-50% MORE RUBIN UNITS from the same bit pool — 4-7GW of incremental power demand, and power is the binding constraint, not just memory.
MSFT unveils Maia 300 in September, GOOG's TPUv8i already in bring-up — hyperscaler silicon real, still a fraction of TPU scale. Neoclouds running hot too: CRWV doesn't own the power or the DCs it leases — survival is the bar.
We'll hit NVDA, INTC, MU first, then get to LITE and the semi cap group.
Five shops raised targets. Not one upgraded the rating. That's the setup in one line — RPD cleared a bar buried six feet under, and the best the street could do was "Hold at a higher number." Stock's +57% over six months on a margin story while ARR bleeds. Makes sense if you frame this as self-help, not growth comeback. The market already got the memo.
PTs from $9 to $13, and every single one got bumped post-Q2. DA Davidson is the loneliest bear at $9 (Underperform). Stifel tops out at $13 (Hold). The cluster — Scotiabank $10.15 (Sector Perform), RBC $12 (Sector Perform), Piper Sandler $12 (Neutral), Stephens $12 (Equal Weight) — all neutral-equivalent, all moving targets up mechanically off the print.
No upgrades. That matters more than any PT. The analysts followed the numbers, not their conviction.
Bull: The margin math is clean. Cut 12% of headcount, reallocate into the Core Platform (Detection & Response + Exposure Management), hit 20% operating margins by Q4 FY26, and FCF actually grows in 2027 on a lower ARR base. At ~8x CY27 FCF and a ~20% LTM FCF yield, the equity looks cheap if management just executes the cost-out. Q2 delivered on that — EPS printed $0.44 vs a $0.35 bogey, FCF $31.9M, full-year op income and EPS guidance raised. Even the street's biggest bear admits the self-help math:
"We expect ARR to remain in decline for the foreseeable future, but the announced 12% RIF will at least result in improved profitability & FCF, w/ FCF expected to increase in 2027 despite a lower ARR base." — DA Davidson
Bear: Look at the trend, not the quarter. 18 consecutive quarters of ARR growth deceleration. Two straight quarters of YoY ARR declines. Net new ARR was -$8.1M in Q2, and Q3 guidance implies another ~$12M QoQ drop. Core ARR +1%. Non-core still dying. New CEO Wael Mohamed's strategy "takes multiple quarters to materialize," which is code for: no growth catalyst on the horizon. Scotiabank flags the real risk — the turnaround could be a value trap:
Improved margins may be rewarded in the short term, but reduced investment could further affect top-line growth. The firm struggles to determine the turnaround timeline. — Scotiabank
Both cases are true simultaneously. Bull says the margin math works. Bear says the top-line math doesn't. The stock re-rates on which one the market decides to pay for. Right now it's paying for the former.
Incremental: the cost structure is now a weapon, not a headwind. The 12% RIF, the 20% op margin target by Q4 FY26, and the first real strategic framework from the new CEO — that's what moved the stock, not the ARR print. Full-year op income and EPS guidance got raised into the cuts.
Already known: ARR was declining. Nobody got surprised by -2%. The Q3 guide of another ~$12M ARR drop was telegraphed. This is a business in a controlled descent, and the market is fine with it as long as FCF keeps compounding up.
RPD is the cleanest example of a legacy security vendor swapping the growth narrative for the FCF narrative in real time. The playbook applies directly to the rest of the exposure management complex — Tenable, Qualys — where land-and-expand has run its course and the mid-market is saturated. The +57% six-month re-rate with ARR in decline tells you exactly where the buy-side rotates in this segment: multiple expansion on FCF beats multiple compression on growth. Watch whether the next leg of bulls are the same sell-side desks who kept Holds — they'll chase once the ARR trajectory inflects. That's the real trade.
The crypto-miner-to-AI-datacenter pivot is officially real, and the street is re-rating it in real time. 241 MW contracted in seven months — from ZERO. Rockdale is now leased to a frontier lab (reportedly Anthropic — the filing doesn't name the tenant), the full 1GW Corsicana site is under LOI with a single tenant, and the bull case is no longer thesis. It's arithmetic.
Seven firms moved after the print. New PT range: $25 to $40 (from a prior $23-30 cluster), average landing around $31. Everyone is Buy/Overweight — there's no wall of worry here, just a scramble to catch up. HCW is the outlier at $40, and they're the ones flagging "advanced discussions" at Corsicana. That's not a coincidence.
The consensus math is straightforward. $9.1B TCV across 20 years at Rockdale (or $16.1B if both renewal options get exercised), guided 80-90% NOI margin on ~$457M average annual revenue, or ~$365-411M of NOI. That works out to roughly $2.1M NOI per megawatt — datacenter economics, not bitcoin mining economics.
Cantor frames it best: the new lease alone is worth $11/sh. Combined with AMD, contracted value is $13/sh. Toss in ~$1/sh for BTC and net cash. Against the $20.28 print, the market is paying ~$6-7/sh for Corsicana and everything else. Which is the entire trade in one sentence.
Corsicana is the free call option at the heart of this name. 1GW under LOI. Cantor's math: if it executes at similar terms, that's ~$41/sh at 16x EV/NOI. Needham thinks it happens near-term.
Bull: The rate of change is violent. Zero to 241 MW / $9.8B of contracted revenue across two counterparties in seven months — that's credibility, not a press release. Available power with interconnect approval is the scarcest commodity in AI infrastructure, and Riot owns 1.7GW of it across two sites. The AMD deal was the first shoe. This is the second, bigger one. Corsicana at 1GW would roughly quadruple the contracted footprint.
