经济观察报记者钱玉娟 在2025世界人工智能大会(下称"WAIC2025")开幕前夜，7月25日，中国人工智能科技企业上海阶跃星辰 智能科技有限公司(下称"阶跃星辰")发布了首个全尺寸、原生多模态推理模型Step3，这一基座大模型将 于7月31日面向全球企业和开发者开源。 阶跃星辰创始人、CEO姜大昕接受包括经济观察报在内的媒体采访时称，当下坚持做基础大模型的公司 越来越少，困难在于投入巨大。特别是当大模型产业过渡至2.0时代后，不少公司会放弃模型训练，去 追求商业化。而他认为，模型的能力会决定应用的上限，应用也会给模型提供具体的场景、数据，因 此，阶跃星辰更坚持超级模型加超级应用的路径。 阶跃星辰联合创始人、副总裁李璟负责商业化，他在接受包括经济观察报在内的媒体采访时明确，上述 收入指确认的合同收入，该目标建立在阶跃星辰2025年上半年已实现数亿元合同收入且毛利水平较高的 基础上。 姜大昕认为，最适合实际应用的大模型需要满足强智能、低成本、可开源和多模态四个特征，缺一不 可。唯有模型全面发展，才能让模型真正用起来。这也是阶跃星辰研发Step3基础模型的出发点：为追 求性能与成本极致均衡的企业和开发者设计。 ...

这个暑假，在学校搞 AI 的你是不是还在卷研究？ 是不是还缺点算力？ 是不是想要点折扣？ 它来了！面向 高校用户， 英博云特别推出「暑期现金消耗返券活动」。 满足规则，A800 低至 4.26 元 / 卡 / 小时起，H800 低至 9.33 元 / 卡 / 小时起。 活动时间 即日起至 8 月 31 日 返券规则 具体返利比例 消耗金额越高，返利比例阶梯式飙升， 满 10000 元及以上直接返 30%！ 福利叠加 三重惊喜： 福利一：注册及首充赠券 福利二：充值满额赠券 福利三：现金消耗满额赠券（8 月 31 日截止） 福利叠加价格示例： 注册成功赠送 100 元代金券（8 月 1 日起赠送减半）； 首次充值满 100 元赠送 200 元代金券（8 月 1 日起赠送减半）； 高校用户单笔或累计消耗现金达指定金额，即可领取对应比例的代金券（可叠加使用） 消费达标：活动期间通过英博云平台消耗现金（支持单笔或累计） 返券流程：根据活动期消耗现金情况统一进行发放（活动结束后） 最后提醒 代金券有效期为 3 个月，建议提前规划，避免过期。 立即扫码，参与活动 （活动最终解释权归英博云所有） 关于英博云 北京英博数科科 ...

2025年4月，美国政府限制英伟达向中国出口H20芯片，导致英伟达蒙受直接经济损失，其在2026财年 第一季度（截至2025年4月27日）计提了45亿美元的库存减值损失， 这一计提直接反映了H20库存因销 售受阻产生的价值缩水 ，导致当期净利润减少45亿美元，毛利率从预期的71.3%降至60.5%。 英伟达计提的损失，是禁令导致用于生产H20的相关物料出现严重库存积压。黄仁勋近日表示将非常努力 地恢复生产，实则是将积压的物料利用起来。据英伟达供应链相关人士透露，以当前积压的晶圆量计算， 可生产的H20芯片数量或达大几十万片，最早一批H20将在8月份上市。 H20仅适配中国市场，海外没有需求。对任何一家商业公司而言， 绝对不会承受几十亿美元的无谓损耗。 此前H800被禁售时， 英伟达曾将全部剩余库存出售给超微，把库存压力转嫁至OEM厂商， 这也是禁售 后市场上H800的货仍源源不断的原因。 " 国内云厂商对向美国备案信息尚有顾虑，下单H20仍存变数。 " 作者丨刘伊伦 编辑丨包永刚 北京时间7月15日，英伟达宣布将恢复H20在中国的销售，并宣布推出面向中国市场的全新且完全兼容的 GPU。黄仁勋在采访中表示，美国 ...

【文/观察者网 王一】7月15日在今年第三次访华之行中，美国英伟达公司创始人兼首席执行官黄仁勋 宣布了一个"非常、非常好的消息"——叫停3个月后，美国特朗普政府已批准英伟达向中国市场销售 H20芯片。 此前还扬言要彻底封锁对华芯片出口的特朗普政府，如今却来了个政策"急转弯"。彭博社、美国消费者 新闻与商业频道（CNBC）等美媒直言，这对黄仁勋来说是一个"巨大的胜利"。 "美国政府已经批准了我们的出口许可，我们可以开始发货了，所以我们将开始向中国市场销售H20。 我非常期待能很快发货H20，对此我感到非常高兴，这真是个非常、非常好的消息。"黄仁勋15日对中 国央视记者透露，英伟达还将发布一款名为RTX Pro的新显卡，"专为计算机图形、数字孪生和人工智 能（AI）设计"。 英伟达也在公司官网发文证实了这一消息，称美国政府已向英伟达保证将发放H20出口许可证，公司希 望能尽快开始发货。 H20是英伟达在2023年年底，为遵守美国当时的出口管制而专为中国市场设计的"减配版"AI加速器。但 今年4月，特朗普政府告诉英伟达，在未经其许可的情况下，禁止在中国销售H20芯片。 彭博社称，此次放行发生在美中关系出现缓和迹象的 ...

