J P M O R G A N North America Equity Research 07 December 2025 Broadcom Inc F4Q25 Preview: Expect Beat/Raise On AI Strength; Strong TPU v6 (Ironwood) And AI Networking; Next Gen TPU v7 Set To Ramp Mid CY26; See >$50B AI Revenues Next Year; Reit OW Broadcom is set to report Oct quarter (F4Q25) earnings Thursday, and we continue to see a strong demand profile for its AI products (custom ASICs and networking solutions), gradually improving non-AI semi business (e.g., enterprise, server/ storage, broadband) and ...

Core Insights - Marvell Technology (MRVL) shares have declined by 14.1% over the past year, underperforming the Zacks Electronics - Semiconductors industry and the Zacks Computer and Technology sector, which grew by 70.4% and 26.8% respectively [1] Financial Performance - In Q3 of fiscal 2026, Marvell's data center segment generated revenues of $1.52 billion, reflecting a year-over-year increase of 37.8% [4][8] - The Zacks Consensus Estimates project a revenue growth rate of 42% and an earnings growth rate of 81% for fiscal 2026 [14] Product and Market Positioning - Marvell's products are essential for AI computation and networking, particularly in hyperscale and enterprise environments, as demand for high-bandwidth interconnects and custom ASICs rises [5][8] - The company is expanding its AI portfolio with new technologies such as interconnect, Co-packaged optics (CPO), and optical DSP [8][12] Strategic Initiatives - Marvell has acquired Celestial AI to enhance its capabilities in the interconnect space, positioning itself at the forefront of next-generation AI data center architectures [6] - Partnerships with industry leaders like Amazon Web Services (AWS) and NVIDIA are crucial for supplying connectivity products for AI and data-center workloads [7][9] Competitive Landscape - Marvell faces competition from semiconductor giants like Broadcom, AMD, and Micron Technology, but its rapid portfolio expansion helps it stay competitive [10][12] - The company is developing advanced solutions such as a co-packaged copper system and a telemetry API for real-time network performance monitoring [12][13] Investment Recommendation - Given its strong fundamentals, deep AI portfolio, and strategic partnerships, Marvell Technology is recommended as a buy [17][18]

Core Viewpoint - Marvell Technology is experiencing a challenging year despite a recent stock rebound, primarily due to competition in the ASIC market and uncertainties surrounding key customers like Amazon Web Services [1][4][10]. Financial Performance - Marvell reported record revenue of $2.01 billion for Q2 of fiscal 2026, marking a 58% year-over-year increase, driven by strong AI demand in data centers [6][7]. - Data center revenue surged 69% year-over-year to $1.49 billion, while enterprise networking and carrier infrastructure sales increased by 28% and 71%, respectively [7]. - The company's gross margin expanded by 420 basis points to 50.4%, and net income turned from a loss of $193.3 million last year to a profit of $194.8 million [7]. - Adjusted earnings per share rose 123% year-over-year to $0.67 [7]. Market Position and Competition - Marvell focuses on specialized silicon for complex and energy-intensive computing workloads, aiming to remain competitive in the evolving tech landscape [2][3]. - Despite its growing role as an AI supplier, Marvell faces strong competition from Broadcom in the ASIC market, which may hinder its growth potential [4][10][11]. - Analysts predict that Marvell's ASIC business may lag behind peers through 2027, with concerns about its ability to capitalize on increasing ASIC spending [10][11]. Customer Dynamics - Amazon remains Marvell's largest ASIC customer, but its involvement in upcoming projects appears limited, raising concerns about potential market share loss [12]. - A delay in a significant ASIC program with Microsoft is anticipated, which could impact Marvell's growth trajectory [12]. Analyst Sentiment - Wall Street maintains a consensus "Moderate Buy" rating for Marvell, with a price target of $92.07, suggesting a potential upside of about 3.28% [13]. - More optimistic analysts project a target of $122, indicating a potential 36.79% upside if Marvell successfully executes its AI strategy [13].

