Substack 周末综述转述 Oracle FQ1'27（OCI +120.7%、RPO $664B）与 Adobe FQ3 财报及一批卖方评级观点，但两项真催化已于 9/11 完成定价并反转

Happy Friday! **QQQs +60bps** as **Oil -3%** pulls back overnight and yields dn 1-2bps across the curve. The drop in Oil attributed to an[ FT article](https://www.ft.com/content/df2d9bcc-59f9-4122-aa4d-2020bef77e0e?syn-25a6b1a6=1) about a mtg in Oman on Monday between GCC ministers and Iranian counterparts with objected to reach a deal on managing traffic in Hormuz. Core CPI slightly hotter than expected at +0.3% MoM vs expects for +0.2%

In Tech, Semis +1% lead the way higher while Internet/Software +50bps. **ORCL +6%** on a better print while **ADBE -5%** continues to frustrate any remaining bulls left, with the print not doing anything to lift the AI overhang.

We’ll hit **ORCL** and **ADBE** first then move onto the usual.

Let’s get to it…


### **ORCL +5%: OCI cleared the buyside bar and the ATM is complete. Better print but ROI/’27 funding concerns still weigh**

A much better print overall with better IaaS growth and ATM being done the big positives, which is what buyside was focused on. OCI's fourth straight quarter of acceleration: +55%, +66-68%, +84% and +93% across FY26, now +121% — the largest step-up so far and mgmt expects IaaS acceleration through the year. Given better OpenAI sentiment recently and low expectations, positioning had a positive skew going into the print but this was still better than what bulls had hoped for - feedback was generally positive. This continues to be a tough stock, but a better print + continued momentum from OpenAI should help keep the bears away for now. Oct 28th investor day next catalyst

FQ1’27 revenue was $19.345B, +29.6% y/y (last q +20.6%), vs. Street ~$19.14B/+28.2%; non-GAAP EPS was $1.92, +30.4%, vs. Street $1.74–1.75.

OCI revenue reached $7.388B, +120.7% y/y (last q +93.2%), vs. Street ~$7.2B/+115%, clearing ~118% buyside growth bogey as capacity delivery accelerated.

Operating margin of 42.1% vs. Street 40.7–40.8% proving expense discipline can offset substantial infrastructure-driven gross-margin dilution.

FQ2 cloud growth guidance of 65–71%, midpoint 68%, better than Street expects of roughly 64–65%, although total revenue and EPS guidance were broadly in line.

### **Key Takeaways:**

-

**Capacity.** ORCL delivered 850MW and more than 300,000 GPUs, almost three times FQ4’s capacity deliveries and 73% of the entire FY26 build. That is meaningful evidence of execution, although 850MW matched some expectations and fell below the more aggressive “approaching 1GW” framing. Abilene’s customer-acceptance process has compressed to 24 hours, reducing the gap between infrastructure delivery and monetization. The first Vera Rubin systems are scheduled for customer delivery in FQ2, with mgmt reporting better-than-expected hardware quality, manufacturing yields and performance.

-

**Callback.** The most important clarification was that FQ1 benefited from both additional capacity and some deliveries pulled forward from FQ2. Even a few days or a week of earlier customer acceptance can move quarterly revenue, so the upside should not all be interpreted as an increase in ultimate capacity. Nevertheless, mgmt continued to expect IaaS growth to accelerate through FQ2, FQ3 and FQ4. The callback also indicated that customer-funded deployments can deliver similar or better profitability than traditional arrangements while improving returns on invested capital.

-

**Site risk.** Mgmt directly pushed back on the idea that New Mexico or Wisconsin threatens the existing FY27 revenue and earnings outlook. Construction was described as on track, but New Mexico’s air-permit process and Wisconsin’s grid-delivery arrangements remain unfinished. The defense is portfolio diversification, phased commissioning and backup capacity, not a claim that every project will arrive on schedule. None of New Mexico, Wisconsin, Michigan or Shackelford contributed to FQ1 deliveries, underscoring the breadth of the capacity program.

-

**Backlog.** RPO reached $664B, +45.9% y/y and +$26B sequentially, vs. Street approximately $642B/+41%; it was broadly in line with the cited buyside expectation. The $26B net increase is distinct from the more than $30B of new AI contracts signed. Most new contracts use prepayments, customer-owned hardware or similar structures, and do not begin affecting revenue or CapEx until FY28 or later. Customer and campus diversification improved, but no customer-concentration percentages were disclosed. Around half of RPO is now expected to convert into revenue over the next 36 months.