"Riot's 1.7-gigawatt power assets across the two sites are among the best for building AI data centers due to interconnect approval and urban proximity offering low latency." — Bernstein
"Riot is among the few high-performance computing and colocation operators with immediately available power." — Needham
Bear: Let's not confuse the headline with the timeline. First 96 MW lands December 2027. The remaining 95 MW in June 2028. That's 16+ months of execution risk, credit-wrap risk, and AI capex-cycle risk between now and first revenue. Q2 was a miss — EPS -$0.33 vs -$0.23 expected, revenue $153.27M vs $155.59M. And there's a funding gap: $2.1-2.3B of capex against a $573M interim facility. The investment-grade backstop is "expected," not closed. If that slips, the timeline slips.
"Management expects an investment-grade credit wrapper to follow, which would cover the full 20 years of lease payments." — Needham (emphasis on expects.)
Also: 3.85 beta. This thing moves 4% on a sneeze. If BTC rolls over or the IG wrap gets delayed, the drawdown will be violent. And for all the Anthropic speculation — the counterpary is still unidentified in the filing.
New: The 191MW Rockdale lease with dollar figures, NOI margin guidance, and delivery schedule. The $573M non-recourse interim facility at SOFR+275, maturing October 2026. The entire Corsicana campus under LOI with a single tenant — that's a new disclosure, and it's bigger than the Rockdale deal.
Known: The AMD lease (~$700M, $2/sh per Cantor). Q2 miss. The strategic shift toward HPC conversions. ERCOT regulatory overhang had put the whole sector in a holding pattern — this deal cracks that narrative open again.
This is the cleanest validation yet of the "miners are the new power landlords" trade. The 75% run in RIOT over the past year and the 53% YTD will drag the whole complex — watch CORZ, IREN, CIFR, WULF on the thesis that the conversion playbook works beyond the first two names. The Corsicana LOI is the next catalyst to track. If it executes — and Needham's "high probability, near-term" language suggests it's close — the entire comp set gets a fresh ceiling to re-rate against.
The numbers: Q2 EPS $0.08 vs $0.07 est, revenue $29M vs $28.14M est, UP 13% Y/Y. FY26 growth guidance hiked to 13-14% from ~12% (TD Cowen books it at 13-15%, close enough). Gross margin sits at 87% — that's the IP licensing model printing money.
Three voices, one narrative: execution is fine, the multiple is the problem.
TD Cowen frames the quarter's strategic significance best:
"The quarter highlighted continued execution on CEVA's strategy to increase value per design through broader system-level engagements, carrying larger license values, higher royalties, and greater customer stickiness."
Not sure the 12% drop is justified on fundamentals — guide went UP, licensing is cycling at three-year highs, 87% GM is absurd. But momentum is momentum, and the Buy-rating crew is shaving PTs even as they hold the line. The 2-3 quarter test: do these platform wins convert to royalty dollars? If yes, $31 looks silly. If not, Benchmark's patience wins. I lean with the bulls here — the platform narrative is showing up in licensing numbers, and that's the leading indicator for royalties.
TWO BUYS, TWO PRICE TARGETS ($20-22), STOCK AT $8.70. That's ~130-150% implied upside. The bull case isn't about bitcoin — it's about the AI data center pivot. The bear case is the company is still a miner with 0.67% gross margins, burning cash, and needing to finance the pivot with dilution. Both analysts see the same thing. They just think the market is too pessimistic on the transition.
Q2 revenue $228.8M, +47% YoY. Beat Needham's top-line and blew through adjusted EBITDA estimates. EPS LOSS OF -$0.37 vs -$0.32 consensus (wider, but nobody cared). Mining efficiency gains drove COGS down. Revenue +128% over the last twelve months. Stock down 23.6% IN THE PAST WEEK — that's the market pricing in dilution, not operational failure.
Needham cut PT to $20 from $22 on a "less robust pipeline" than previously forecast. Benchmark held $22. Both maintained Buy. Both cite the AI cloud story, not the mining business, as the reason.
The bull narrative: Bitdeer is an AI data center play that happens to mine bitcoin while it builds. Shareholders are pushing management to emphasize the AI side — Benchmark acknowledges the logic, since AI data centers carry contracted cash flows and get higher market multiples.
The problem: ALL of the company's revenue currently comes from mining. Mining pays the bills. The AI side is still capex-heavy infrastructure in progress.
Needham's operational detail:
"The mining business is paying the company's bills."
That's Benchmark, and it cuts to the core tension. The market wants to pay AI data center multiples. The company is currently a low-margin miner that needs to spend billions to become that AI story.
This is the real bear case. Cloud operations carry much higher capex than mining. Needham raised its 2027 adjusted EBITDA slightly on improved mining margins, but also raised the share count estimate and lowered the target multiple. That's the honest way to underwrite this: the pivot works, but current shareholders get diluted along the way.
Cash burn is accelerating. Gross margin is 0.67% — essentially breakeven mining economics. If bitcoin stays weak, there's no internal cash to fund the cloud buildout. It was 43.9% below its 52-week high. The company's top and bottom line are moving in the right direction, but at this run-rate, BTDR will likely need to raise capital.
Bull: AI cloud expansion + mining efficiency = EBITDA inflection. Revenue +128% LTM. Management has utility-scale power assets (Tydal, Rockdale) that are scarce and hard to replicate. Contracted AI cash flows deserve data center multiples, not mining multiples. InvestingPro's own fair value says the stock is undervalued at $8.70.
