当地时间8月25日，被英伟达称为机器人"新大脑"的Thor芯片，以开发者套件的形式正式发售，售价为 3499美元（约2.5万元人民币）。这款芯片旨在支撑具身智能机器人与物理世界进行实时、智能的交 互。 "机器人需要丰富的传感器数据和低延迟的 AI 处理来支撑实时运行，这必须依靠强大的AI算力和内存 来并行处理来自多个传感器的数据。"英伟达表示。 具身智能机器人面临的算力不足问题已引发业内关注。英伟达上一代Orin芯片被众多机器人厂商使用， 但其最高算力仅为275Tops（每秒275万亿次运算）。在一些业内人士看来，如果机器人在场景中自主运 行，这一算力规模远远不够。 官方数据显示，基于Blackwell架构GPU的新一代Thor芯片，在FP4精度下提供高达2070 TFLOPS（每秒 2070万亿次浮点计算）的峰值算力，比上一代Orin芯片提升了7.5倍，能效（单位功耗下的算力输出） 提升3.5倍。此外，其CPU性能比上代提升了3.1倍，内存128GB，提升了2倍。 自2014年以来，英伟达快速绑定机器人领域的开发者。公司8月25日披露，目前已经有超过200万开发者 加入英伟达机器人技术生态，覆盖制造业、食品 ...

Two and a half years into the AI boom and things are still progressing at a torrid pace. Though markets experienced an especially sharp correction from March-May, much of the volatility stemmed from politically driven tariff headlines. With those policies now dialed back to more sustainable levels, growth forecasts have stabilized, and the AI trade is back in focus.What’s remarkable is that for the first time in modern history, the pace of technological progress is being constrained not by ideas, but by phy ...

Core Insights - The humanoid robot market is projected to reach a record $1.2 billion in funding in 2024, with expectations to double to $2.3 billion by 2025, indicating rapid growth [1][2]. Group 1: Market Trends and Investment - The funding scale and valuations in the humanoid robot sector are rising, with a notable concentration of capital among leading companies. In 2024, industrial humanoid robots saw a funding surge to $904 million across 40 transactions, with 2025 already raising $644 million through just 12 deals [2][5]. - The top ten companies have collectively raised nearly $7.2 billion, with Meta, Xiaomi, UBTECH, and Tesla capturing around 60% of the total funding [2][12]. - The emergence of Physical AI is redefining the balance between hardware scarcity and software proliferation, with companies like Figure and Skild AI achieving valuations of $2.7 billion and $1.5 billion, respectively, within three years of establishment [2][18]. Group 2: Competitive Landscape - The competition between the US and China in the humanoid robot market has evolved from algorithmic leadership versus manufacturing scale to capital dominance versus price wars. US companies hold 32% of the global market, while Chinese firms account for 27% [23][27]. - US manufacturers are focusing on scaling production, with companies like Figure and Agility Robotics planning to build factories capable of producing over 10,000 humanoid robots annually [26][27]. - Chinese manufacturers are leveraging competitive pricing strategies, with prices ranging from $13,700 to $27,500, but face challenges in reliability and brand trust [25][31]. Group 3: Technological Developments - Major tech companies are laying the groundwork for humanoid robots through foundational models that enhance robots' capabilities in perception, decision-making, and task execution [41][42]. - Companies like Google, Nvidia, and Apple are developing advanced AI systems to support humanoid robots, emphasizing the importance of computational power and data integration [41][46]. - The market is witnessing a bifurcation between open-source collaboration and proprietary systems, with companies adopting different strategies to enhance their competitive edge [30][34]. Group 4: Commercialization and Deployment - The deployment of humanoid robots is shifting from merely replacing human labor to embedding them in collaborative environments, as seen in partnerships with companies like BMW and Mercedes-Benz [36][38]. - The integration of humanoid robots into existing automation ecosystems is expected to provide advantages in cost, flexibility, and modular scalability [39][40]. - The anticipated widespread adoption of humanoid robots in various sectors, including industrial, retail, and healthcare, is projected to occur within the next decade, fundamentally altering labor dynamics [28][29].