-

**Funding.** The completed $20B ATM removes the continuing issuance pressure from that program; it does not mean all future financing is secured. Up to another $20B of FY27 financing remains contemplated. Descriptions differ on the remaining debt/equity mix, which was not clarified on the call. Equally important, mgmt explicitly corrected the suggestion that new AI contracts require no additional CapEx: they require no additional ORCL cash funding under the arrangements described. Capital is still being deployed, but customers and suppliers are bearing more of the upfront funding burden.

-

**Cash flow.** OCF of $23.1B, +184% y/y, vs. Street ~$9.7B/+19%, included $11.4B of customer prepayments, nearly half the quarter’s operating cash generation. Gross CapEx was $28.5B, +235%, vs. Street ~$19.8B/+132%. Approximately half the expected full-year prepayments arrived while only about 31% of midpoint annual gross CapEx was incurred, making the quarterly FCF improvement particularly sensitive to timing. Mgmt retained the annual spending framework and explicitly declined to provide a date for consolidated FCF turning positive.

-

**Margins.** Non-GAAP gross margin fell to 61.0%, approximately 770bps lower y/y and 530bps lower sequentially, vs. Street ~62.0%. However, operating margin reached 42.1% vs. Street 40.7–40.8%, supported by a 9.1% y/y reduction in non-GAAP operating expenses. Mgmt’s structural argument is that infrastructure carries lower gross margins but also requires less incremental selling and development expense than software. That helps explain the operating-margin resilience, but does not eliminate the need to assess depreciation, financing costs and eventual cash returns. EPS upside also benefited from a lower tax rate and greater other income, not just operating execution.

-

**Pricing.** Capacity coming up for renewal was renewed or resold at an average 20% premium, with most of those GPUs already four years old or older. Utilization increased to 97.9% from 97.5% last quarter. Those are useful counterpoints to the GPU-obsolescence thesis, but the renewal uplift applies to that quarter’s renewal cohort, not the entire fleet. On component inflation, mgmt said higher costs are being reflected in customer pricing and do not change the prior infrastructure gross-margin framework. That is distinct from promising stable consolidated gross margins, which remain under pressure from mix and commissioning costs.

-

**Applications.** SaaS revenue was $4.219B, +9.9% y/y (last q +10.3%), vs. Street $4.257B/+10.9%. In CC, growth improved to 10% from 9%, so the acceleration depends on the currency basis. Fusion grew 14% versus 12% last quarter, while NetSuite slowed to 6% from 9%. Mgmt attributed NetSuite’s weakness to slower purchasing decisions last fiscal year rather than a newly deteriorating demand environment. NetSuite Next and AI-assisted implementations provide potential acceleration drivers, but the revenue turn remains something to demonstrate rather than assume.

-

**Database.** Cloud database revenue was approximately $0.9B, +26% y/y (last q +29%); multicloud database revenue grew 353% (last q +404%). No comparable Street growth benchmarks were quoted for these disclosures. Multicloud remains a potentially important higher-margin contributor, with distribution now spanning 70 regions and 119 availability zones. However, its rapid expansion has not yet produced acceleration in the overall cloud database business, which is the more important profitability check than the multicloud growth percentage alone.

-

**AI monetization.** Embedded AI usage rose 42% sequentially to more than 150 million uses, while production agent executions nearly doubled to more than 3.5 million. These demonstrate adoption, not a separately quantified incremental AI revenue stream. The more direct financial mechanism is faster implementation: AI can bring contracted applications into production sooner and accelerate revenue recognition. The AI Data Platform also extends beyond ORCL databases and applications, creating potential standalone consumption and cross-selling opportunities, but mgmt did not quantify the revenue contribution.

-

**Guide.** FY27 revenue moved from $90B to “at least $90B,” implying at least 33.6% growth vs. Street approximately $89.7B/+33.1%. EPS increased only $0.05 to $8.10, +6.2% on the reported FY26 base, vs. Street $8.07/+5.8%, despite roughly $0.18 of FQ1 upside. This is a modest confidence upgrade, not a wholesale earnings reset. Mgmt’s callback explanation was that substantial capacity still needs to be delivered and guidance reflects achievable outcomes rather than a predetermined pattern of quarterly raises. October 28 Investor Day is the next opportunity to clarify longer-term margins, funding and cash conversion.