Bear: The pipeline keeps slipping. Rockdale on hold. Gross margin 0.67%. Cash burn accelerates exactly when capex needs to ramp for cloud. The wider-than-expected EPS loss and dilution from future raises cap the stock until the AI revenue actually shows up. The 23.6% weekly drop says the market has stopped giving credit for promises.
The stock trades like a broken miner, not a build-out story. Both analysts see enough in the AI pipeline to sit at Buy with $20-22 PTs — but the path to that value runs through dilution, execution risk, and bitcoin prices. If you believe the Tydal/Clarington/Tennessee buildout lands, the current price is cheap. If you think Rockdale stays on hold and financing gets messy, $8.70 is just a falling knife. Right now the market is pricing in the bear case. The analyst community is pricing in the bull. But in a tape like this, dilution warnings carry more weight than PTs. This one's a wait-and-watch until the AI revenue line starts showing up as something other than a cost center.
Verdict: The bull case is dead until active clients and GSV stop bleeding. Two firms cut PTs to $8 — RBC from $9, Scotiabank from $10, both Sector Perform. Needham keeps the Buy flag flying at $11 (from $15), but only after slashing 2026/2027 EBITDA estimates 10%/22%. This is a quality-of-earnings story trapped inside a top-line decline.
Google broke the demand faucet, and AI broke the growth narrative. RBC's Brad Erickson flags May and June SEO changes still driving traffic headwinds into Q3 — not stabilizing. Scotiabank goes further, arguing this is no longer just low-end AI substitution but a shrinking Google/SEO referral channel that threatens customer acquisition efficiency across the board. Same conclusion from both: THE OFFSETS ARE REAL BUT TOO SMALL.
The prints back that up:
The company is angling toward a higher-quality, higher-margin marketplace. Connects, talent subscriptions, and ads now = 15% of total revenue. Business Plus grew 174%. AI-related services hit a $330M annual run rate (~10% of GSV). Lifted on track for 25% FY growth. Management also deploying $5-10M incremental H2 marketing into paid channels that are showing traction against the search decline.
Scotiabank isn't dismissing any of it — just the sizing:
"Upwork is becoming a higher-quality marketplace but has not yet proven it can return to sustainable growth."
Bull: The mix shift is working. Higher-margin revenue streams compounding, Business Plus inflecting, AI services now 10% of GSV, and paid marketing channels give management a lever Google took away. At these levels — sub-$10, 50% off YTD — the bar is low for a stabilization trade.
Bear: Every growth initiative combined is still smaller than the core bleed. Clients are leaving, GSV is negative, and visibility is deteriorating. Scotiabank put it plainly: investors won't reward 30%+ EBITDA margins if revenue stays in decline, so multiple expansion stays blocked until GSV and active clients flatten. The Q2 beat-and-guide-down was the tell — management doesn't see the bottom either.
Not a value trap if you think paid channels fully replace SEO by 2027. It's a show-me story at best. Needham's $11 PT vs the $8 Street cluster is the whole debate in two numbers — either you believe the offsets compound fast enough to re-accelerate growth, or you wait for GSV to inflect before touching it. With Google's algorithm changes still rippling and agentic AI eating more of the low-end freelance workflow, the risk skews bearish until proven otherwise.
DA Davidson keeps the Buy at $498, and the note reads like ADI has the cleanest setup in analog heading into H2. Not just a PT reiteration — the thesis is that the recovery is broadening, not hanging on one end market.
Volume drove the FQ2 beat. Channel inventories stay lean. DA sees FQ3 revenue ABOVE its $3.90B estimate (+7.6% QoQ) and FQ4 upside to $3.98B — above-seasonal in a group that's spent two years in the gutter. Demand drivers: ATE, aero/defense, auto content, AI data centers. That's four independent engines firing, not one bounce.
PTs are climbing around him — Argus to $460, Cantor to $550. The cluster matters more than any single target: bulls are adding upside, not cutting. The $498 DA target is just 15x CY27 EV/Sales for the high-end B2B analog franchise. That's the quality premium, and it's not stretched.
"Analog Devices is entering the second half of fiscal 2026 with one of the strongest setups in the analog group."
Bear case: this is a guided recovery, not a confirmed macro turn. But with lean inventories and improving order visibility, the risk skews north. The Empower Semiconductor acquisition closes the loop on AI power delivery — another growth vector the consensus hasn't fully repriced.
Verdict: Contrarian bull case is intact at ~12x EV/FCF, and the Street is refusing to blink into FQ2. Stifel holds Buy/$250 with expectations for in-line to modest upside on revenue (+10% CC) and cRPO (+13% CC). Raymond James still at Strong Buy/$290 citing better-than-expected demand. Erste upgraded to Buy. Three desks, one thesis: the bar is low enough that even a boring print works.
Stock's down 25% YTD vs IGV down 1%. Market's pricing in the bear case — Agentforce is vapor, 2H acceleration won't materialize, Data Cloud 360 is a science project. PEG of 0.58 says otherwise. Stifel's customer and partner checks suggest stable demand and the AI business can compound. The $1.6B VA deal (three-year Agentic Enterprise License on Missionforce) is a real anchor-tenant proof point, not a press release.
"At approximately 12 times enterprise value to free cash flow, Stifel views the valuation as presenting a low bar heading into the second-quarter results."
The whole debate reduces to one question: does 2H subscription and support acceleration actually show up? Stifel's checking the boxes on durability. At this valuation, the risk/reward skews long into the print. CRM needs to clear the low bar, not jump a high one.
