2025年，人类在探索未知的征程中留下了坚实的足迹。从"深度求索"用算法点亮AI效率革命，到谷 歌量子计算首次实现可验证的算力超越；从鲁宾天文台以巨眸凝视宇宙起源，到跨物种脑图谱揭开神经 发育的奥秘……科技的力量既指向星辰大海，也照进生命精微深处。 在本报评选的十大国际科技新闻展示的图景中，有深沉的警醒，也有科技的温度：首个气候临界点 的到达警示着气候的脆弱，脑科学与手术机器人的突破则承载着对个体健康的深切关怀。科技的真谛， 从来不仅是追求极致，更是守护共生，是智慧与责任同行。我们相信，科技的内核永远是温暖的，引领 人类走向更坚韧、更包容的未来。 "深度求索"引领开源人工智能模式 在人工智能（AI）浪潮奔涌的2025年伊始，中国"深度求索"（DeepSeek）公司以其开源大模型 DeepSeek-R1的突破性成果，在世界AI发展史上留下了深刻印记。 DeepSeek-R1开创性地采用纯强化学习训练大规模推理模型，在提升模型能力的同时，显著降低了 对标注数据的依赖。该团队今年9月在《自然》杂志发表的论文，系统阐述了这一技术路径的科学价值 与工程实现。其实现了在有限算力条件下达到顶尖性能的目标，展现了中国科研人员 ...

入选的2025年国内十大科技新闻分别是：国产人工智能大模型DeepSeek引发全球关注；中国"人造 太阳"创造"亿度千秒"世界纪录；超导量子计算原型机"祖冲之三号"成功构建；国际上首次制备大面积 二维金属材料；我国首例侵入式脑机接口临床试验开展；月球背面演化历史首次揭开；单个体细 胞"变"完整植株奥秘揭示；高精度可扩展模拟矩阵计算芯片研制成功；四中全会《建议》突出科技创新 引领作用；我国第一艘电磁弹射型航空母舰入列。 科技日报北京12月24日电 （记者陆成宽）24日，由科技日报社主办、部分两院院士和媒体负责人 共同评选的2025年国内、国际十大科技新闻揭晓。 入选的2025年国际十大科技新闻分别是："深度求索"引领开源人工智能模式；脑机接口将脑活动解 码为连续语句；人工智能"从零开始"设计蛋白酶；光子计算芯片性能超越传统电子硬件；鲁宾天文台发 布宇宙摄影"首作"；智能机器人自主完成胆囊切除手术；新算法取得首个可验证量子优势；引力波信号 验证霍金黑洞面积定理；地球达到首个灾难性气候临界点；跨物种哺乳动物脑细胞发育图绘成。 ...

公众号记得加星标⭐️，第一时间看推送不会错过。 在莱布尼茨超级计算中心安装第一个光子处理器一年后，Q.ANT 带着更加雄心勃勃的计划卷土重 来。 该公司推出了第二代原生处理单元（NPU：Native Processing Unit），这款设备的设计目的不仅在 于加速人工智能工作负载，更在于重新思考其背后的数学和能量模型。第一代NPU证明了光可以作为 一种实用的计算介质，而第二代NPU则展现了光子架构一旦摆脱晶体管尺寸限制后，能够以惊人的速 度发展。 此次第二代NPU 不再以独立PCIe卡的形式推出，而是封装成一个完整的19英寸机架式服务器。每个 机箱都包含多个第二代NPU、一个集成x86主机处理器和一个Linux操作系统。 在短短两年内，Q.ANT 的 NPU 的性能进步就赶上了标准数字处理器过去 30 年的性能进步 Q.ANT 的 C、C++ 和 Python API 与现有的高性能计算 (HPC) 和人工智能 (AI) 框架保持兼容，同 时其光子算法库 (Q.PAL) 提供针对光学硬件优化的非线性函数。对于开发者而言，关键的转变不在 于接口，而在于底层计算模型：以往需要庞大而复杂的架构才能实现的算法，现在 ...

Core Viewpoint - The article discusses the development of a novel all-optical chip named LightGen, which addresses the significant computational power shortage faced by large-scale generative artificial intelligence, particularly in visual generation tasks [1][4]. Group 1: Research Development - A research team led by Assistant Professor Chen Yitong from Shanghai Jiao Tong University published a paper in the journal Science on December 18, 2025, detailing their work on the LightGen chip [2]. - The LightGen chip is designed for large-scale intelligent semantic vision generation and integrates millions of photonic neurons to achieve high-resolution image generation, denoising, style transfer, and 3D generation and manipulation [3][4]. Group 2: Performance and Efficiency - Experimental results indicate that the LightGen chip's end-to-end computational speed and energy efficiency exceed that of the most advanced electronic chips by more than two orders of magnitude, paving the way for the advancement of large visual generative models [4].

