News Summary: The Rubin platform harnesses extreme codesign across hardware and software to deliver up to 10x reduction in inference token cost and 4x reduction in number of GPUs to train MoE models, compared with the NVIDIA Blackwell platform.NVIDIA Spectrum-X Ethernet Photonics switch systems deliver 5x improved power efficiency and uptime.New NVIDIA Inference Context Memory Storage Platform with NVIDIA BlueField-4 storage processor to accelerate agentic AI reasoning.Microsoft’s next-generation Fairwater ...

@ TVIDIA "Everything Everywhere All at Once" AI Scales Beyond LLMs 80% of Startups Building on Open Models 1-in-4 OpenRouter Tokens Reasoning Generated by Open Models 2024 2025 Clara Earth-2 AI-Physics Model Open Models Kimi K2 - DeepSeek V3.2 Reach the Frontier ● ● - GPT-OSS 0 0 - Qwen Frontier Models . ● Nemotron ● . Open Models ● Jan 2025 Jan 2026 Jan 2023 Jan 2024 Release Date | | 160 M | | --- | --- | | | 140 M | | | 120 M | | | 100 M | | Monthly Hugging Face Downloads | 80 M | | | 60 M | | | 40 M | | ...

Group 1 - TSMC has solidified its leading position in the global foundry market after inventory adjustments at the end of 2022, with high utilization rates in advanced process technologies [1] - The 3nm process has achieved full capacity utilization in its fifth quarter of mass production, driven by strong demand for Apple A17 Pro/A18 Pro chips and other application processors, setting a new record for initial market demand [1] - Future growth is expected to continue due to the introduction of NVIDIA Rubin GPUs and specialized AI chips from Google and AWS, driven by increasing demand in AI and high-performance computing (HPC) applications [1] Group 2 - In contrast, the smartphone market has seen slower initial capacity growth for existing processes like 7/6nm and 5/4nm, with the latter experiencing a resurgence in 2023 due to surging demand for AI acceleration chips [2] - The demand for AI computing chips is accelerating the construction of AI data centers and significantly enhancing the overall capacity of the 5/4nm process [2] Group 3 - The 2nm process is projected to achieve full capacity utilization in its fourth quarter of mass production, driven by dual demand from smartphones and AI applications, aligning with TSMC's strategic outlook [5] - Potential customers for the 2nm technology include Qualcomm, MediaTek, Intel, and AMD, which is expected to maintain high utilization rates for the 2nm process [5] Group 4 - TSMC is investing $165 billion in its Arizona facility to meet growing U.S. consumer demand and mitigate geopolitical risks, with the facility covering 4nm, 3nm, and 2nm processes [11] - The dual-layout strategy enhances TSMC's geopolitical resilience and ensures capacity meets customer demand, particularly in AI and HPC, while maintaining high utilization rates for advanced processes beyond 2030 [11]

Core Viewpoint - TSMC has solidified its leading position in the global foundry market following inventory adjustments at the end of 2022, with high utilization rates in advanced process nodes showcasing its technological superiority [1][4]. Group 1: Advanced Process Utilization - The 3nm process node has achieved full utilization within five quarters of mass production, driven by strong demand for Apple A17 Pro/A18 Pro chips and other application processors, setting a new record for initial market demand in advanced processes [1]. - TSMC's 5/4nm process is experiencing a resurgence in demand, particularly due to the surge in AI accelerator chips like NVIDIA's H100 and B100, which has significantly boosted overall capacity [2][4]. - TSMC's advanced process utilization rates are projected to remain high, with expectations that the 2nm process will reach full capacity within four quarters of mass production, driven by dual demand from smartphones and AI applications [7]. Group 2: Future Developments and Investments - TSMC plans to allocate 30% of its 2nm process capacity to its Arizona facility, enhancing geopolitical resilience while ensuring capacity meets customer demand, especially in AI and high-performance computing [9]. - The company anticipates that the diverse customer base for the 2nm technology, including major players like Qualcomm, MediaTek, Intel, and AMD, will help maintain high utilization rates [7]. - TSMC's investment of $165 billion in its Arizona plant will support advanced process technologies, including 4nm, 3nm, and 2nm, ensuring the company remains at the forefront of the semiconductor industry [9].