Core Viewpoint - The liquid cooling industry is expected to experience rapid demand growth starting in 2026, while supply will lag behind, creating a highly favorable market environment for early movers in the sector [17]. Demand Side - As the power consumption of AI computing cards and cabinets increases, traditional air cooling solutions become inadequate, necessitating the adoption of liquid cooling solutions for effective heat dissipation. Specifically, cabinets with power consumption exceeding 35-40KW cannot utilize air cooling [4]. - North America faces severe electricity shortages, and liquid cooling can significantly reduce Power Usage Effectiveness (PUE), thereby saving energy costs and alleviating delays in data center projects caused by power shortages. Liquid cooling solutions can achieve a PUE of less than 1.2 [4]. - Many data centers in North America are located near economic centers or residential areas, and liquid cooling can significantly reduce noise levels compared to air cooling, accelerating project implementation and reducing financial costs [6]. Supply Side - The supply landscape for liquid cooling in North America is currently dominated by American and Taiwanese manufacturers, with limited participation from mainland Chinese firms. The primary technology used is cold plate liquid cooling, which accounts for over 98% of the market [11]. - The verification process for liquid cooling products is lengthy and complex, creating high barriers to entry for new suppliers. This process can take anywhere from six months to two years, making it challenging for many manufacturers to enter the market [11]. Company-Specific Insights - NVIDIA is projected to ship 100,000 cabinets of the GB series (primarily GB300) by 2026, with a liquid cooling value per cabinet estimated at $90,000 to $100,000, leading to a total liquid cooling value of approximately $10 billion [7]. - Google is expected to ship 2.2 to 2.3 million TPU V7 and above chips by 2026, translating to a need for around 35,000 cabinets with liquid cooling solutions, with a total liquid cooling value estimated at $2.6 billion [9]. - Meta anticipates shipping 1 million MTIA V2 chips, requiring about 14,000 cabinets, with a projected liquid cooling value of approximately $1.05 billion [10]. - Amazon's AWS Trainium3 is expected to ship 1.2 million chips, corresponding to around 16,000 cabinets, with an estimated liquid cooling value of $1.25 billion [10]. Market Opportunities for Mainland Chinese Manufacturers - North American CSP technology companies are increasingly looking to mainland Chinese liquid cooling manufacturers to ensure supply chain security, as existing production capacities in North America and Taiwan may not meet the surging demand by 2026 [13]. - Mainland Chinese manufacturers can offer liquid cooling products at a lower cost compared to their foreign counterparts, potentially enhancing project profitability for data centers [13]. - Some leading mainland Chinese manufacturers possess competitive technological advantages in liquid cooling products, which can improve the efficiency of downstream data center products [13]. Future Outlook - The liquid cooling industry is expected to see a significant increase in demand starting in 2026, with early adopters benefiting from the first wave of industry growth. The process of engaging with overseas manufacturers, obtaining samples, and securing initial orders will be crucial for companies looking to capitalize on this trend [17].

Summary of NVIDIA's Optical Interconnection Developments Company Overview - **Company**: NVIDIA (Ticker: NVDA US) - **Industry**: AI Networking and Optical Interconnection Key Points Industry Developments - NVIDIA is expected to incorporate CPO (Co-Packaged Optics) into its 2027 Rubin Ultra architecture, following Google's adoption of OCS (Optical Circuit Switching) for its TPU V7, which interconnects over 9,000 chips, surpassing NVIDIA's projected deployment of 576 dies in 2027 [1][2] - The transition to CPO for rack-to-rack interconnects is anticipated to enhance power consumption, latency, density, and cost efficiency compared to AOC (Active Optical Cable) [2] Product and Technology Insights - NVIDIA's scale-up optical solutions may arrive sooner than expected, with CPO being considered for the 576-die architecture starting in the second half of 2027 [2] - Compute trays and switch trays will continue to use PCB backplane connectivity, while rack-to-rack interconnects are likely to adopt CPO-based optical interconnects [2] - The scale-up CPO presents an incremental opportunity for the optical interconnect supply chain, with key beneficiaries including companies like LITE, Sumitomo, and Browave [3] Market Expectations - NVIDIA's scale-out CPO switch is projected to have deployment figures of 2,000, 20,000, and 35,000 units in 2025, 2026, and 2027 respectively [4] - The anticipated demand for NVIDIA's OIO (Optical Interconnection) solution is expected to coincide with the Feynman architecture, driving demand for CW lasers, FAUs, and optical engines [4] Risks - Potential risks include deceleration in AI demand, geopolitical uncertainties, and increased competition within the industry [5] Rating and Performance Expectations - NVIDIA is rated as a "Buy," indicating an expectation to outperform the benchmark by more than 15% [6] Additional Considerations - The report emphasizes that the CPO penetration rate may not be significant in the 1.6T generation due to factors such as maturity and technical reliability [4] - The supply chain for scale-up CPO is similar to that of scale-out CPO, indicating a consistent market structure [3] This summary encapsulates the critical insights and projections regarding NVIDIA's advancements in optical interconnection technology and its implications for the industry.