Wells Fargo doubles down — Overweight, $315 PT — after NVDA drops a $500B+ AI INFRASTRUCTURE FINANCING PARTNERSHIP with Apollo, BlackRock, Blackstone, Brookfield, Goldman, and KKR. That's not a supplier move. That's NVDA building the capital stack for its own demand. The structure: independent compute financing platforms with long-duration, usage-linked revenue sharing. Translation: NVDA wants a toll booth on every AI Factory, not just the shovel sale.
$5.27T cap. 33.5x P/E. 0.3 PEG. Nobody moves the valuation debate until the model shifts. Wells Fargo's Aaron Rakers flags the obvious — this is NVDA beyond the traditional box-shifter role. The real watch item: does management start guiding toward recurring compute revenue? Rubin racks ship in Q4. Lancium gets $2B for power infrastructure. DeepSeek is building 1GW in Inner Mongolia. Demand signals keep stacking.
"The partnership mobilizes over $500 billion of third-party capital for AI infrastructure buildouts."
The financing structure is the story. Long-duration, usage-linked revenue sharing means NVDA captures upside from AI compute utilization, not just the initial GPU sale. That's the kind of model shift that rewrites the multiple — IF it shows up in the P&L. For now, it's a conviction add to a 33x growth story that still seems cheap on a PEG basis.
BofA hikes PT to $75 from $70 but stays Underperform. A PT raise that's really a valuation warning — the rare "growth is real, price is wrong" print. Stock sits at $66.40, mcap $1.54B, +124% YTD.
The quarter: Q2 REVENUE +89.7% YoY TO $33.9M, FY guide up to ~$135M. Pro forma GM 47.2% — 166bps above consensus — and BofA now models FY26 GM ~47% vs the 45-46% the company was guiding. Apollo5 more than doubled; Apollo3/4 grew double digits. DEMAND EXCEEDS SUPPLY across wafers, packaging, substrates, test — and 80-90% of growth is unit-led, not price. The healthy kind of beat.
Here's the rub: BofA raised FY27/FY28 sales estimates by 19%/20% and pulled first non-GAAP profit into Q3 FY27. New PT at 7x CY28 EV/Sales vs 6x peer avg, justified by 32% CY26-28 sales CAGR vs 17% peers. So the analyst is chasing the story with the model but refusing to bless the multiple. That's a big tell.
"Demand now exceeds available supply across wafers, packaging, substrates and test."
The bull case in one line — edge-AI is supply-constrained, not demand-constrained, heading into FY27. But the r/r is the issue. Paying 7x CY28 sales for a company that only turns non-GAAP profitable five quarters out? BofA literally added 20% to the revenue line and still can't get to Buy. Valuation ceiling is low. Growth ceiling is high. PMs get to pick.
Argus trims CEG to $325 from $350, keeps Buy. The cut is rates-driven, not fundamental — higher discount rates hit the multiple on a long-duration cash flow story like nuclear power. Stock at $270.43, so that's still ~20% upside implied.
The AI demand narrative stays intact. Data center power demand is the tailwind, management signaling confidence via aggressive buybacks and dividend hikes. Q2 beat: adjusted EPS $2.55 vs $2.41 est. Revenue missed — $7.5B vs $7.94B est — but full-year guidance went UP anyway on PJM capacity prices and Calpine integration. That's the part that matters.
Argus sees the premium as justified.
"The current premium valuations are reasonable given the company's leading industry position and its growth and profitability outlook."
The rates move trims the target, not the thesis. P/E 25.3 on a raised guide — market still pricing this as a growth story, not a utility. Fine r/r if you believe the AI power supercycle extends; the risk is rates stay higher for longer and the multiple compresses further.
Cantor's not blinking — reiterated Overweight and $40 PT while the stock sits at $19.09. That's >100% implied upside, and the tape agrees: shares +60% YTD, +10% in the last week alone. The Q2 revenue miss ($5.82M vs $8M forecast — a 27.25% shortfall) is the obvious bogey, but nobody owns this name for the current top line.
This is a platform story, and Q2 pushed the platform across the finish line. Definitive Commerce Dept agreements for funding locked. Serra Verde acquisition signed. Blacksburg, SC picked for the second metal-making/magnet facility. Hydrometallurgical demo plant commissioned in Wheat Ridge. Post-quarter: first commercial-grade oxide samples from magnet swarf and the TMRC acquisition closed. That's the mine-to-magnet strategy moving from PowerPoint to production.
"With the platform now largely assembled and funding secured, our focus shifts to execution across the magnet, metals and Round Top segments."
The CEO transition (Humpton → Moraitis, effective Oct 1) is a watch item, not a red flag. Analysts still model 44% revenue growth this year with no profitability in 2026 — so this remains a story stock until the magnet line proves out at scale. The $1.5B cash balance buys the runway to get there. Targets cluster at $30-$45, so the street's broadly constructive; Cantor's just the boldest. Downside cushioned by that cash, upside tied to execution. At $19, the r/r skews favorably for a name this far along in the build-out.
Stock rips 26% in a week to $155 and the street is playing catch-up with a fresh wave of PT hikes. But Benchmark's post-earnings sit-down with CFO Hoffmeister and IR chief Kleinstein adds the real new layer: the GMV engine is now explicitly international-first. That's the signal to watch, not just the headline beat.
"International markets [are] the likely largest source of incremental GMV dollars over the next 12 months, followed by enterprise and point-of-sale."
That sequencing matters. EU and LatAm carry the near-term load — Brazil already compounding — while SE Asia and Africa are the next layer. China and India? Management waving them off as long-term "regional leadership" plays, not near-term numbers. Same-store growth and new merchant adds roughly balanced, so this isn't a land-grab story; the existing base is spending more.