Core Viewpoint - The H20 market is experiencing high demand, with potential buyers urged to act quickly due to limited supply and significant interest from Chinese companies [1][4]. Supply and Demand - Current H20 supply consists of existing inventory, with estimates ranging from 300,000 to 400,000 units or 600,000 to 1,000,000 units, indicating a limited availability [1]. - Chinese enterprises are rapidly purchasing H20, with large companies submitting substantial applications [1]. Technical Aspects - Discussions on transitioning from H200 (or H800) to H20 suggest the use of "point cutting" technology for hardware downscaling, differing from previous software methods [2]. - There are indications that after the ban on H20, Nvidia considered reverting H20 back to H200, but the high costs led to abandonment of this plan [2]. Market Impact - The release of H20 is expected to negatively impact certain sensitive companies, although specific names are not disclosed [3]. - Once the existing H20 inventory is sold out, it is unlikely that new H20 units will be produced, as Nvidia is focusing on the Blackwell series [3]. Buyer Recommendations - Potential buyers are advised to act without hesitation, as future availability may become constrained [4].

每经记者｜朱成祥 杨卉 每经编辑｜陈俊杰 7月15日，《每日经济新闻》记者从英伟达方面获悉，公司将恢复H20（一款面向国内市场推出的 GPU）在中国的销售，同时还将推出面向中国市场的全新且完全兼容的GPU（图形处理器）。本月，英 伟达创始人兼首席执行官黄仁勋在美国和中国推广AI（人工智能），强调了AI将为全球商业和社会带 来的诸多益处。 与非研究院资深分析师张慧娟认为："H20恢复销售，可以稳住云计算等大客户，避免份额进一步流 失。另外最新推出的RTX PRO GPU，宣传'是为智能工厂和物流打造数字孪生AI的理想选择'，英伟达 此举既避开敏感的高算力训练场景，又切入中国工业数字化转型的蓝海市场，可谓定位精准。" 独立国际策略研究员陈佳在接受《每日经济新闻》记者采访时指出："除了英伟达，本月还有多家巨头 恢复了对华业务，这一行为背后，是国内科技产业链外部环境有所改善的事实。" 为何急于恢复H20供应？ 此次，黄仁勋在为恢复向中国销售H20方面付出了相当大的努力。 据公司方面透露，在美国华盛顿，黄仁勋会见了政策制定者们，重申了英伟达在支持政府创造就业机 会、加强美国AI基础设施和本土制造业，以及保持美国在AI领 ...

Core Insights - Nvidia is filing applications to restart sales of its H20 artificial intelligence chips to China, following a period of regulatory changes and discussions with U.S. officials [1][6] - The company anticipates receiving U.S. government licenses soon and plans to introduce a new "RTX Pro" chip tailored for the Chinese market, which is claimed to be fully compliant with regulations [2] Group 1: Regulatory Environment - The H20 chip is central to the U.S.-China tech standoff, being the most powerful chip Nvidia can legally sell to China under current export controls, designed for inference tasks rather than training new AI systems [3] - The Trump administration's restrictions in April could have cost Nvidia between $15 billion to $16 billion in revenue, based on Chinese firms' spending in the first quarter [5] - The restrictions were briefly lifted after Nvidia's CEO met with Trump, where promises of U.S. investments and job creation were made in exchange for continued access to chip sales [6] Group 2: Market Dynamics - Chinese tech giants like ByteDance, Alibaba, and Tencent have been stockpiling H20 chips in anticipation of stricter export controls, attracted by the chip's superior memory bandwidth and Nvidia's software ecosystem [4] - The ongoing situation highlights the balancing act U.S. policymakers face between national security concerns and commercial interests, suggesting potential future reversals in policy [10]

Core Viewpoint - Kunlun Core has recently completed a new round of financing, but there is a noticeable divergence in information disclosure, with investors and financial advisors announcing the news while Kunlun Core remains silent, raising questions about the underlying issues [1][2]. Financing Situation - Kunlun Core's latest financing round has seen participation from various investors, including Shanghe Momentum Capital and Shanzheng Investment, with the latter confirming their involvement just a day before the announcement by Shanghe [1]. - The company has undergone six rounds of financing since its establishment, with Baidu remaining the largest shareholder, though its stake has been diluted from 76.17% to 67.49% [4][6]. IPO Context - The financing is taking place against the backdrop of a surge in domestic GPU and AI chip companies preparing for IPOs, with nearly ten companies expected to file for listings soon, collectively referred to as the "Chinese Nvidia" [6][9]. - The current market conditions have made it difficult for GPU projects to attract investment, with many investors expressing reluctance to invest in the GPU sector due to significant losses reported by existing companies [7][9]. Product Development and Market Position - Kunlun Core, which originated from Baidu's smart chip division, has launched its first and second-generation chips, with plans for a third-generation chip (P800) set for release in 2025 [11]. - The company has achieved a notable milestone by creating a fully self-developed 30,000-card cluster capable of training large models, which is a significant achievement in the domestic market [11][13]. Competitive Landscape - In terms of performance, Kunlun Core's P800 ranks among the top in the domestic market, but it still lags behind some competitors like Birun's BR100, which boasts superior specifications [18][19]. - According to IDC, Kunlun Core ranks second in the domestic AI accelerator card market, although the accuracy of this data has been questioned by industry insiders [19][20]. Production Capacity and Future Outlook - The production capacity required to meet the projected demand for Kunlun Core's products is substantial, with estimates suggesting that 1,400 wafers would be needed to achieve the anticipated output [20]. - The upcoming regulatory changes may impact the production capacity of domestic AI chip companies, making it crucial for them to secure sufficient manufacturing resources to maintain their market positions [21].