Investment Rating - The investment rating for NVIDIA Corp. is "Overweight" with a price target of $160.00, while the stock closed at $135.40 on May 16, 2025 [4]. Core Insights - NVIDIA announced NVlink Fusion, enhancing interconnect technology for a broader ecosystem, which could boost communication infrastructure revenues [2][6]. - The company introduced RTX PRO servers aimed at the enterprise AI inference market, featuring advanced graphics cards and networking technologies [2]. - A new partnership with Foxconn and the Taiwan government was established to develop a supercomputer with 10,000 Blackwell GPUs, indicating strong R&D collaboration [2]. - Despite challenges from the US Commerce Department's ban on H20 in China, which poses a $5 billion headwind, there is a clear path for re-acceleration in the second half of the year [2][6]. Summary by Sections Product Developments - NVlink Fusion extends capabilities for custom ASICs and non-NVIDIA CPUs, potentially enhancing competition and customer choice [2]. - RTX PRO servers are designed for enterprise AI inference, featuring up to 8 Blackwell RTX Pro graphics cards [2]. - The DGX Spark and DGX Station workstations are nearing customer availability, with Spark launching in July 2025 [2]. Market Position and Challenges - The report acknowledges near-term challenges but emphasizes a strong potential for growth in the second half of the year [2][6]. - Intermediate-term concerns regarding cloud vendor support and supply chain bottlenecks are being addressed [2]. Financial Metrics - NVIDIA's market capitalization is approximately $3.39 billion, with a 52-week price range of $86.62 to $153.13 [4].

Core Insights - NVIDIA's CEO Jensen Huang presented a grand vision for the future of artificial intelligence (AI) at Computex 2025, showcasing new products and technology plans aimed at reshaping the tech ecosystem from cloud to edge and virtual to reality [1] - Huang emphasized that NVIDIA is transitioning from being a tech company to a crucial AI infrastructure company, defining its operations as an "AI factory" that produces valuable outputs known as tokens [2][3] - The chip industry is valued at $300 billion, while the data center opportunity is evolving into a nearly $1 trillion market, driven by the concept of "AI factories" and infrastructure [5] Product Launches - NVIDIA introduced the RTX 5060 GPU and a new MSI laptop equipped with it, set to launch in May [5] - The company announced the Grace Blackwell GB300 system, designed for AI inference performance, featuring 72 NVIDIA Blackwell Ultra GPUs and 36 ARM-based Grace CPUs, with significant performance improvements [7][8] - The new RTX Pro AI platform was launched for enterprises, offering 30 PFLOPS AI computing power and supporting complex AI model training and inference tasks [9] Strategic Developments - NVIDIA is establishing a new headquarters in Taipei, Taiwan, symbolizing its commitment to building a global AI ecosystem [12] - Huang highlighted the importance of robotics in the future of AI, stating that all mobile devices will become robots, leading to an industrial revolution [12][14] - The company is investing in a robot simulation training platform, Isaac GR00T, and has partnered with various companies to advance its integrated hardware and software strategy [14]

Core Insights - NVIDIA is entering a new phase of development as a leading provider of AI infrastructure, with its market value continuing to rise [1] - The company announced the launch of new technologies and products, including the NVIDIA NVLink Fusion chip, during the Computex 2025 event [1][2] - NVIDIA is transitioning from a single hardware supplier to an open ecosystem platform by allowing third-party access to NVLink IP [1][2] Group 1: New Technologies and Products - The NVLink Fusion technology enhances ecosystem compatibility, enabling integration with CPUs from Fujitsu and Qualcomm to build high-performance AI infrastructures [1][2] - NVIDIA introduced the RTX PRO server for enterprise-level AI inference, capable of supporting up to 8 Blackwell RTX Pro Graphics 6000 cards [2] - The company updated its robot foundational model, Isaac GR00T, and introduced a synthetic data generation framework for humanoid robot training [2][4] Group 2: Partnerships and Collaborations - Initial adopters of NVLink Fusion include MediaTek, Marvell, and others, but no Chinese companies are among the first group [2] - NVIDIA is collaborating with Foxconn and Taiwan partners to build a supercomputer with 10,000 Blackwell GPUs, with TSMC as a primary customer [5][6] - The company plans to establish a new office in Taiwan to strengthen its ecosystem partnerships [5][6] Group 3: Market Presence and Future Plans - NVIDIA's revenue in China is projected to be approximately $17 billion in 2024, accounting for about 14% of its global total [6] - The company is leasing new office space in Shanghai to accommodate current employees and prepare for future expansion [6]