Core Viewpoint - The successful listing of Southern Power Grid Digital Grid Research Institute Co., Ltd. marks a significant milestone in Southern Power Grid's integration into the capital market and signifies a new phase in its comprehensive digital and intelligent transformation [1][6][7] Group 1: Company Development - Southern Power Grid aims to strengthen and expand its digital grid industry ecosystem to better support the intelligent, green, and integrated transformation of the energy and power industry [2][3] - The company has established itself as a leader in the digital grid sector, with a focus on high-quality development and delivering excellent performance to investors and society [2][7] Group 2: Technological Advancements - Southern Power Grid has made significant progress in key core technologies, including the development of a unified IoT operating system for the power industry and a large-scale domestic digital system [4][5] - The company has been recognized as a model enterprise for scientific reform and has received multiple accolades for its contributions to technology and innovation [4][5] Group 3: Market Positioning - The rapid listing process of Southern Power Grid Digital, taking only 145 days, reflects the company's strong compliance and preparation efforts, as well as the capital market's support for technology-driven enterprises [6][7] - The successful listing enhances the company's position as a high-quality platform serving multiple industries across the nation, providing a solid foundation for future growth [7]

Group 1 - The core viewpoint of the article highlights that Xizhi Technology has successfully completed a C-round financing of over 1.5 billion yuan, with participation from major investors including China Mobile and Shanghai Guotou [2] - Xizhi Technology, founded in Boston in 2017 and based in Shanghai, is recognized as a global unicorn in the photonic mixed computing sector, having developed its own platforms for photonic networks and computing [3] - The funds from the recent financing will be primarily allocated to three areas: accelerating the iteration and mass production of CPO photonic co-packaging technology, expanding the deployment scale of the LightSphere X super nodes, and investing in the development of the next-generation photonic mixed computing cards that support AI large models [4] Group 2 - The investment rationale is based on the recognition of the bottleneck in computing power in the post-Moore's Law era, with Xizhi possessing complete self-research capabilities in photonic and electronic chips and system-level packaging, which positions the company as a key player in defining the next generation of computing infrastructure [5] - The investment community has noted the supportive policies from the National Development and Reform Commission and the Ministry of Industry and Information Technology regarding photonic chip advancements, as well as the rapid response from the Shanghai Pudong "14th Five-Year" special funds to meet enterprise needs [6]

Group 1 - The core viewpoint of the news is that Xizhi Technology has successfully completed a C-round financing exceeding 1.5 billion yuan, indicating strong market confidence in its technology and commercialization prospects [2] - The financing round was participated by well-known investment institutions such as China Mobile, Shanghai Guotou, and Pudong Venture Capital, along with continued investment from existing shareholders [2] - Xizhi Technology is a leading unicorn in the field of optoelectronic hybrid computing, focusing on three core technologies: Optical Matrix Computing (oMAC), On-Chip Optical Networks (oNOC), and Inter-Chip Optical Networks (oNET) [2] Group 2 - In the field of optical networks, Xizhi Technology has successfully delivered multiple clusters with thousands of cards, focusing on Scale-Up solutions for intelligent computing centers [3] - The company launched the world's first distributed optical interconnect optical switching GPU supernode, LightSphere X, at the 2025 World Artificial Intelligence Conference, winning the highest honor, the "SAIL Award" [3] - Xizhi Technology is actively developing higher integration and larger bandwidth density CPO solutions to address bandwidth and latency bottlenecks in large-scale computing clusters [3] Group 3 - Xizhi Technology's optical hybrid computing card, released in March, integrates the world's largest 128×128 optical matrix, achieving practical application in complex commercial models [4] - The innovative optical hybrid computing architecture was published in the prestigious journal "Nature," marking a dual breakthrough in technology development and industry influence [4] - The company is accelerating the development of the next generation of optical hybrid computing cards to fully support the training and inference needs of AI large models [4] Group 4 - The rapid development of trillion-parameter large models has led to an explosive growth in global computing power demand, while traditional electronic chip performance is nearing physical limits [4] - Xizhi Technology addresses this bottleneck through two main paths: its optical interconnect solutions enhance multi-card collaboration efficiency and GPU utilization, while its optical hybrid computing reduces latency and total cost of ownership (TCO) [4] - The recent financing will be used to further expand technological advantages, accelerate product iteration, and enhance market applications, positioning China favorably in the global next-generation computing competition [4]

Core Viewpoint - The commercialization of photonic networks and photonic computing is reaching a critical turning point, with significant advancements in technology and funding opportunities for companies in this sector [1]. Group 1: Company Overview - Xizhi Technology, a leading player in the photonic computing sector, has completed over 1.5 billion CNY in Series C financing, with participation from major investors including China Mobile and Shanghai Guotou [2]. - The company is the only domestic unicorn focused on photonic mixed computing and is accelerating the development of the next-generation photonic mixed computing card to support AI large models [2][3]. Group 2: Technological Advancements - The company has developed a high-bandwidth, low-latency optical interconnect solution that significantly enhances GPU utilization and demonstrates advantages in computing density, energy efficiency, and cost reduction [3]. - Xizhi Technology launched the world's first distributed optical interconnect GPU supernode, LightSphere X, during the 2025 WAIC, winning the highest award, the SAIL Award [3][5]. Group 3: Research and Development - The latest generation of photonic mixed computing cards, Xizhi Tian Shu, features a 128×128 photonic matrix, marking the first application of photonic mixed computing in complex commercial models [5]. - The company has established strategic partnerships with leading domestic photonic/electronic wafer manufacturers and system vendors, facilitating breakthroughs in key areas such as supernode construction and CPO [5].

