以下文章来源于智能涌现 ，作者邱晓芬 智能涌现 . 直击AI新时代下涌现的产业革命。36氪旗下账号。 老黄来华，诚意和动作都有了，但英伟达在"禁令"后丢失的蛋糕，已迅速被国产芯片公司瓜分。 文 ｜ 邱晓芬 编辑 ｜ 苏建勋 来源｜ 智能涌现（ID：AIEmergence） 封面来源 ｜ 网络 为了抢回中国的AI算力市场，英伟达创始人黄仁勋正进行一场意义重大的中国之行。 此次黄仁勋行程满满，除了政府侧的拜访，他与小米创始人雷军的合照也已传遍网络。7月16日，他还会出席链博会（中国国际供应链促进博览会）。此 外，英伟达官方还为他照例组织了一场中国媒体会。 要面对这么多官员、客户和媒体，黄仁勋自然不会空手而来。7月15日，在黄仁勋的中国之行途中，他亲自向客户们宣布——英伟达已得到美国政府保 证，将恢复H20在中国的销售。 或许是为了避免清库存嫌疑，黄仁勋还表示，将向中国市场推出全新兼容的RTX Pro系列GPU。 在Computex 2025期间，黄仁勋直言，英伟达在中国AI芯片的份额已经从2022年出口管制期间的95%，下降至50%。 从今年4月份禁售，到如今重新恢复供货，英伟达H20芯片的禁令只维持了短短三个月 ...

Core Viewpoint - AMD is positioning itself to challenge NVIDIA in the AI chip market, with new product launches and a projected market growth exceeding $500 billion by 2028 [2][3]. Product Launches - AMD unveiled its flagship data center AI chips, the Instinct MI350 series, along with a new AI software stack, ROCm 7.0, during the Advancing AI 2025 event [2]. - The MI350 series includes MI350X and MI355X GPUs, both utilizing 3nm technology with 185 billion transistors and HBM3E memory, achieving a performance increase of 4 times and a 35 times faster inference speed compared to the previous MI300X [2][3]. - The MI350 series is set to launch in Q3 2025, with adoption from major companies like Oracle, Dell, Supermicro, HPE, and Cisco [3]. Future Roadmap - AMD plans to release the next-generation MI400 series GPU in 2026, developed in collaboration with OpenAI, which has already provided feedback based on its use of the MI300X chip [3]. Financial Performance - AMD reported Q1 2025 revenue of $7.44 billion, a 35.9% year-over-year increase, exceeding market expectations, with a net profit of $710 million [4]. - In contrast, NVIDIA achieved Q1 2025 revenue of $44.1 billion, a 69% year-over-year increase, with a net profit of $18.78 billion [4]. Market Position - According to TrendForce, AMD ranks fourth among global chip design firms due to weak data center performance and declining gaming product demand, while NVIDIA remains the leader in AI chip design [5]. - AMD aims to ramp up production of the MI350 series in the second half of 2025 to compete directly with NVIDIA's offerings [5].