Core Viewpoint - Huawei's Ascend CloudMatrix 384 super node has demonstrated performance that surpasses NVIDIA's products in certain aspects, indicating a significant advancement in domestic AI chip capabilities [1][11][13]. Group 1: Huawei's Ascend Chip Overview - Ascend is a dedicated AI processing chip (NPU) designed specifically for AI tasks, with the Ascend 910 being its main product [3][6]. - Previously, Ascend chips were used as backup options due to the unavailability of high-end NVIDIA and AMD chips, but they have now emerged as leaders in the domestic chip market [3][6]. - The Ascend chips have primarily been utilized in AI inference tasks, with limited use in model training due to performance and ecosystem limitations [4][6]. Group 2: Performance and Capabilities - In 2024 and 2025, Huawei transformed Ascend from a backup option to a primary player capable of training large models, achieving significant results documented in research papers [5][6]. - Ascend has successfully trained models with 135 billion parameters using 8192 chips and 718 billion parameters using over 6000 chips, showcasing the ability to train large-scale models with domestic chips [6][10]. - Key performance indicators such as MFU (Modeling Function Utilization) reached over 50% for the dense model and 41% for the MoE model, indicating high efficiency in resource utilization [9][10]. Group 3: Competitive Comparison with NVIDIA - In direct comparisons, Ascend's 384 super node demonstrated comparable performance to NVIDIA's H100 and H800 in real-world applications, achieving the best utilization rates [11][12]. - Although a single Ascend chip's performance is only one-third of NVIDIA's Blackwell, the overall system performance of the 384 super node exceeds NVIDIA's GB200 due to the higher number of chips used [13][21]. - This indicates that Ascend is not just a replacement but has the potential to lead in certain performance metrics [13]. Group 4: Technological Innovations - The CloudMatrix 384 super node consists of 384 Ascend 910 chips and 192 Kunpeng CPUs, interconnected using advanced optical communication technology, which enhances data transmission efficiency [16][30]. - Huawei's approach focuses on a system-level engineering breakthrough rather than relying on single-chip performance, utilizing a combination of communication, optical, thermal, and software innovations [21][22]. - The architecture allows for high-speed, peer-to-peer communication among chips, significantly improving data transfer rates compared to traditional copper connections used by competitors [28][30]. Group 5: Market Position and Future Outlook - Despite still trailing behind NVIDIA in chip technology and software ecosystem, Huawei's Ascend has gained traction in the Chinese market, especially as companies adapt to domestic chips due to restrictions on NVIDIA products [36][38]. - The domestic semiconductor industry is evolving under pressure, with Huawei's strategy representing a unique "technology curve" that prioritizes system optimization over individual chip performance [38][39]. - The advancements made by Ascend may signify the beginning of a significant shift in the AI computing landscape, positioning domestic capabilities for a potential resurgence in the global market [40].

Core Viewpoint - Huawei's Ascend 384 Super Node has demonstrated performance that surpasses NVIDIA's products in certain aspects, indicating a significant advancement in domestic AI chip capabilities [2][3]. Group 1: Product Overview - Ascend is an AI chip developed by Huawei, specifically designed for AI tasks as an NPU, distinguishing it from traditional GPUs and CPUs [4]. - The main product, Ascend 910, has transitioned from being a backup option to a primary solution for training large models due to restrictions on high-end chips from NVIDIA and AMD [4][6]. Group 2: Performance Metrics - In recent developments, Huawei has successfully trained large models using Ascend chips, achieving a dense model with 135 billion parameters and a MoE model with 718 billion parameters [6]. - The key performance indicator, MFU (Modeling Function Utilization), reached over 50% for the dense model and 41% for the MoE model, indicating efficient utilization of computational resources [9]. Group 3: Competitive Analysis - In a direct comparison with NVIDIA's H100 and H800 during the deployment of large models, Ascend demonstrated comparable performance, achieving the best utilization rate in the competition [10]. - Although a single Ascend chip's performance is only one-third of NVIDIA's Blackwell, the 384 Super Node configuration, which utilizes five times the number of chips, results in an overall computational power that exceeds NVIDIA's GB200 [10].