Core Viewpoint - Shanghai Xizhi Technology Co., Ltd. has completed a C round financing exceeding 1.5 billion RMB, aimed at accelerating the development of core technologies and the large-scale implementation of photonic hybrid computing power [1][9] Company Overview - Founded in February 2018, Xizhi Technology has a team of nearly 250 people, with over 80% being R&D personnel, primarily from MIT and other industry experts [2] - The company focuses on photonic networks and photonic computing, transitioning photonic hybrid technology from R&D to industrial application [2] Financing History - Xizhi Technology raised $10.7 million in seed funding in 2018, with investors including Baidu Ventures and ZhenFund [3] - In April 2020, the company completed a $26 million Series A financing led by Matrix Partners China and CICC Capital [3] - The company also completed a Series A+ financing in July 2020, with investments from HeLi Capital [4] - The B round financing details remain undisclosed, but it was reported that Shanghai International Group led the round, with subsequent investments from Tencent and Sequoia Capital [4] Technological Advancements - Xizhi Technology has made significant breakthroughs in key areas such as supernode construction and CPO (Co-Packaged Optics), with multiple technology and commercialization advancements announced recently [5] - The company launched the world's first distributed optical interconnect GPU supernode, LightSphere X, at the 2025 World Artificial Intelligence Conference, which won the highest award, the SAIL Award [5] - The LightSphere X aims to expand Scale-Up networks, introducing advanced optical interconnect technology to overcome physical limitations of traditional interconnect methods, significantly enhancing GPU utilization [5] Market Position and Strategy - With the increasing demand for computing power driven by large model parameters, Xizhi Technology is strategically positioning itself in the supernode market, recognizing it as a significant opportunity [6][7] - The company has completed multiple thousand-card clusters and aims to deploy ten-thousand-card clusters within the year [7] - Xizhi Technology is accelerating the development of the next generation of photonic hybrid computing cards to fully support AI large models while deepening commercial applications and ecosystem development [8]

Core Viewpoint - Alibaba is developing a new generation of AI chips focused on multifunctional inference scenarios, aiming to fill the market gap left by NVIDIA's H20 exit [1][3]. Chip Development and Specifications - The new chip utilizes domestic 14nm or more advanced processes, supported by local foundries like Yangtze Memory Technologies, integrating high-density computing units and large-capacity memory with an expected LPDDR5X bandwidth exceeding 1TB/s, targeting a single-card computing power of 300-400 TOPS (INT8), comparable to H20's approximately 300 TOPS [1][3]. - Compared to NVIDIA's H20, Alibaba's chip offers full-scene compatibility, supporting FP8/FP16 mixed precision computing and seamless integration with the CUDA ecosystem, reducing migration costs by over 70% [3]. - Alibaba has urgently increased its order for the Cambricon Siyuan 370 chip to 150,000 units, which is based on a 7nm process and utilizes Chiplet technology, integrating 39 billion transistors and achieving a measured computing power of 300 TOPS (INT8) with a 40% improvement in energy efficiency [5]. Market Strategy and Production Capacity - The Cambricon Siyuan 370 chip is expected to cover 60% of Alibaba Cloud's inference demand by Q2 2025 and supports multi-card interconnection via PCIe 5.0, facilitating user growth for Tongyi Qianwen [5]. - Alibaba collaborates with Yangtze Memory Technologies to develop AI chips focusing on overcoming storage bottlenecks, achieving a storage density of 20GB/mm² and read/write speeds of 7000MB/s, a 40% improvement over the previous generation, expanding local storage capacity to 128GB [5][6]. - To ensure mass production, Alibaba employs a dual-foundry backup strategy, with SMIC's 14nm production line handling basic chip production, achieving a stable yield of over 95% and a monthly capacity of 50,000 units [6]. Future Roadmap - Alibaba's three-step strategy includes: - Short-term (2025-2026): Focus on 7nm/14nm inference chips to quickly capture market share through ecosystem compatibility [10]. - Mid-term (2027-2028): Launch 4nm training chips targeting a computing power of 1 EFLOPS, competing with NVIDIA's H100 [10]. - Long-term (post-2030): Explore disruptive technologies like photonic computing and integrated storage-computing solutions, with the first commercial photonic AI chip already released, promising a speed increase of 1000 times and a 90% reduction in power consumption compared to GPUs [10]. - Alibaba's path to domestic computing power is characterized as a dual battle of technological breakthroughs and ecosystem reconstruction, aiming to disrupt NVIDIA's monopoly through a "compatibility-replacement-surpassing" strategy [10][11].