Key Takeaways from J.P. Morgan's Expert Discussion on AI Datacenters Industry Overview - The discussion focused on the **AI Datacenter** industry, particularly the scaling up of AI Datacenters and the evolving architecture for hyperscale AI workloads. Core Insights 1. **Shift in Compute Capex**: - There is a rapid shift in compute capital expenditures (capex) towards inference workloads, with techniques like distillation and multi-step optimization yielding significant near-term gains. By approximately **2027**, the share of compute dedicated to inference is expected to surpass that of training workloads [3][4][5]. 2. **Preference for Smaller Models**: - Enterprises are increasingly adopting smaller, fine-tuned models over larger ones, accepting slight quality trade-offs for reduced costs in inference workloads. This trend is exemplified by Cursor's new coding model [3][4]. 3. **Standardization in Hardware**: - The industry is witnessing a move towards standardization in inference-related networking hardware, with expectations for more rack-level standardization in the coming year. White-box solutions are gaining traction through Open Compute Project (OCP) initiatives [3][4]. 4. **Training Constraints**: - Training workloads are facing constraints primarily due to power supply issues, while inference workloads are less affected. The power demands for training are significantly higher, estimated at **5-10 times** that of inference [4][5]. 5. **Longer GPU Lifespan**: - Buyers are now planning for a useful life of **five to six years** for GPUs, an increase from the previous **four years**. This shift reflects a strategic move to repurpose GPUs from training to inference tasks [5]. 6. **Storage Solutions**: - The storage landscape remains hybrid, with HDDs maintaining cost leadership while Flash/NAND is preferred for high-performance needs. Advances in HDD technology, such as HAMR, are helping HDDs remain competitive [5]. 7. **Beneficiaries of Capex Shift**: - Companies like **Broadcom**, **Marvel**, and **Celestica** are expected to benefit from the shift towards inference workloads. Broadcom's work with custom ASICs for major players like Google and Amazon positions it favorably in this evolving market [5]. Additional Important Points - The discussion highlighted the growing comfort among operators in mixing branded and white box solutions, indicating a trend towards flexibility and cost-effectiveness in hardware choices [1][3]. - The preference for Ethernet and PCIe for inference workloads is driven by cost considerations and the ease of capacity expansion, contrasting with the continued use of InfiniBand for training clusters [3][4]. - The call emphasized the importance of co-packaged optics for high bandwidth requirements, particularly for workloads exceeding **1.6T** [3][4]. This comprehensive analysis provides insights into the current trends and future expectations within the AI Datacenter industry, highlighting key shifts in technology, investment strategies, and market dynamics.

Summary of the Conference Call on Data Center & Networking Equipment Industry Overview - The conference focused on the **data center and networking equipment market** and its trends as of October 2025, featuring insights from **650 Group** [1] Key Points and Arguments 1. **Capital Expenditure Growth**: There is a significant increase in companies spending over **$5 billion** in capital expenditures annually, rising from **5 in 2020** to **11 in 2024**, indicating a growing demand for transformative AI technology [5][2] 2. **AI Project Participation**: Major suppliers such as **Dell, SMCI, Fox Conn, and Celestica** are expected to participate in significant AI cluster projects like **Stargate**, focusing on reducing time to first token [5] 3. **Networking Demand**: Companies like **Arista** and **Celestica** are experiencing strong demand for their core data center networking capabilities, with each vendor addressing different areas of the network and customer segments [2][6] 4. **TAM Expansion**: The Total Addressable Market (TAM) is expanding in both **scale-up** and **scale-across** (DCI Ethernet networking), with expectations for new rack designs and liquid-cooled switches at the upcoming **2025 OCP Global Summit** [2][5] 5. **Custom ASIC Adoption**: There is a broadening adoption of custom ASICs, particularly as inference demand grows, with expectations of **12-15 major custom ASIC programs** for companies like **OpenAI, Apple, and Anthropic** [5] 6. **Market Share Dynamics**: **Arista** and **Celestica** hold leading positions in the data center networking category, with their competition being more about addressing different use cases rather than direct competition [6] 7. **Networking Fabric Competition**: The scale-up market is currently dominated by **NVLink**, with **UALink** and **Ethernet** competing for the second position. Ethernet is expected to be well-positioned for scale-up due to its ubiquity and ease of operation [6] Additional Important Insights - The **2025 OCP Global Summit** is generating high investor enthusiasm, with expectations for innovative designs and technologies to be showcased [2] - The report emphasizes the importance of hardware engineering differentiation and software integration capabilities of leading vendors compared to white box vendors [6] This summary encapsulates the critical insights and trends discussed during the conference call, highlighting the growth and competitive dynamics within the data center and networking equipment industry.

Core Insights - Marvell Technology has a strong history of delivering shareholder returns, primarily known for its chips in telecom and consumer sectors since its establishment in 2005 [1] - The company is now focusing on custom ASICs for artificial intelligence (AI), which has become a critical area for growth despite a temporary revenue growth pause earlier this year [2] Company Developments - Marvell has shifted its product offerings to include ASICs that cater to AI and data centers, utilizing advanced process technologies and multi-chip packaging [4] - The company supports chips ranging from 14nm to 3nm and custom AI accelerators, emphasizing power efficiency and rigorous testing methodologies [5] - In May 2023, Marvell partnered with Nvidia to enhance NVLink Fusion technology, integrating connectivity with hardware and software for next-generation AI infrastructure [6] Financial Performance - In the first half of fiscal 2026, Marvell reported $3.9 billion in revenue, marking a 60% increase compared to the same period in fiscal 2025 [7] - Revenue for fiscal Q1 was nearly $1.9 billion, while Q2 was about $2.0 billion, indicating a 6% quarter-over-quarter growth, although growth remains inconsistent [7] - The positive revenue growth led to a turnaround in operating income, resulting in a net income of $373 million for the first half of the fiscal year, a significant improvement from a $409 million loss in the previous year [8]