### **Bull vs. Bear Debate**

The broader bull thesis is that ORCL can become a substantially larger cloud infrastructure provider without surrendering the economic advantages of its database and applications franchises. The durable advantage is not simply access to GPUs. Bulls point to the ability to assemble, commission and operate large clusters, allocate workloads across a multitenant fleet, support enterprise customers and connect infrastructure to established software and database products. This quarter strengthens that argument because OCI exceeded both Street and the buyside growth expectation while delivered capacity rose sharply. Fast customer acceptance, near-full utilization and higher renewal pricing suggest that ORCL is monetizing operational capability rather than merely accumulating hardware. The distinction matters for whether infrastructure growth deserves a durable earnings multiple or a more cyclical asset-rental valuation.

There are also growth opportunities beyond the largest AI customers. Multicloud database allows ORCL to monetize workloads inside other hyperscalers’ environments rather than requiring customers to move their entire application estate to OCI. Applications provide another route: embedded agents, faster implementations and the AI Data Platform could deepen ORCL’s role in business processes while pulling through database and infrastructure consumption. Sovereign and Alloy deployments broaden the customer base further. On the financial side, customer-owned hardware, prepayments and supplier arrangements could let deployed capacity grow faster than ORCL-funded investment. Bulls therefore see a route to better returns on capital even if infrastructure mix keeps consolidated gross margins below historical software levels. Higher estimates and a rerating would require continued delivery, resilient contract economics and progressively clearer evidence that cash generation is catching up with investment.

**The bear thesis** does not require AI demand to disappear. It requires the economics accruing to common shareholders to be less attractive than the revenue trajectory suggests. ORCL is committing capital today against long-duration customer demand while facing delivery, funding and future replacement-cost risks. Customer prepayments are a real financing benefit, but they also bring forward cash associated with future service obligations. Mechanically excluding the quarter’s $11.4B prepayments would move FCF from -$5.4B, versus Street approximately -$10.1B, to roughly -$16.8B. That is not a normalized FCF forecast, but it illustrates how much the reported cash result depended on upfront collections. The CFO’s refusal to set a positive-FCF date leaves an important issue unresolved: attractive project cash generation after commissioning is not the same thing as positive consolidated cash flow while the company keeps building.

Bears also question whether operating-expense reductions can continue offsetting mix dilution, depreciation and financing costs indefinitely. Gross margin was below Street even while operating margin exceeded expectations, and the annual EPS increase captured only a fraction of the quarter’s upside. The diversification story is improving, but the largest AI counterparties still matter disproportionately; a broader number of customers does not automatically eliminate correlated exposure to AI funding and monetization. Meanwhile, the software businesses that could cushion infrastructure economics did not produce uniformly strong results: applications were below Street and overall cloud database growth slowed.

Competition adds another reason not to extrapolate current scarcity pricing indefinitely. As outside context, AMZN’s AWS markets Trainium2-based instances for both training and inference and claims better price-performance than its own GPU-based alternatives. That is not a direct ORCL comparison, but it demonstrates that customers have an expanding set of architectural alternatives. The bear inference is that lower compute costs, greater supply or customer-specific silicon could eventually compress GPU rental economics even while total AI workloads keep growing. Today’s successful renewal cohort is encouraging evidence, not proof that every subsequent hardware generation will retain the same economic value.

### **ADBE -5%: AI adoption is working, but slowing ARR and a soft FQ4 guide leave monetization unproven. The Q slightly strengthened the adoption story without resolving whether AI can offset slowing legacy monetization.**

Not much to like here and bears remain in control as the “AI disintermediation” story continues to be front and center and the numbers not doing anywhere near the heavy lifting required to lift that overhang.

FQ3 revenue of $6.760B, +12.9% y/y (last q +12.7%), topped Street $6.694B, while EPS of $6.13 exceeded Street $6.08; the operating margin was slightly below expectations at 44.0% versus 44.1%.

Creative freemium MAU surpassed 100M, +70% y/y, and AI-first ARR exceeded $650M, +150%, but mgmt remains focused on acquisition and engagement rather than committing to a conversion timetable. NNARR of $400M exceeded Street $367M but declined roughly 39% y/y, while the FY26 ARR growth target stayed at 10.2%.

FQ4 revenue guidance of $6.800–6.850B, midpoint $6.825B and +10.2% y/y, slightly below Street $6.841B; the required bookings ramp rests primarily on enterprise seasonality and Acrobat innovation, not renewed pricing.

### **Key Takeaways**

-

**RR.** The issue is growth quality, not an outright Street disappointment. Ending ARR of $27.50B, +11.2%, was modestly above Street approximately $27.47B, but incremental ARR generation continued to deteriorate. Debate around what’s the right organic ARR growth excluding Semrush vs 10.5% last quarter.

