US stocks were mixed on Friday as investors digested two days of gains driven by easing geopolitical tensions and turned their attention to corporate earnings and next week's Federal Reserve decision.... ...

Core Viewpoint - Semiconductor stocks in the US showed pre-market strength, with notable gains from AMD, Arm Holdings, Nvidia, and TSMC, indicating positive market sentiment in the sector [1] Group 1: Stock Performance - AMD increased by over 2.9% [1] - Arm Holdings rose by more than 2.7% [1] - Nvidia saw a gain of over 1% [1] - TSMC experienced an increase of more than 1.3% [1] Group 2: Company Developments - Nvidia's founder and CEO, Jensen Huang, visited China, starting with a new office in Shanghai [1] - During the visit, Huang engaged with employees, addressing their concerns and discussing key events for the company in 2025 [1]

Market Analysis by covering: Intel Corporation, Advanced Micro Devices Inc. Read 's Market Analysis on Investing.com ...

Group 1 - Semiconductor stocks in the U.S. showed strength before the market opened on January 23, with AMD rising over 3% [1] - Nvidia experienced an increase of nearly 2% in pre-market trading [1] - Other notable performers included Arm Holdings and TSMC, both of which saw gains of over 1% [1]

为了防个小偷，不得不把家里一半的房间门焊死，这种令人"智熄"的操作，最近却在使用AMD处理器的云服务商身上上演了。 事情的起因是近期德国一家安全机构发现了一个名为StackWarp的芯片级漏洞，这个漏洞主要影响AMD的Zen系列处理器，导致云计算最核心的"隔离能 力"出现了缺口。 什么意思？对于云服务平台而言，作为平台，是无权也无法触碰用户的数据，这就像房东不该进租客的卧室。 如此严重的漏洞，必须马上拿个解决方案出来，AMD给的解决方案很直接，既然是同步多线程(SMT)引起的问题，那把同步多线程禁用不就完了。 这一堪称"甜菜"操作背后，直接导致了算力的缩水，同步多线程一旦禁用，处理器的并行处理能力会大幅下降。 对于云服务商而言，原本能支持100个虚拟服务实例的服务器，在执行防御措施后，支撑能力将直接减半，在算力需求激增的当下，这种折损背后的成本有 多高，不言而喻。 但在中国市场，海光C86处理器的用户不需要担心这个问题。 因为从最底层的架构设计上，海光全系处理器就对StackWarp漏洞天然免疫，用户无需升级固件，无需禁用超线程，更不需要牺牲任何计算性能，服务器可 以保持满负荷运转，业务连续性不受影响。 在由 ...

Intel Corporation (NASDAQ:INTC) shares tumbled over 12% in premarket trading on Friday, January 23, 2026, extending losses from the previous day's after-hours session where the stock had already plunged 14% following the chipmaker's fourth-quarter earnings release. Despite beating Wall Street expectations on both revenue and earnings per share for Q4, investors fled the stock after Intel issued a dismal first-quarter forecast and CEO Lip-Bu Tan warned of persistent manufacturing problems. The weak guidance ...

Core Viewpoint - The article emphasizes the booming demand for server CPUs driven by major cloud service providers, leading to a supply-demand imbalance that has resulted in price increases by Intel and AMD [3][4][6]. Group 1: Demand Drivers - The demand for server CPUs is likened to a race car suddenly hitting the gas, fueled by three main factors: the hardware upgrade cycle among global cloud giants, the cost-effectiveness of CPUs for AI inference, and bulk purchasing by cloud service providers fearing future supply shortages [4][5]. - Major cloud service providers like Google, Amazon, and Alibaba are in a phase of upgrading their data centers, replacing old CPUs with new models such as AMD's EPYC "Turin" and Intel's Xeon "Granite Rapids," which offer more cores and higher efficiency [4]. - In the context of AI applications, CPUs are becoming the preferred choice for high-concurrency, lightweight inference tasks due to their cost advantages and software compatibility, complementing GPUs that handle model training [4]. Group 2: Supply Constraints - The supply side is facing challenges, particularly with high-end manufacturing capacity being fully booked, as advanced nodes (3nm/5nm) are in high demand from various tech companies, limiting the availability for server CPUs [6]. - The role of CPUs in complex AI server configurations is critical, as they manage task scheduling and memory, making them indispensable for efficient operation alongside powerful GPUs [6]. Group 3: Investment Insights - Investors are advised to focus on Intel and AMD as direct beneficiaries of the price hikes, monitoring their capacity planning and pricing strategies in upcoming financial reports [7]. - The tight supply from international giants presents an opportunity for domestic CPU manufacturers to gain market share, particularly in sectors like government and finance [7]. - The increase in server demand will also benefit the supply chain, including testing and packaging sectors, as well as semiconductor equipment and materials companies that may see growth from potential capacity expansions [7]. Group 4: Future Outlook - Continuous monitoring of AI model iterations and capital expenditure plans from global cloud providers is essential, as these will serve as indicators of sustained demand for server CPUs [8]. - The supply-demand imbalance in server CPUs reflects the broader upgrade wave in computing infrastructure driven by the AI era, highlighting the importance of high-end computing needs, domestic alternatives, and the ripple effects across the industry [8].