Benchmark holds Buy/$170. JPMorgan leads the bulls at $185 Overweight; Evercore at $175. The neutrals — Cantor and UBS, both $145 — are now UNDER the tape, which tells you the debate shifted from "is growth durable" to "what multiple do you pay for it." Q2: EPS $0.42 vs $0.40 est, rev $3.58B vs $3.45B, and GMV +32% Y/Y — FIVE STRAIGHT QUARTERS ABOVE 30%. Business is working. The question is whether the 26% week just front-ran the next leg.
Neutral and $38.50 PT — this is a merger arb now, not a fundamental long.
D.A. Davidson held the line after RAMP's Aug 5th print: slight revenue beat, profitability BETTER THAN EXPECTED. But no guidance — the pending $2.5B Publicis acquisition has the company in quiet period mode, with closing still on track for YE26. So the analyst work here is mostly modeling for the deal, not the standalone.
The valuation math does the heavy lifting: $38.50 PT = 10x FY27 free cash flow. At 15.6x P/E and a 9% FCF yield, the equity already screens like a bond with a call option attached. That's what a takeout date does to a stock.
LiveRamp launched the LAB program (AI agents for marketing planning and data transformation) around the print. Nice narrative, but nobody's paying up for standalone AI storylines when Publicis holds the keys. D.A. Davidson's model tweaks are modest — the real r/r is in the deal spread and the close timing, not quarterly fundamentals.
Source is thin, so no blockquote here. The takeaway is simple: hold if you want the deal, don't kid yourself about the organic story.
Stifel isn't blinking into the print. Reiterates Buy and $125 PT with the stock at $90.48 — that's ~38% implied upside. They're calling for a top-line beat of ~5pp toward the high end of the historical range and an FY27 guide in the low-20s percentage growth zone. The base business already prints 24% LTM growth ($1.01B revenue), so this is about acceleration, not stabilization.
The real narrative driver is Max. Company DOUBLED Max capacity in FQ2, and Stifel frames their AI assumptions (Max, virtual agents) as conservative — meaning any upside there is pure option value. Management's stated endgame is every customer on Max, which is a multi-year platform re-rate, not a feature launch. Early data supports the hype: Max users seeing call booking rates +500bps and close rates +1,000bps vs peers. Piper Sandler (OW, $115 PT) goes further, flagging Max as a potential 2x subscription revenue uplift.
Truist is in the same camp on multi-year growth drivers. The Street is coalescing around one narrative: Max is the accelerant, and the quarter should prove it.
"The firm views steady uptick in customer and site adoption as a long-term expansion driver that can contribute to sustained 20%-plus growth."
One caveat to log: Iconiq and a director sold ~$6M of Class A recently. Not a red flag at that size, but insider supply into strength is worth tracking.
DA Davidson bumps DSP to $17.50 (Buy), Citizens to $17 (Market Outperform) — both reacting to the same Q2 beat. Contribution ex-TAC $60.2M, +24.5% YoY, 1% above consensus and high end of guidance. CTV is the engine, up NEARLY 50%. Q3 EBITDA guide of $18.5-19.5M clears the $17.8M bogey by a mile — margin flow-through is the real story here.
Bull case writes itself: PEG 0.19, P/E 37, profitable LTM, AI buying platform (Outcomes) scaling into a record enterprise pipeline. Caveat: 43% six-month run has reset the base. This is momentum confirmation, not a fresh entry.
"The firm noted Viant's successful execution against a record enterprise pipeline and its autonomous AI buying platform, Outcomes, as key growth drivers."
Hold stays, PT bumps to $328 from $304 — still BELOW the $337 tape. Not a buy signal, a "pain is priced" signal. TD Cowen models FY27 revenue guidance that MISSES STREET at SUB-10% growth, and explicitly expects cuts to GBS and TurboTax targets.
The real bear hook: DIY tax ARPU compression from pricing reductions. That hits margins, not just the top line. But the bull case is quietly assembling — 80.8% GM, 20.4x P/E, 0.59 PEG, stock DOWN 52% OVER THE PAST YEAR. Shorts are stacked, expectations are negative, and the downgrade wave (Truist, MS, Stifel all cut to Hold/EW recently) already ran its course.
"Uncertainty related to headwinds to its do-it-yourself tax average revenue per user from pricing reductions, balanced against a low bar following ongoing underperformance in shares and positioning skewing short."
That's the key asymmetry. The bad news is an open secret. The question is whether management resets FY27 guide to the whisper numbers — if they undercut even that, this thing gets uglier before the May 2027 recovery path TD Cowen flags. For now, the r/r is neutral. Not a fresh short, not a fresh long. Wait for the print.
One name on the tape and it's a fence-sit. Cantor reiterates Neutral, $10 PT — stock at $8.81, so ~13% upside but zero conviction (no upgrade, no target bump, score 40).
The whole call is a discounted cash-flow parlor trick. Cantor models QUBT taking 5% of the quantum hardware/software/services market by 2035, discounts that back at 10%, and lands at ~5.8x EV on the PV'd 2035 revenue. The kicker: $1.3B cash on the balance sheet against a ~$2.2B implied market cap.
"The firm projects Quantum Computing can capture 5% of the quantum hardware, software, and services market by 2035, which would equate to $375 million in sales in 10 years."
Translation: you're mostly buying the balance sheet and paying a couple turns for the optionality. The underlying business is still unproven — Q2 revenue printed $5.61M (beat vs $5.15M est, +9,074% YoY off a near-zero base), but the adjusted loss of -$0.05 was worse than the -$0.03 bogey. Revenue growth is easy when you start at nothing. Profitability is the tell, and that's still horrid.