-

**Freemium.** The customer-acquisition strategy is gaining traction: creative freemium MAU exceeded 100M, +70% y/y, versus over 90M last quarter, while total MAU surpassed 1B, +20%. The roughly 10M sequential creative-user addition was similar to last quarter, so strong adoption has not yet become an accelerating acquisition curve. Mgmt is shifting discovery from searches for a product name to searches for a specific task, routing users into workflow-specific “micro-studios,” increasing engagement, and adjusting when the paywall appears. There is no committed date or disclosed conversion-rate target for turning that activity into material incremental ARR.

-

**Callback.** The most important clarification was that weaker RPO primarily reflected contract mix, not worsening retention. Newer creator offerings increasingly monetize through month-to-month subscriptions rather than annual commitments; enterprise contract duration also shortened modestly, but mgmt characterized that impact as limited and explicitly did not view the shift as competition-driven. RPO of $22.16B, approximately +8% y/y, was below Street $22.60B, while cRPO growth slowed to 9% from approximately 13% reported last quarter. Shorter duration helps explain RPO weakness, but it does not, by itself, explain weaker annualized recurring-revenue additions.

-

**Demand.** Business Professionals & Consumers remained the strongest financial contributor: subscription revenue of $1.910B, +15.8% y/y (last q +15.6%), exceeded Street approximately $1.880B, supported by Acrobat, SMB channels and enterprise AI adoption. Creative & Marketing Professionals delivered $4.650B, +12.9% y/y (last q +12.9%), versus Street approximately $4.625B, but benefited from a full quarter of Semrush. Mgmt highlighted a healthy enterprise pipeline and stable retention, not a broad acceleration in enterprise IT budgets. On the outlook, mgmt attributed the less-than-full revenue upside flowing into the annual raise to FX, saying nothing had changed in the business.

-

**AI.** AI-first ARR exceeded $650M, more than +150% y/y and approximately +30% sequentially against the prior disclosed threshold. That is real monetization, but still only approximately 2.4% of total ARR, and the definition includes Semrush AI products. Firefly app and credit-pack ARR grew 40% q/q, with video usage and broader credit consumption supporting expansion. Importantly, that metric excludes Firefly Enterprise and cannot be multiplied against the prior, broader $300M Firefly ARR disclosure. Acrobat AI Assistant MAU doubled q/q, but usage growth is not equivalent to growth in paid subscriptions.

-

**Enterprise.** ADBE continues to show better momentum where AI is embedded in enterprise workflows rather than sold as a standalone generation tool. ARR grew more than 20% y/y in each of AEM and agentic web apps, GenStudio, and AEP and apps. Brand Visibility paid customers doubled q/q, while CX Enterprise Coworker had more than 1,700 customers and early adopters following its June launch, not necessarily 1,700 paying customers. The strategic opportunity is to combine content creation, brand context, marketing execution and measurement into larger engagements. No comparable Street forecasts or incremental contract-value breakdowns were disclosed for these measures.

-

**Pricing.** Deferred Creative Cloud pricing actions remain deferred, with no contribution assumed in FQ4. This is a continuation of the prior-quarter strategy, not a newly announced pricing retreat. Shantanu explicitly said he was pleased ADBE had prioritized new-user adoption over the short-term relief pricing could have provided. The trade-off remains meaningful: the earlier combination of deferred line optimizations and freemium expansion was estimated to reduce FY26 ending ARR by approximately $500M.

-

**Margins.** Non-GAAP OM of 44.0% versus Street 44.1% fell approximately 230bps y/y and 50bps q/q; gross margin of 89.5% versus Street approximately 89.7% declined approximately 100bps y/y. Cloud, inference and AI development costs are being incurred before the full monetization benefit arrives. The callback emphasized model-routing optimization, inference efficiencies and balancing proprietary versus third-party models as offsets, but supplied no quantified AI contribution-margin or cost-per-generation bridge. EPS growth of 15.4% versus Street 14.5% materially exceeded operating-income growth of 7.2% versus Street approximately 6.5%, underscoring the importance of share-count reduction.

-

**Guide.** The unchanged annual ARR target implies approximately $777M of FQ4 NNARR, down roughly 16% y/y but nearly double FQ3, versus approximately $731M implied by the Street ending-ARR target using the same FQ3 starting point. Mgmt’s bridge is enterprise year-end bookings, Acrobat and productivity releases, and Firefly/Creative Cloud innovation around MAX. It is not a guide to an immediate freemium-conversion inflection. The annual revenue midpoint increased approximately $51M versus the quarter’s approximately $65M upside to the prior guide midpoint; mgmt attributed the difference to incremental FX pressure.

