这份路线图，或许正是那块厚实跳板 第一次发出的、清晰可闻的蓄力声。 当大部分同行还在用"对标"作为宣传话术时，这家公司直接把超越Hopper、Blackwell、Rubin的时间 点写进了2025-2027年的日历。底气何来？答案是：过去七年没走捷径。 在AI芯片发布会"人均对标英伟达"的语境下，一份真正敢把超越日期写进标题的架构路线图，本身就 构成了新闻。 1月26日，天数智芯公布了其2025至2027年的四代架构路线图。与寻常的未来可期不同，它像一份产 品手册般精确： 2025年天数天枢超Hopper，2026年天数天璇架构对标Blackwell，同样在2026年，天 数天玑架构超越Blackwell，2027年天数天权超Rubin ，2027年之后将转向突破性计算芯片架构设计。 未来3年，天数智芯将基于此次发布的四代架构，陆续发布多款产品，持续提升计算性能。 更不同寻常的是，这份这份未来宣言的第一行，已被验证。在天数智芯路线图中的最近一代架构—— 2025年推出的天枢，在关键的大模型场景——DeepSeek V3上，平均性能已比英伟达Hopper架构高出 约20%。 "他们不是在预测未来，他们是在汇报进 ...

Core Insights - The current anxiety surrounding AI financing due to perceived market vulnerabilities is largely exaggerated, as stated by Sung Cho from Goldman Sachs Asset Management [1] - AI investments are primarily funded by strong internal cash flows from established tech giants rather than speculative debt, which is crucial for understanding the long-term resilience of the AI market [1] - Approximately $700 billion to $1 trillion is expected to be spent on AI in the coming years, with 90% of this funding coming from operational cash flow [1] Debt Financing - Only 10% of AI investments are financed through debt, with a significant portion issued by high-rated entities, such as Meta, which has a credit rating superior to that of the U.S. government [2] - The debt financing in the AI sector is concentrated among stable, high-credit companies, reducing systemic risk [2] Market Dynamics - The rapid turnover of perceived leaders in the AI model space is a defining characteristic of this emerging market, with investor sentiment and valuations shifting quickly among key players like Meta, OpenAI, and Google [3] - The financial impact of this "model volatility" is significant, as seen in Google's recent market capitalization increase of $1 trillion due to perceived advancements in their AI model, Gemini [3] Future Trends - The competitive landscape will continue to evolve rapidly, with new AI models expected to emerge, particularly those based on NVIDIA's Blackwell architecture, leading to further volatility and shifts in leadership [4] - The AI market is characterized by robust foundational capital supported by sustainable operational cash flows, rather than precarious leverage, indicating a healthy financial environment despite the competitive dynamics [5]

AI这条主线已经火热了两年，2023年、2024年在资本市场都很受欢迎，今年也不例外。今年行情的热 度，可能是从国内DeepSeek的软件发布后便引发了一场世界级的旋风——大家没想到竟有中国公司能 用美国十分之一的算力，做出一个几乎等效的软件，或者说一个大模型。这一现象不仅在国内卷起了一 个旋风，在海外也卷起了旋风，核心原因是因为大家没有想到这是中国的公司做出来的。外界此前普遍 认为中国人工智能水平与海外存在代差，且差距达数年之久，而DeepSeek的出现打破了这一认知。 （文章来源：每日经济新闻） 在DeepSeek之后，我们也看到英伟达发布了新一代Blackwell架构，且在今年便开始放量。放量后，英 伟达在一季度发布会上提到，其GB200机架的周部署量约为1000台；如此一来，1000台机架对应的周度 GPU销量便达7万多张，且单卡价格较高，每张达几万美金，因此Blackwell的销售额爬坡非常快。此 外，GB200后续还会切到GB300，可见整个需求侧表现亮眼，这是硬件侧的核心驱动。 算力领域的细分环节包含哪些？主要包括通信和芯片两大领域。在通信领域，光模块、PCB、服务器的 基本面比较强，且2026 ...

Core Viewpoint - G42's subsidiary Khazna Data Centers is collaborating with NVIDIA to accelerate the development of AI infrastructure in the Middle East and Africa [1] Group 1: Partnership Details - Khazna Data Centers has announced a partnership with NVIDIA to build AI factories in the Middle East and Africa [1] - NVIDIA has certified Khazna's next-generation data center facility design to support the NVIDIA Blackwell architecture [1] Group 2: Development Plans - Khazna plans to develop up to 250 megawatts of AI clusters as part of this initiative [1]