Bernstein reframes the story: CPU is the new growth vector — and it is NOT a CoWoS story. PT to $554 from $430, Outperform maintained. Agentic AI workloads pulling CPU demand across AMD, Intel, Apple, Amazon, Google, Arm. CPU REV GOES FROM $15–16B LAST YEAR TO HIGH-$30B BY 2027 — mid-teens of total revenue. EPS +67% THIS YEAR, 31% CAGR 2027-28. The 28x P/E prints fine against that; PEG is 0.53.
The nuance is the capex mix. $64B/$75B/$82B in 2026-28 looks chunky, but most is plain wafer capacity. CPUs don't sit on CoWoS; if anything, Amkor/ASE do the packaging. So this is a unit-growth story, not a fresh packaging-bottleneck bid.
N2 ramping revenue ahead of prior nodes extends the runway. PT is 20x forward earnings (NT$3,300 equivalent).
"CPU and chiplets are also increasing demand for N3, N5 and N7 technologies."
Elsewhere: TSMC-Sony $4.69B image-sensor JV in Japan (Sony controls; TSMC minority), Needham to $530 on AI wafer demand, Fitch holds Taiwan at AA, GDP to 9.4%. Noise, not signal.
Verdict: the market sold the guide-down, not the beat. TLS fell 17% on a quarter that cleared top AND bottom line — that's the tell. Investors see the same trade Kessinger sees: the company is trading ~$30M of 2027 revenue for 600bps of gross margin, and nobody pays up today for margin that prints in 2027.
Q2 REVENUE $47.7M (+33% YoY, above the $44-46M guide) and ADJUSTED EBITDA $6.9M, both clearing. Company raised FY26 EBITDA by slightly more than the Q2 beat — that's the good stuff. But FY26 revenue guidance came down ~$2.5M, and the softer Q3 guide spooked the tape. LTM revenue is $193.7M at 42% gross margin, growing 66%... and still guiding down. Growth that guides down is growth worth distrusting.
DMDC program phase-out starts in Q4: Telos stops passing through single-digit gross margin third-party software revenue. Costs ~$30M of 2027 revenue but adds 600bps to CASH GROSS MARGIN on run-rate. The TSA PreCheck renewal then adds another 400bps in H2 2027. This is a proper margin-quality trade — the problem is timing. Revenue steps down first, the margin recast shows up later. Between now and then, you carry a growth story that's telling you it won't grow.
Kessinger holds Neutral/$4. Single coverage, no cross-firm spread to consolidate. Stock at $4.23 and down 17% on the week — the tape is the real rating. Wait for the margin expansion to print in reported results before stepping in front.
Verdict first: D.A. Davidson hosted Leighton and McGowan Friday and came away MORE confident on the growth reacceleration — reiterated Buy and $185 PT (~32% upside from $118.39). This is a "keep riding" call, not a fresh one: stock's ALREADY +67% OVER THE PAST YEAR.
The whole debate is margins. Davidson says CY2027 EBIT margins may compress MORE than management guides — but flags 2027 as the bottom. HSBC took the other side, cutting to Hold after Q2 non-GAAP op margin printed 24.62%, below both their estimate and consensus. Non-GAAP EPS -8.1% YoY. At 42.92x P/E, you're paying for a reacceleration that hasn't printed yet.
The bull case has real substance though: cloud infra services +39% in Q2, a $600M four-year robotics deal with a US tech company, and insulation from rising component costs. The question isn't whether growth is coming — it's whether you can stomach the margin bleed until CY2027.
Rosenblatt stays Buy at $53 after the Utah DOT expansion — and this one's more about the TAM narrative than the quarter. The bigger picture: 160 incremental intersections is ~$1.6-3.2M at $10-20K a pop (small), but Utah has ~2,100 signalized intersections total → >$20M statewide TAM. BlueCity now in TN, GA, NJ, CT.
The math gets interesting when you zoom out. ~300,000 signalized intersections in the US → roughly a $3B addressable market. That's the bull case: every state DOT award is another referenceable lighthouse, and the install base compounds.
QUARTER WAS MIXED. REVENUE +56% YOY TO $55M, beat the $51.55M bogey — 14TH STRAIGHT QUARTER OF PRODUCT REVENUE GROWTH. But the adjusted loss of -$0.27 vs -$0.12 expected was horrid, and that's the tension PMs should care about: growth is real, profitability isn't. Stock's up 51% over 12 months and trades above the platform's calced fair value — multiple expansion is doing heavy lifting here.
Bear case: the $3B national TAM assumes broad DOT adoption that hasn't happened yet. One state — even a good one — doesn't validate the whole thesis. Not sure we can read too much into a single contract expansion, but the direction of travel is right. Watch the margin trajectory next print; that's the swing factor.
Stifel flips back to $225 from $200, doubling down on the FQ4 ARR beat story. The call hinges on >$10M OF ARR UPSIDE in Guidewire's seasonally strong Q4 — roughly 1pp over Street — after FQ3 missed by a hair ($26M net new ARR, within guidance, so noise). They also expect the initial FY27 guide to land ABOVE STREET'S ~18% ARR GROWTH with revenue above ~16% consensus. Channel and survey work shows sustained cloud demand and, the new angle, real P&C carrier interest in AI/agentic functionality — Guidewire as the gateway via GWCP plus PricingCenter, UnderwritingCenter, ProNavigator. Germania Mutual running ProNavigator on underwriting and claims is the cross-sell proof point.