-

**Leadership.** Anil Chakravarthy becomes CEO on December 1, with Shantanu becoming Executive Chair, but the broader leadership transition remains unfinished: the CFO is still interim and David Wadhwani is stepping down from Creativity & Productivity. Anil rejected the implication that his enterprise background means a retreat from Creative, emphasizing all three product categories and customers from individuals through enterprises. His emphasis is on agentic software, model flexibility and enterprise context. MAX in November is the next product checkpoint; a fully articulated financial framework may take longer as the leadership team is assembled.

-

**Topaz.** The pending acquisition adds AI enhancement capabilities across imaging and video, with a natural fit in Creative Cloud, Firefly and enterprise production workflows. Mgmt expects a FQ4 close; callback commentary characterized the quarterly contribution as immaterial. One estimate assumes approximately $50M of acquired ARR embedded in the unchanged annual target, which would modestly weaken the implied organic outlook, but that amount is not a mgmt disclosure. The acquisition should therefore not be treated as a quantified organic-growth offset.

### **Bull vs. Bear Debate:**

**The broader bull thesis** is that ADBE can remain the monetization layer for professional creative work even when it does not own the best underlying AI model or the interface where a task begins. Its durable advantages are the installed workflow, precision editing, document expertise, brand context, enterprise integration and distribution across creative production and marketing. AI can increase the volume and variety of content produced, creating opportunities to charge for consumption, automation and larger enterprise workflows rather than relying entirely on additional seats. Current Firefly product capabilities reinforce this possibility: partner models sit alongside proprietary models, with outputs flowing into editing and production tools. Bulls therefore do not need Firefly to win every model comparison; they need ADBE to retain the customer relationship and capture enough value from completing the work.

This quarter supplied evidence that ADBE is attracting users rather than simply defending a shrinking audience. The rapidly expanding freemium base, growing paid credit consumption, durable Acrobat demand and stable enterprise retention support a transition thesis rather than an immediate franchise-breakdown thesis. Enterprise adoption offers a potentially better economic path than monetizing casual users: brand-specific models, governed workflows and customer-context integration could expand contract values even if creative teams become more productive with fewer seats. The rerating catalyst would be evidence that stronger engagement is translating into incremental paid ARR while inference efficiencies stabilize margins. Clearer leadership, more consistent disclosure and sustained repurchases could help, but bulls ultimately need the new products to improve growth economics, not merely usage statistics.

**The broader bear thesis** is not that creative activity disappears. It is that AI makes useful creative output easier to produce while reducing the premium users will pay for specialized tools, established workflows and professional seats. Growing content volume therefore need not translate into growing software spend at historical pricing. Competition is also moving beyond isolated generation tools into more complete workflows. Affinity, for example, offers professional design, photo editing and page layout free for individuals, supports common creative file formats and integrates with Canva, with optional AI features tied to premium plans. That does not establish equivalent capability or prove customer losses, but it raises the burden on ADBE to demonstrate why users should convert to paid products rather than remain free or use alternatives.

The quarter fits that concern: revenue and EPS modestly exceeded Street, yet organic ARR growth slowed, the annual ARR target did not improve, contract duration shortened and margins contracted before large-scale freemium monetization. The Acrobat analogy also has limits. Reader historically put ADBE inside an existing document workflow where editing or signing created a natural paid conversion point; Creative freemium often has to acquire users externally around a discrete task that competing products may also solve. AI-first ARR is encouraging but remains a small portion of the total, and rising credit usage does not reveal the profitability of serving that demand. Leadership changes add the possibility of an investment-heavy or conservatively reset FY27 framework. Buybacks can support EPS, but they cannot indefinitely offset weak operating growth, and the extra week could temporarily obscure that distinction in reported numbers.

# **TECH RESEARCH/NEWS**

### **Microsoft Plans Data Center Push to Triple Its Computing Power - Bloomberg**

Bloomberg reports Microsoft is planning to more than triple global data center capacity to **>38GW by 2032 from ~12GW today**, with AI-specific capacity rising from ~2GW to roughly one-third of the total. The buildout is aimed at relieving compute shortages that have constrained Azure/GitHub growth and pushed some customers to rivals, though Microsoft says Bloomberg’s specific planning figures are not accurate.


