J P M O R G A N North America Equity Research 03 August 2025 Hardware & Networking Weekly Tech Talk: Soaring AI Spend into 2026 to Support Soaring Valuations This week, we are focusing on the implications of the messages we heard from all the Cloud companies regarding their spending plans for IT infrastructure, understandably with a priority on AI. The year 2024 set a high benchmark for 2025, due to the revival of new datacenter builds following a lean period in 2023. However, the upsizing of capex plans by ...

North America Equity Research 28 July 2025 Hardware & Networking Scale-Up AI Networking in Numbers: Key Forecasts from 650 Group for Scale-Up AI and Technology of Choice We recently hosted a conference call with Alan Weckel of 650 Group and Harlan Sur, J.P. Morgan's Semi & Semi Cap Equipment / IT Hardware analyst, to discuss Scale-up AI Networking. Please see Harlan Sur's report here on the takeaways from the call on Scale-up AI Networking. Below we walk through the key forecasts in relation to AI Networkin ...

Summary of European AI & Semis Symposium Industry Overview - The symposium focused on the **semiconductor** and **artificial intelligence (AI)** industries, featuring over **500 participants** and **25+ speakers** from various sectors including academia, corporate, and private companies [1][2] Key Insights - **Cost Efficiency of Generative AI**: The capabilities of Generative AI are being validated as cost efficiency improves, with some large language models being up to **400x cheaper** for training and inference compared to earlier versions [4][5] - **Productivity Gains**: AI is driving measurable productivity gains, with claims that AI can equate to a workforce with **6x the bandwidth** in software development tasks [4][5] - **AI Workload Management**: At Salesforce, AI now handles up to **50% of workloads**, while at Google, AI accounts for **25% of new code lines** [4][5] - **Advancements in AI Capabilities**: OpenAI and DeepMind's models have achieved gold-medal scores at the International Mathematical Olympiad, outperforming most human participants [4][5] - **Multi-modality in AI**: AI is evolving to execute algorithms and generate various sensory inputs and outputs, expanding its applications across sectors like healthcare, retail, and industrials [5][6] Sector-Specific Applications - **Healthcare**: AI is significantly impacting healthcare, with **95% of certain cancer procedures** in the US being performed robotically. AI-linked mirrors can detect **90% of known diseases** [7][8] - **Entertainment**: Netflix reported a **10x acceleration** in production through the integration of Generative AI into its visual effects pipeline [7][8] - **Retail**: AI-powered robots in retail can perform stock-taking and predictive analysis, achieving a **3x to 4x ROI** within **3 to 4 months** [7][8] - **Education Technology (EdTech)**: AI created **1,200 learning modules** of **35 hours each** in just three days, enhancing consumer engagement by **3x** through AI-driven media solutions [7][8] Challenges and Solutions - **Power Consumption**: Data centers may require as much power as **New York City**, necessitating new power generation sources and more efficient distribution grids [6][9] - **Guardrails for AI Content**: The need for human reviewers and technical safeguards is emphasized to mitigate risks associated with AI-generated content [9][10] - **Transparency and Bias Mitigation**: Mechanisms to explain AI's decision-making processes and avoid bias are crucial for reliability [9][10] Technological Innovations - **Photonics and Quantum Computing**: These technologies are seen as transformative, with potential for **70% reduction** in energy consumption for certain accelerators [9][10] - **EUV Lithography Tools**: ASML's tools are essential for improving yield and reducing cycle time in semiconductor production, supporting AI training with increased power and reduced latency [9][10] - **Hybrid Bonding**: This technology offers advantages in chip performance and thermal efficiency, critical for data centers [9][10] Company Highlights - **Logitech**: Leveraging thousands of autonomous robots for continuous operation, resulting in clear cost benefits [7][8] - **Infineon**: Positioned positively due to its role in providing AI power semiconductors with a reliable track record [9][10] - **Nebius**: Focused on cost and energy-efficient GPUs, indicating a strong market position [9][10] Conclusion The symposium underscored the rapid advancements in AI and semiconductor technologies, highlighting both the opportunities for productivity gains and the challenges that need to be addressed for sustainable growth in these sectors [1][2][8]

Core Viewpoint - Palo Alto Networks is negotiating to acquire identity management software maker CyberArk for over $20 billion, marking the largest acquisition under CEO Nikesh Arora's leadership [1] Company Summary - CyberArk specializes in identity management and faces competition from major players such as Microsoft, Okta, and IBM [1] - The potential acquisition is part of a broader trend of consolidation in the cybersecurity industry, following significant mergers like Cisco's acquisition of Splunk and Google's acquisition of Wiz [1]