Core Insights - Marvell Technology's stock has surged approximately 25% since a previous note, driven by a $5 billion stock repurchase authorization and management's optimistic outlook on growth related to AI and data centers [2] - The AI market is experiencing significant investments, with major firms like Amazon and Microsoft projected to spend substantial amounts on AI infrastructure, which could benefit Marvell's specialized chips [5][6] Company Performance - Marvell's stock declined nearly 30% earlier this year due to sporadic orders for its custom AI accelerators, but these fluctuations are common in the sector [3] - The company's data center business has tripled since 2023, with AI-related revenue increasing sixfold, indicating strong growth potential [7] AI Market Dynamics - Marvell's engagement in the AI market focuses on high-speed interconnect solutions essential for data transfer in AI tasks, with application-specific integrated circuits (ASICs) expected to be a major growth driver [4] - The shift from large-scale AI models to smaller, specialized ones may favor niche players like Marvell, which can provide customized solutions [6][7] Valuation Metrics - Marvell is trading at approximately 50 times trailing earnings and about 28 times estimated FY'26 adjusted earnings, which is lower than peers like AMD and Nvidia, suggesting a favorable valuation [8] - The company's balance sheet is robust, with a low debt-to-equity ratio of 7% and a cash-to-assets ratio of 5.9% [8]

Core Insights - The article compares two semiconductor companies, Credo Technology Group Holding Ltd. (CRDO) and Marvell Technology, Inc. (MRVL), both of which are integral to the infrastructure supporting AI-driven data centers [1][2]. Group 1: Company Overview - CRDO specializes in high-speed connectivity solutions, particularly in its AEC product line, which has seen significant growth due to its reliability and lower power consumption compared to optical solutions [4][5]. - Marvell has a diversified product portfolio, focusing on custom ASICs, data center switches, and 5G chips, with a strategic pivot towards the data center market [9][10]. Group 2: Financial Performance - CRDO reported a strong fiscal first quarter, with three hyperscalers contributing over 10% of revenues, and anticipates continued revenue growth from these clients [5][8]. - MRVL's data center revenues surged 69% year over year, with a significant contribution from custom silicon and electro-optics, making data center the largest segment for the company [9][11]. Group 3: Growth Drivers - CRDO's growth is driven by its AEC business, optical solutions, and PCIe retimer products, with expectations of substantial revenue growth in fiscal 2026 [6][7][8]. - MRVL's growth is supported by its custom XPU products and a robust design pipeline, with management targeting a significant increase in market share in the data center space [11][12]. Group 4: Market Position and Valuation - CRDO's stock has increased by 76% over the past three months, while MRVL's stock has declined by 1.7% during the same period [9][16]. - In terms of valuation, CRDO has a forward price/sales ratio of 26.94X, which is higher than MRVL's 7.17X, indicating differing market perceptions [18]. Group 5: Analyst Estimates - Analysts have significantly revised earnings estimates for CRDO upwards, while MRVL has seen only minor revisions [20][21]. - CRDO holds a Zacks Rank of 1 (Strong Buy), while MRVL has a Zacks Rank of 3 (Hold), suggesting a more favorable outlook for CRDO at this time [23].

Group 1: Compute Segment - The compute segment has shown consistent growth, primarily driven by AI applications [1] - Growth is significantly influenced by custom ASICs at vertically integrated producers (VIPs), also known as hyperscalers [1] - Shipments in this segment are expected to be lumpy due to project ramps occurring in Q3, Q4, and Q1 [1] Group 2: Mobility Segment - The mobility segment was primarily influenced by supply chain transitions in the first half of the year [2] - The challenges faced in this segment are described as transitory and not driven by end market demand, but rather by supply chain complexities [2]

Group 1: Company Overview - Advanced Micro Devices (AMD) is leveraging a strong partner ecosystem, including Microsoft, Oracle, Google, Dell, HPE, Lenovo, and IBM, to enhance its presence in the AI infrastructure market, facing competition from NVIDIA and Broadcom [1][10] - AMD's EPYC processors are in high demand for cloud and enterprise workloads, with significant adoption by major cloud hyperscalers [2][10] Group 2: Product Developments - The launch of the Instinct MI350 series has strengthened AMD's system-level capabilities, supporting deployments with AMD CPUs, GPUs, and NICs [3] - AMD has formed a multibillion-dollar collaboration with HUMAIN to develop AI infrastructure powered by its CPUs, GPUs, and software [3] Group 3: Financial Performance - AMD anticipates third-quarter 2025 revenues of $8.7 billion, reflecting a 28% year-over-year increase and 13% sequential growth [5][10] - The Zacks Consensus Estimate for third-quarter 2025 earnings is projected at $1.17 per share, indicating a 21.2% year-over-year growth [11] Group 4: Competitive Landscape - AMD faces tough competition from NVIDIA, which is central to AI computing with its advanced GPUs, and Broadcom, which provides networking chips and custom ASICs for efficient hyperscale data centers [6][7] Group 5: Stock Performance and Valuation - AMD shares have increased by 39.2% year to date, outperforming the broader Zacks Computer and Technology sector and the Zacks Computer Integrated industry [8] - AMD's stock is considered overvalued with a forward 12-month price/sales ratio of 7.46X compared to the sector's 6.7X [9]