全球半导体市场出现结构性变化，传统意义上被视为成熟品类的CPU领域，正成为资本市场的关注焦点。 截至1月23日，英特尔股价一度创下近四年新高，年内涨幅近38%，AMD延续连续上涨态势；A股市场中，龙芯中科与海光信息分别录得20%涨停及单日涨幅 超13%。 这一行情背后 ，反映的是市场对"算力税"传导逻辑的重新定价，继GPU因AI训练需求大涨之后，CPU正在成为算力成本上涨的第二波承载者。 机构的共识正在迅速形成。国联民生证券与西部证券在近期报告中不约而同地指出， 当前CPU市场的供需变化并非周期性波动，而是由AI智能体应用规模化落 地所驱动的结构性变革。 与以GPU为核心的AI训练不同，在智能体工作负载中 ，CPU承担了包括工具调用、任务编排、实时决策等大量非AI原生计算，相关处理耗时占总任务延迟的 比例高达80%-90%。这使得CPU在系统层面可能比GPU更早成为性能瓶颈。 在此趋势下，国产CPU厂商迎来产业与政策的双重机遇。CPU，这一曾长期被视为"传统"的算力组件，正在AI智能体浪潮中重新确立其系统级价值。 AI智能体催化"域外CPU"需求重塑 传统AI计算将算力重心完全置于GPU之上，主要用于模型的训 ...

AI代码平台Claude Code在半小时内完成了英伟达CUDA代码向AMD ROCm平台的移植，展示了生成式AI在打破计算生态壁垒方面的潜力。 1月22日，一位用户在社交平台Reddit上披露，他使用Claude Code将整个CUDA后端移植至AMD的ROCm平台，且无需中间转换层。 这一案例引发市场关注，部分人士认为这可能削弱英伟达长期以来依赖CUDA构建的技术护城河。 不过，业内人士指出，这一成果可能仅适用于较简单的内核代码。对于需要深度硬件优化和复杂上下文的代码库，AI工具的移植能力仍面临显著局限。 英伟达的CUDA平台长期主导AI计算领域，其生态系统的封闭性使得开发者难以将应用迁移至竞争对手AMD的ROCm平台，这也是英伟达维持市场优势的关键 因素之一。 智能代理实现快速移植 业内人士认为，真正的挑战在于相互关联的复杂代码库。 此类移植需要智能代理系统理解大量上下文信息才能有效完成向ROCm的转换。 更关键的是，编写内核代码的核心在于实现深度硬件优化。 有观点指出，Claude Code在针对特定缓存层级等硬件细节的优化方面仍力有不逮，这限制了其在 高性能计算场景中的实用性。 据用户johnnyt ...

Core Insights - The AI code platform Claude Code has successfully migrated NVIDIA's CUDA code to AMD's ROCm platform within half an hour, showcasing the potential of generative AI in breaking down computational ecosystem barriers [1] - This case has raised market concerns that it may weaken NVIDIA's long-standing technological moat built around CUDA [3] Group 1: Migration Process - A user reported that the only issue encountered during the migration was the "data layout" difference [4] - Claude Code operates using an intelligent agent framework that can smartly replace CUDA keywords with their ROCm counterparts while maintaining the underlying logic of specific kernels [4] - The tool simplifies the migration process, allowing developers to complete the work directly through a command-line interface without needing to configure complex conversion environments like Hipify [4] Group 2: Limitations and Challenges - The user did not specify the exact type of codebase processed, but the design of ROCm mimics several aspects of NVIDIA's CUDA platform, making simple code migration relatively easy for AI tools [5] - Industry experts believe that the real challenge lies in migrating complex, interrelated codebases, which require the intelligent agent system to understand a significant amount of contextual information for effective conversion to ROCm [5] - There are concerns that Claude Code may struggle with deep hardware optimizations, particularly regarding specific cache hierarchies, limiting its practicality in high-performance computing scenarios [6]