### **AI Financing/GOOGL/AVGO: Inside Blackstone’s Bid to Rule Over AI Financing - [TheInformation](https://www.theinformation.com/articles/inside-blackstones-bid-rule-ai-financing?utm_campaign=article_email&utm_content=article-17799&utm_medium=email&utm_source=sg&rc=h3on51)**

**

*The Information reports Blackstone **plans to buy “multiples” more Google TPUs than the $5B/500MW initially announced, implying tens of billions of dollars and multiple gigawatts of AI compute.** Blackstone says compute remains its “biggest conviction bet,” viewing the AI infrastructure buildout more like an industrial revolution than a normal tech cycle. The firm is also financing Nvidia GPUs and separate Broadcom/Anthropic TPU deployments, positioning itself to become one of the largest capital providers to the AI ecosystem*

### **OpenAI: OpenAI Is Open to Slowing Cutting-Edge AI, CEO Sam Altman Tells Staff - [Bloomberg](https://www.bloomberg.com/news/articles/2026-09-11/openai-is-open-to-slowing-cutting-edge-ai-ceo-sam-altman-tells-staff?srnd=homepage-americas)**

**

*Bloomberg reports Sam Altman told OpenAI staff the company could slow frontier AI development, potentially alongside other major labs, as safety concerns intensify across the industry. OpenAI has already paused parts of model development over safety issues, while Chief Scientist Jakub Pachocki has called for voluntary slowdowns until shared safety standards are established.*

### **Anthropic*****: *****Anthropic disrupts bioweapons research efforts, Russian hacking, Chinese Claude misuse - [Reuters](https://www.reuters.com/legal/litigation/anthropic-disrupts-russian-chinese-ai-campaigns-targeting-its-claude-models-2026-09-10/)**

**

*Reuters reports Anthropic disrupted misuse of Claude spanning bioweapons research, weapons development and a suspected Russia-linked cyber campaign targeting Ukrainian officials, highlighting how rapidly improving models are enabling more sophisticated offensive activity. Anthropic also says seven China-based AI labs, including Alibaba, Moonshot and DeepSeek, tried to extract Claude’s capabilities, with Alibaba allegedly generating 151M+ exchanges*

Actual report from Ant: [here](https://www.anthropic.com/threat-intelligence-report-september-2026#scams-and-fraud-sep-26)

### **INTC: Silver Lake, Intel-backed Altera prepares IPO that could raise over $2 billion as early as 2026- [Reuters](https://www.reuters.com/world/silver-lake-intel-backed-altera-prepares-ipo-that-could-raise-over-2-billion-2026-09-10/)**

**

*Reuters reports Silver Lake/Intel-backed Altera is preparing an IPO that could raise >$2B, with a confidential filing expected in coming weeks and a listing possible as early as this year. Altera was valued at $8.75B when Silver Lake bought 51% last year, while Intel still owns 49%, setting up a potentially meaningful mark-up if the IPO prices well.*

### **HUT/IREN/GLXY/CORZ/CIFR: Jefferies says ERCOT Batch Zero de-risks some 2027 power, but scrutiny remains a meaningful timing risk**

Jefferies says HUT and IREN secured conditional base-load status for key Texas projects, allowing power studies to proceed and materially de-risking 2027 energization, while GLXY and CORZ had more mixed outcomes with some capacity left as conditional studied load. Jefferies flags CIFR as the biggest uncertainty, with 3.2GW in Batch Zero but no disclosed base/studied split, potentially implying less favorable allocations. Jefferies views Texas’ audit of the ~474GW interconnection queue as a broad headwind to construction timelines, though projects with signed tenants and proven developers should be best positioned to clear the process.

### **AI Semis/AMD/INTC: MS sees 2.5D packaging capacity +~50% in 2028 vs CSP capex +12%, reinforcing compute as the priority spend**

MS says aggregate advanced-packaging capacity should reach ~370kwpm of CoWoS plus 40–45kwpm of EMIB-T in 2028, materially outgrowing overall AI capex as larger chips and broader GPU/XPU/CPU/networking adoption drive silicon intensity higher. MS sees AMD as a major incremental 2027 driver, with CoWoS consumption +165% y/y to ~345K wafers on MI450/455 and Venice, while Maia 200 and a U.S. AI lab/AVGO ASIC add further demand. MS also expects Intel EMIB-T to reach ~10–15% of 2.5D capacity by 2028, largely consumed by MediaTek/Google TPU v9

### **NFLX: Hearing M-sci mixed saying gross adds in U.S./Europe weak but Asia better**

### **SPOT: Hearing M-sci saying net adds tracking above street**

### **CHWY: JPM downgrades to Neutral, cuts PT to $24 as organic growth remains muted**

JPM says CHWY’s organic growth is still pressured by macro and a weak pet-industry backdrop, limiting near-term upside despite continued share gains. JPM remains constructive on strategic initiatives, margin expansion and capital returns, but sees the risk/reward as more balanced after 2Q. PT falls to $24 from $29.