Bull case: 94.7X P/E stops mattering when PEG is 0.26 and ARR compounds ~19%. Bear case: rich multiple into an 18% guide is 2023 all over again, and Wells Fargo's $190 UW is the margin-of-safety tell. Stock UP ~8% ON THE WEEK with 11 upward revisions — momentum sides with Stifel. Still DOWN 14% YTD, though. This is a "prove it on the FY27 guide" name, not a derisked compounder.
INTC — BULLISH. The $20B upsized equity raise at $95 drew more than $100B in demand. That's a war chest for the foundry buildout, and Fab 38 activation means every Intel fab is now filling — a utilization inflection, not a capex story. EMIB-T packages Google TPU v9 starting late 2027. Real foundry demand, not just a promise.
ASML — BULLISH. Up 4% on the INTC raise — semi cap reading higher foundry capex. Quietly overtaking KLAC in YTD performance. Samsung pushes High-NA volume to ~2030, so the EUV monopoly gets paid later. But it gets paid.
KLAC — NEUTRAL. ASML overtaking KLAC in YTD says the marginal dollar wants EUV and logic capex over process control. KLAC's not broken, just not the chosen expression of the AI capex trade.
LITE — BULLISH. Q4 revenue $1.01B, up 109% YoY. Q1 guide $1.225-1.275B vs $1.16B CONSENSUS — demand accelerating, not decelerating. Ignore the GAAP loss; that's a $7.8B ONE-TIME NON-CASH DEBT EXTINGUISHMENT CHARGE. InP substrate is the emerging bottleneck, and LITE basically said it needs more AXTI help.
AXTI — BULLISH. The bottleneck supplier. LITE explicitly flagged InP substrate as gating the datacom optics ramp. That's pricing leverage. Watch for capacity expansion announcements.
DELL — BEARISH. Hearing SPCX and CRWV buy AI servers direct from Taiwan ODMs instead of 100% from DELL. The biggest neoclouds are disintermediating the server OEM. Negative for AI server margin and mix; DELL keeps enterprise but loses the highest-growth AI capacity buyers.
AAOI — BEARISH. Commodity product, mixed management track record, and TSMC subsidiary Visera entering 1.6T transceivers. CPO ELSFP capacity at 400K units/month in 2028, ~$400 ASP, ≈$1.92B annualized revenue capacity on top of a $5.6B run rate. That's a supply wave, not a blessing.
MPWR — BULLISH. Initiated at $1,706, 35x 2027E. Enterprise data sales +153% YOY in 2026E, +81% in 2027E. Initial high-speed DDR5 orders secured, 800V AC-DC sampling. Power delivery is the quiet AI bottleneck and MPWR owns the torque.
NVTS — BEARISH. Renesas suing the company; CEO and about half the workforce are Renesas alumni. Legal and governance overhang muddy the GaN/SiC story. Narrative problem more than a fundamentals break.
AMD / AVGO — NEUTRAL. Schwab clients sold both while buying MU and INTC. Retail rotation into memory and laggards. Contrarian signal at extremes; no fundamental change in either story.
SONY — BULLISH. ~¥1T ($6.4B) JV with TSMC in Kumamoto for next-gen image sensors by 2029, primarily Apple iPhone plus physical AI. Sony ~60%, TSMC ~40%. Secures advanced supply and moves Sony into AI-perception hardware.
MU — BULLISH. Gen6 enterprise SSDs supply-constrained; 245TB high-capacity demand very high. KIOXIA CAN ONLY FULFILL 40-50% OF DEMAND. Strategic Customer Agreements are binding take-or-pay through calendar 2030 with upfront cash — a structurally de-risked cycle. NVDA's HBM4 downspec isn't a negative; UBS says it RAISES 2027 HBM consumption from 58.7B to 61.5B Gb. Pricing power flows to MU.
SKHNY — BULLISH. $26.5B US ADR debut. Resumed Dalian Fab 2 after four years, +50K wpm. Market expects FY27 HBM contract pricing up >50%; JPM more conservative at <40%. Memory stays tighter for longer through 2027.
SNDK — BULLISH. Enterprise SSD squeeze is direct NAND pricing leverage — prices nearly 10x year-ago levels. Memory strength isn't just HBM; SSDs are the next squeeze.
AAPL — BEARISH. Reportedly cutting shipments on DRAM shortage. CXMT LPDDR5X too limited for near-term adoption. Memory allocation favors AI customers. Input-cost and supply-allocation problem, not demand.
MSFT — BULLISH. Maia 300 unveiling in September; 300K+ units in 2027, capacity for 1M+. In talks with Anthropic as a potential customer — that flips MSFT from self-supply to merchant silicon. Changes the NVDA negotiation dynamic either way.
GOOG — BULLISH. GCP earnings approaching Search parity by 2028. TPUv8i already in software bring-up on g3 codebase. TPUv9 on Intel EMIB-T from late 2027 — packaging diversification beyond TSMC. Bear case is execution culture, not technology.
META — BULLISH. Muse Glimmer 30B runs on a single consumer GPU — top-3 model optionality. Confirmed first hyperscaler customer for Pure's Everpure in SEC filings. Building AI infra like an owner, not a renter.
AMZN — BULLISH. Alchip July sales NT$7.43B, +181.8% YOY, more than double June. Trainium3 ramp on TSMC N3P confirmed. Real silicon volume, not roadmap noise.
ORCL — BEARISH NEAR TERM. $20B ATM adds to the AI-infra equity issuance wave. Market sells first, asks later on spending opacity. Structurally, ORCL drifts toward Neo economics — higher growth, lower-quality margins, more CoreWeave-style competition. 52% probability it's the second Everpure customer, which would be a quiet proof point.