Summary of Key Points from the Conference Call Company and Industry Involved - **Company**: Google (Alphabet Inc.) - **Industry**: Cloud Computing, Hardware & Networking Core Insights and Arguments - **Capex Growth**: Google reported a significant increase in capital expenditures (capex) for Q2 2025, with a rise of **+70% year-over-year** to **$22.4 billion**, exceeding the consensus estimate of approximately **$18 billion** [1] - **Full-Year Outlook**: The company raised its full-year capex outlook for 2025 to **$85 billion**, up from a previous estimate of **$75 billion**, indicating a year-over-year growth of **60%+** [1] - **Investment Focus**: The majority of the capex is directed towards technical infrastructure, with **two-thirds** allocated to servers and the remaining to datacenters and networking equipment [1] - **Future Projections**: Management hinted at further increases in capex for 2026, driven by strong customer demand and growth opportunities [1] Additional Important Information - **Implications for Other Hyperscalers**: Google's capex results are expected to set a precedent for other U.S. hyperscalers, suggesting a similar trend in spending appetite when they report their earnings [1] - **Supplier Impact**: Companies with exposure to AI infrastructure spending, such as Celestica, Flex, Arista, and others, are anticipated to benefit from this increased capex [1] - **Historical Capex Trends**: The report includes a historical overview of Google's quarterly capex, showing fluctuations and significant increases in recent quarters, particularly in Q2 2025 [2] This summary encapsulates the critical financial insights and future expectations regarding Google's capital expenditures and their implications for the broader cloud computing and hardware industry.

标普道琼斯指数：美国数据服务公司Datadog将取代Juniper Network加入标普500指数。 ...

Group 1 - The three major U.S. stock indices opened higher, with the Dow Jones up 0.5%, the Nasdaq up 0.43%, and the S&P 500 up 0.32% [1] - Huyue Technology surged over 11% after the U.S. Department of Justice resolved the company's $14 billion acquisition of Juniper Networks in an antitrust case [1] - Oracle's stock rose over 7% as the CEO announced a strong start to the fiscal year 2026, having signed multiple large cloud service agreements [1]

Group 1 - The U.S. Department of Justice has reached a settlement regarding Hewlett Packard Enterprise's $13 billion acquisition of Juniper Networks [1]

Core Viewpoint - Palo Alto Networks (PANW.US) reported mixed financial results for Q3 due to market volatility in April, but analysts remain optimistic about the stock, anticipating clearer growth prospects in Q4 [1] Financial Performance - Q3 FY2025 sales increased by 15% year-over-year to $2.29 billion, slightly above analyst expectations of $2.28 billion [1] - Net profit decreased by approximately 6% to $262 million, with adjusted earnings per share at $0.80, exceeding the expected $0.77 [1] - Non-GAAP gross margin was 76%, below the anticipated 77.2% [1] Subscription Revenue Concerns - Subscription revenue grew by 18% to $1.234 billion, falling short of the market expectation of 20.1% and lower than the 20% growth rate in Q2 [1] - Analysts noted that the slowdown in subscription growth is a key concern, similar to trends observed in Fortinet [1] Analyst Ratings and Price Targets - Morgan Stanley reiterated an "Overweight" rating with a target price of $205, highlighting the CEO's candid discussion about challenges in Q3 [2] - Bank of America maintained a "Neutral" rating with a target price of $125, citing concerns over peak financial metrics and high non-GAAP P/E ratios [2] - Evercore ISI kept an "Outperform" rating and raised the target price from $215 to $220, acknowledging healthy platform and new product metrics [2] - Wedbush also maintained an "Outperform" rating with a target price of $225, viewing 2025 as a transformative year for the company [2] Industry Context - Cybersecurity peers experienced stock price declines, with Fortinet and Check Point Software Technologies down 0.5% and 0.6%, respectively, while CrowdStrike and SentinelOne fell by 1% and 2.5% [3]

Core Insights - Palo Alto Networks reported Q3 revenue of $2.3 billion, exceeding analyst expectations of $2.28 billion [1] - The company revised its full-year revenue forecast to $9.17 billion to $9.19 billion, up from the previous estimate of $9.14 billion to $9.19 billion [1] - For Q4, the expected revenue is between $2.49 billion and $2.51 billion, slightly below the analyst expectation of $2.5 billion [1] - Following the earnings report, Palo Alto Networks' stock fell by 2.84% in after-hours trading [1]