Arista Networks FY Conference Summary Company Overview - **Company**: Arista Networks (NYSE: ANET) - **Date of Conference**: February 25, 2026 - **Key Speakers**: John McCool (Chief Platform Officer), Rudolph (Head of Investor Advocacy), Mark Newman (Bernstein Analyst), Daniel Zhu (Bernstein Networking Analyst) Industry Insights AI Opportunity - **Ethernet's Role**: Ethernet is crucial in the AI landscape, particularly in connecting back-end networks and GPUs. Significant progress has been made in standardizing these networks [4][5] - **Network Size and Efficiency**: Networks are expanding, necessitating innovations to maintain efficiency. Arista has a history of driving performance improvements [5] - **Front-end vs. Back-end Networks**: The distinction between front-end and back-end networks is blurring, with Arista's products being versatile enough to operate in both environments [6][7] Scale Opportunities - **Scale-Up Networking**: Currently limited due to proprietary protocols like NVLink from NVIDIA. Expected to become an opportunity for Arista around 2027 [8] - **Scale-Out and Scale-Across**: Scale-out networking is where most AI revenue is generated, connecting multiple racks and buildings. Scale-across involves interconnecting data centers, presenting unique challenges and opportunities for Arista [9][10] Network Architecture - **Shift in AI Model Training**: The transition from pre-training to post-training and inferencing impacts network architecture, requiring broader storage and more machines [14][15] - **Complexity as an Advantage**: The increasing complexity of networks necessitates higher R&D and engineering, which plays to Arista's strengths [16] Technology Roadmap Transition to Higher Speeds - **Speed Transition**: The transition from 400G to 800G has been compressed, with expectations for even faster transitions to 1.6T [25][26] - **Co-Packaged Optics**: While co-packaged optics are seen as a future trend, Arista is currently able to maintain performance with existing technologies [28][29] Optical Circuit Switching - **Emerging Technology**: Optical circuit switching is gaining traction but remains niche compared to Ethernet switching. Arista sees limited competition from this technology [34][37] Market Dynamics Campus Networking - **Growth Target**: Arista aims for $1.25 billion in campus networking revenue, leveraging low market share to capture growth [56][57] - **Market Discontinuity**: The merger of HPE and Juniper presents opportunities for Arista as customers seek alternative sources [59][62] Memory Supply and Pricing - **Memory Constraints**: Arista is not currently memory constrained and has made adjustments to pricing structures to address rising costs [71][72] - **Impact of Memory Pricing**: Memory is a significant part of the BOM, but Arista's complex switches mitigate the impact of rising memory prices [70][72] Key Takeaways - **Value Proposition**: Arista's ability to provide a unified operating system across various networking needs is a strong selling point [95] - **Market Perception**: There is a need for better communication of the complexities and advantages of Arista's networking solutions to investors [97] Conclusion Arista Networks is well-positioned to capitalize on the growing AI and networking markets, leveraging its technological strengths and addressing emerging challenges in power consumption and memory supply. The company aims to expand its market share in campus networking while maintaining a focus on innovation and efficiency.

Summary of Credo Technology Group FY Conference Call Industry and Company Overview - **Company**: Credo Technology Group (NasdaqGS:CRDO) - **Industry**: Technology, specifically focusing on AI infrastructure and connectivity solutions Key Points and Arguments AI Investment Cycle - The current AI investment cycle is characterized by over $3 trillion in announced spending, indicating that the industry is in the early stages of a significant transformation that will last over a decade [3][8][9] Product Focus and Reliability - Credo has shifted its product focus towards enhancing reliability, particularly for AI clusters that utilize large numbers of GPUs [8][9][11] - The company emphasizes the importance of maintaining a reliable link between GPUs and switches to prevent productivity losses [9][10] Customer Diversification and Competitive Position - Credo has successfully diversified its customer base, moving from one large customer to multiple significant clients, which has raised questions about sustainability and competition [12][13] - The company believes its unique approach of taking complete ownership of the system—from design to production—sets it apart from competitors [13][14][15] Innovation and System-Level Solutions - Credo is focused on delivering differentiated solutions that go beyond standard offerings, allowing customers to innovate [20][22] - The company is developing a system-level solution that provides real-time telemetry data to prevent link failures, enhancing reliability [41][42] Co-Packaged Optics (CPO) and Future Technologies - The industry is moving towards co-packaged optics, but Credo believes that existing technologies will continue to be effective for the foreseeable future [31][35] - Credo is exploring alternatives like ALCs (Active Optical Cables) that offer equal reliability and power efficiency compared to traditional copper solutions [38][39] OmniConnect Portfolio - The OmniConnect portfolio aims to address bandwidth limitations between CPUs and memory, particularly for AI applications that require high memory capacity [46][48] - The first product, Weaver, is designed to overcome physical limitations in memory placement, allowing for greater capacity and bandwidth [49][50] Gross Margin Expectations - Credo expects long-term gross margins to remain within the range of 63%-65%, driven by the introduction of new products and increasing scale [55][56] Additional Important Insights - The company has seen a year-over-year increase in gross margins, attributed to scaling operations and product innovation [55][56] - Credo's competitive strategy includes being first to market with new solutions and maintaining flexibility to adapt to customer needs [14][15] This summary encapsulates the critical insights from the conference call, highlighting the company's strategic focus, competitive advantages, and market outlook.

Core Insights - Broadcom and OpenAI have initiated a multi-year partnership to deploy 10 gigawatts of custom AI silicon and rack-level systems, enhancing performance and scalability of next-generation technologies [1][3][6] Company Developments - JP Morgan analyst Harlan Sur maintains an Overweight rating on Broadcom with a price target of $400, while Goldman Sachs analyst James Schneider has reiterated a Buy rating and increased the price target from $360 to $380 [2][6] - The partnership with OpenAI allows Broadcom to integrate compute Application-Specific Integrated Circuits (ASICs) with its Ethernet, PCIe, and optical connectivity solutions, extending beyond compute chips to networking silicon and server racks [1][3][6] Financial Projections - Each gigawatt of capacity is estimated to generate $25–$30 billion in revenue for Broadcom, leading to a total potential revenue of $250–$300 billion for the 10 GW infrastructure [4] - Deployment is expected to start in the second half of 2026 and be completed by 2029, with initial annual revenue projected between $70 billion and $90 billion over four years [5] - Goldman Sachs estimates that each gigawatt of AI datacenter deployment could yield $10–$15 billion in incremental revenue and contribute approximately $1.00–$1.50 in EPS for Broadcom [7] Market Position - Broadcom is positioned as a leader in custom silicon and networking solutions, with a strong competitive stance against companies like Advanced Micro Devices (AMD), which has a 6GW partnership with OpenAI [6][7] - The announcement indicates a growing trend towards custom silicon in the accelerator market, reinforcing Broadcom's leadership in custom ASIC and Ethernet networking [8]