CRWV — BULLISH WITH DEBT CAVEAT. Q2 revenue $2.58B, ~2x YoY. Backlog $99.4B, up 50% QoQ. Managed inference went from $1M to $100M in one quarter and 2026 capacity is sold out — the highest-quality signal in the print. Funding cost dropped from ~11% to 6% with an A3 rating, but that rating came in a shortage. The real test comes in a demand air pocket.
SMCI — BULLISH. Guided Q1 FY27 to $14.5-15.5B vs $11.99B est; FY27 $65-72B vs $54.4B Street. 20-30% ABOVE CONSENSUS — mgmt doesn't guide that hard without visibility. Q4 profit $1.17B vs $195M a year ago. Taiwan ODM July revenue +64% YoY confirms the read. Accounting-distrust discount unwinds.
NBIS — BULLISH. Hyperscaler contracts reportedly worth more than the entire market cap. Differentiation is real — stringing dedicated InfiniBand between separate smaller clusters. Purest public neocloud bet, but the market oscillates between bullishness and credit risk.
CORZ — BULLISH. Leveraged downstream derivative on CRWV survival. If CRWV merely survives, CORZ +50%. It's a call option on neocloud capital deployment, not a moat.
P — BULLISH. Everpure's first hyperscaler design win is META, confirmed in SEC filings. OCI the most likely second at 52% probability. Customer-procured NAND model: 75-85% hyperscaler gross margins, revenue-light. If OCI signs, repeatability is proven.
NOW — BULLISH. AICT pipeline broadened materially over the past six weeks. AI governance and security are now operational priorities, not debates. Armis moving directly into agent discovery. NOW becomes the workflow governance and remediation layer around IAM. This is a buying cycle, not a narrative.
SAP — BEARISH. >75% of growth comes from expansion, upsell, and pricing — exactly the levers AI compresses. Toll-road thesis ignores cannibalization risk; charging for AI as a hostage fee invites a Digital Access-style revolt. 2% organic grower risk masquerading as 8% LFL.
NET — NEUTRAL. Up 50%+ YTD at 22x 2-year forward revenue, yet investors still argue SaaSapocalypse. Market pays up for platform and agentic winners. Edge relevance to AI workloads is real; the multiple is rich.
SNOW — NEUTRAL. Up 50%+ at 12x 2-year forward. No longer priced as a laggard — data cloud plus AI optionality. Crowded consensus or durable platform shift? That question is the trade.
MNDY — BEARISH. Down 9% after weak Q2. Work management feels budget pressure from AI reprioritization. Reps still see legacy competitors, not AI-native entrants — disruption early, but the risk is budget shifting to agents.
ASAN — BEARISH. Reps name ASAN a top legacy competitor, but nobody frames it as an AI-native threat. Caught in the middle if work-management budgets shift to agents.
HUBS — NEUTRAL. In competitor lists, not a primary target. AI-native disruption hasn't landed. Watch whether HubSpot becomes a threat or a victim of agentic marketing workflows.
TEAM — NEUTRAL. monday reps cite Atlassian as a competitor. Developer/ITSM moat is more defensible than pure work management. No new signal.
PLTR — NEUTRAL TO BULLISH. Burry's sub-$1 target looks disconnected from fundamentals — net cash ~$3/share and Q2 EPS $0.41. Hard floor exists. Multiple still needs growth to justify it.
PANW — NEUTRAL. Most referenced incumbent in the security competitive set — ZS, MSFT, CSCO, CRWD, FTNT all position against them. Multiples compressing. Leadership real, target on the back growing.
CRWD — NEUTRAL. PANW and others position against CRWD. No new print. Endpoint stays sticky; open question is whether AI-native agents reduce EDR need.
ZS — NEUTRAL. Competitor reference for PANW's GTM. SASE/zero trust still solid. Platform-vendor intensity rising; multiple compression is the main risk.
CSCO — NEUTRAL. Appears in security competitive maps, not a primary leader. Networking plus security bundling; AI networking a potential tailwind. No new data.
FTNT — NEUTRAL. Price-performance value prop remains a recurring threat to premium platforms. No new fundamental print.
BX / APO / KKR / BAM / BLK / GS — BULLISH. All trading well on the NVDA $500B data center MOU. Private credit is the marginal dollar funding AI infrastructure. The collateral is the compute — the A3-rated CRWV facility proves it. GS explicitly frames this as a market for credit backed by NVDA compute.
CME — BULLISH. First compute futures for H100 and B200 launch OCTOBER 5, cash-settled on Silicon Data benchmarks from ~100 rental platforms across 40+ countries. Financialization of GPUs — price discovery and hedging for the AI asset class. Legitimizes the whole NVDA financing thesis.
MARA / CLSK / HUT — BULLISH OPTIONALITY. Miner power capacity is in the next AI-deal wave after RIOT-Anthropic. Power is the binding constraint; miners with grid access, PPAs, and cooling convert to AI compute. Optionality, not contracted revenue yet.
BABA — BULLISH. Cloud claims 100-day large AI data center buildout, ~10% cost cut via CUBE 5.0. Cost and velocity advantage. China builds aggressively despite export controls.
VZ — NEUTRAL. GenAI ROI of 5-10x reported. Good enterprise adoption datapoint; not a needle-mover.
HRB — NEUTRAL. FY26 revenue $3.95B +4.9%, EPS $5.69 +28.7%. Solid execution, no AI narrative. Residual data point.
BWEN — BULLISH. 1.5x book-to-bill, revenue +67% YoY, $17M cash vs $2.5M debt. Small-cap infra momentum with a clean balance sheet.