### **CDNS: FBN initiates OP/$375, sees agentic AI as an EDA demand accelerator rather than commoditization risk**

FBN says agentic AI should increase EDA consumption because autonomous design agents invoke CDNS’s physically accurate engines far more frequently than human designers, reinforcing the value of its mission-critical position alongside SNPS. FBN highlights F2Q26 revenue +24% y/y and record $8.1B backlog, while modeling FY26/FY27/FY28 EPS of $8.14/$9.59/$11.38. FBN sees the current ~31.6x NTM P/E as an attractive entry versus CDNS’s historical mid/high-30s multiples and initiates Outperform with a $375 PT.

### **DELL: RBC initiates OP/$640, sees AI backlog + storage mix driving durable growth and margin expansion**

RBC says DELL exited 2Q27 with ~$95B of AI server backlog and a pipeline that should convert to >$132B of orders over the LTM, with demand increasingly constrained by memory supply rather than customer appetite and visibility extending into FY28. RBC sees Dell IP storage as the bigger margin lever, with mix shift and the completion of the legacy storage transition supporting ISG operating margins in the 10–14% range. RBC also sees incremental upside from agentic-AI-driven CPU/server demand and the enterprise refresh cycle, and values DELL at ~21x CY27 EPS versus a ~24x five-year NTM average.

### **HPQ: RBC initiates Sector Perform/$33, sees AI PC mix reaching 60–70% by FY27 but near-term margin pressure caps upside**

RBC says HPQ is well positioned for the AI PC refresh, with AI PCs expected to reach 60–70% of shipments by FY27 versus ~45% in 3Q26, but sees near-term upside constrained by memory/storage inflation and ongoing Print declines. RBC expects PC units to fall mid-teens y/y in 2H26 while revenue holds up better through pricing, premium mix and share gains, with peripherals/services attach supporting profitability. RBC sees Print margins remaining in the upper half of the 16–19% long-term range but initiates at Sector Perform given valuation and execution risk.

### **SHOP: Bernstein initiates OP/$160, argues AI expands the commerce market more than it disintermediates Shopify**

Bernstein says AI-native builders and purchasing agents create real long-term risks to SHOP’s storefront and payments economics, but sees those threats as slow-moving given consumer behavior and the complexity of omnichannel retail. Bernstein argues AI should also lower the barrier for new entrepreneurs, expanding SHOP’s addressable market while its unified commerce platform already spans international, physical and B2B, now >60% of sales volume. Bernstein also sees continued pricing power, with revenue/GMV rising from ~2.5% to ~3% over the past decade, and initiates Outperform with a $160 PT.

### **PANW: Wedbush adds to Best Ideas, raises PT to $400 on acquisition-driven platform expansion**

Wedbush says PANW is positioned to reaccelerate growth as recent and future acquisitions add modules across the platform without compromising cash generation. Wedbush views the broader product suite as strengthening platformization and increasing wallet share per customer. Wedbush assumes Outperform, raises PT to $400 from $340 and adds PANW to its Best Ideas List.

### **DDOG: Wedbush assumes OP, raises PT to $275 on AI-native observability leadership**

Wedbush says DDOG is accelerating growth while maintaining strong margins and remains the best-positioned pure-play beneficiary of rising cloud- and AI-native workloads. Wedbush sees observability demand strengthening as AI infrastructure and applications increase system complexity and monitoring requirements. Wedbush assumes Outperform and raises PT to $275 from $260.

### **YELP: JPM assumes UW/$19, says AI transition is promising but growth visibility remains poor**

JPM says YELP’s AI transformation is strategically encouraging, but limited visibility and an unfavorable macro backdrop keep the near-term risk/reward unattractive. JPM wants to see a return to revenue growth and further cost rationalization before becoming more constructive. JPM assumes Underweight with a $19 PT.

### **QCOM/Samsung: Samsung and Qualcomm’s 2nm foundry deal may slip into next year due to pricing disagreements, not technical problems - [TheBell](https://www.thebell.co.kr/front/newsview.asp?code=0403&key=202609091437016680105190).KR**

### **PLTR: Rosenblatt sees commercial momentum sustaining through 2026+, reiterates Buy/$225**

Rosenblatt says AIPCon customer checks point to a strong pipeline of commercial conversions across multiple verticals, supporting continued momentum into 2027. Rosenblatt highlights Titan, expanded implementation capacity through PwC and the hire of former AIG Chairman Peter Hancock to lead financial services as signs PLTR is scaling both product reach and enterprise execution. Rosenblatt reiterates Buy and a $225 PT.