Summary of ASIC Industry Update Industry Overview - The document provides an update on the ASIC (Application-Specific Integrated Circuit) projects of major North American hyperscaler companies, including AWS, Microsoft, Meta, Google, OpenAI, Apple, and TikTok [2][3] Key Companies and Their ASIC Projects AWS (Amazon Web Services) - **Tranium 2 Chip**: Expected to reach its end phase in Q4 2025, with a transitional chip, Tranium 2.5, to be produced in Q4 2025 and Q1 2026. Marvell is expected to ship approximately 200,000 units per quarter [4][3] - **Tranium 3 Chip**: Forecasted production volume of around 2.5 million units, with potential allocation of up to 500,000 units to Marvell if Tranium 2.5 production is successful [8][9] - **Tranium 4 Chip**: Designed by Annapurna and Alchip, expected to start mass production in Q4 2027 [9][10] Microsoft - **Cobalt 200 CPU and MAIA 200 Sphinx**: Designed by GUC, with MAIA 300 Griffin facing challenges in its development with Marvell. Microsoft may shift to Broadcom if confidence in Marvell wanes [14][16] - **MAIA 200 and MAIA 300**: Part of the second-generation ASIC accelerator series, with the contract with Marvell expiring in H1 2026 [15][16] Meta - **ASIC Roadmap**: Includes multiple generations of chips, with the first-generation inference chip, Artemis, already in mass production. The second-generation training chip, Athena, is set for Q4 2023, and the third-generation chip, Iris, is planned for Q3 2024 [17][18] - **Arke Chip**: A simplified inference-only chip designed by Broadcom and Marvell, expected to help Meta keep pace with NVIDIA's chip iterations [19][20] Google - **TPU Development**: The first-generation ASIC Server CPU, Axion, is designed by Marvell, while the second-generation, Tamar, is designed by GUC. Google expects to produce about 4 million TPUs in 2026, with significant internal use [22][24] - **Demand Surge for Optical Modules**: Due to the increase in TPU production, demand for 1.6T optical modules is expected to rise dramatically from 3 million units in 2025 to 20 million in 2026 [25][26] OpenAI - **Titan 1 and Titan 2 Chips**: Broadcom is developing these chips, with expected shipments of 300,000 units in 2026 and at least 600,000 units in 2027 [28][29] - **Collaboration with ARM**: OpenAI is also working with ARM on ASIC projects, indicating a dual approach to chip development [30][31] Apple - **ASIC Projects**: Apple is customizing two ASIC chips, with mass production not expected before 2027 [32][33] TikTok - **Neptune Chip**: After negotiations, TikTok is expected to resume mass production of its ASIC chip in Q1 2026, with an anticipated production volume of 500,000 units [34][35] GUC (Global Unichip Corp) - **Controversial Position**: GUC is involved in the production of Google's Tamar CPU but is also engaged in more profitable projects like Tesla's AI5 chip, which could generate significant revenue in 2027 [41][43] Additional Insights - The document highlights the competitive landscape among major players in the ASIC market, with companies like Marvell, GUC, and Broadcom playing crucial roles in the design and production of these chips [41][42] - The anticipated growth in demand for ASIC chips, particularly in the context of AI and machine learning applications, suggests a robust market outlook for the coming years [25][26] This summary encapsulates the key developments and projections within the ASIC industry, focusing on the major players and their respective projects.