### **TWLO: BTIG sees contextual data + identity as key moats for agentic AI, reiterates Top Pick/$285**

BTIG says TWLO is emerging as foundational communications infrastructure for agentic AI, with its CDP + Conversation Memory reducing token costs by supplying only relevant customer context while telecom metadata and identity products help solve the trust/authentication problem. BTIG sees Conversation Memory as one of Signal’s strongest early products, carrying a 15–30% uplift, with Retail/Food/E-commerce likely to adopt first and regulated verticals later. BTIG reiterates Buy/$285 and Top Pick, arguing the current setup still offers improving growth, margin expansion and capital discipline despite the stock’s ~67% YTD move.

### **EQIX: Evercore sees AI inference as a new interconnection tailwind underpinning raised +10–13% revenue growth outlook**

Evercore says inference workloads should favor EQIX’s dense interconnection footprint because they are more distributed and latency-sensitive than training, creating incremental demand across IBX facilities. Evercore highlights management’s upgraded medium-term targets of +10–13% revenue growth and +9–12% AFFO/share growth, up from +7–10% and +5–9% previously, with EBITDA margins reaching 53%+ by CY29. Evercore also sees new Fabric One and Inference Exchange products as additional high-margin monetization levers and reiterates OP/$1,270.

### **RBLX: Oppenheimer says RDC focus will be AI creation/search, but developer-tooling gaps remain**

Oppenheimer says RDC 2026 should center on AI content creation, search/discovery and analytics, with management likely showcasing Build, Roblox Reality and potentially Moment 1.0, while major updates on ads, commerce and developer payouts look unlikely. Oppenheimer says developers still want better programmatic tooling, first-party analytics, native UA/marketing tools, licensing support and more algorithm stability, suggesting a persistent gap between platform priorities and creator needs. Oppenheimer also flags continued studio consolidation, though recent M&A appears more specialized and less competitive than last year.

### **HPQ: RBC initiates Sector Perform/$33, sees AI PC mix reaching 60–70% by FY27 but near-term margin pressure caps upside**

RBC says HPQ is well positioned for the AI PC refresh, with AI PCs expected to reach 60–70% of shipments by FY27 versus ~45% in 3Q26, but sees near-term upside constrained by memory/storage inflation and ongoing Print declines. RBC expects PC units to fall mid-teens y/y in 2H26 while revenue holds up better through pricing, premium mix and share gains, with peripherals/services attach supporting profitability. RBC sees Print margins remaining in the upper half of the 16–19% long-term range but initiates at Sector Perform given valuation and execution risk.

### **CART/DASH/UBER: Wolfe says grocery delivery price gap is collapsing, supporting broader adoption and share gains for DASH/UBER**

Wolfe says third-party grocery prices now average only MSD% above in-store versus ~10% previously, while CART was cheapest after fees/promos and overall platform pricing differed by just ~1%. Wolfe notes AMZN remains LDD% cheaper than the 3P apps but has narrower selection and the highest out-of-stock rates, while CART/DASH offer a more seamless shopping experience.

### **TSLA: MS sees autonomous Semi software as a high-margin earnings lever, adds $20/share to bull case**

MS says autonomous trucking could deliver ~2.3x higher utilization, ~20% lower cost/mile and ~6x higher profit per truck versus human-driven fleets, supporting a ~$0.85–$1/mile FSD fee or ~$12–18K/month per Semi. MS estimates ~82K autonomous Semis by 2040 could generate ~$17B of recurring software revenue and ~$7.5B of incremental EBIT, ~10% upside to its base case, excluding vehicle/charging revenue. MS keeps EW/$400 but raises its bull case to $840 from $820, with Musk targeting self-driving Semi capability by 2027.

### **CRM: UBS says Claudeforce strengthens the apps layer, but sees limited near-term revenue impact**

UBS views the expanded Anthropic partnership positively, arguing tighter model-to-app integration, pre-built Salesforce tasks and native permissioning should deliver better AI ROI than simple MCP connectivity and reinforce the value of the application layer. UBS says the recent ~20% stock move was largely a narrative reset rather than an estimates change, while “headless” AI access is likely inevitable but could ultimately increase Salesforce usage and customer satisfaction.

### **FTNT/CHKP/DT: Downgraded to Neutral from O/P at Wedbush**