楚汉骄雄

OpenAI自研AI芯片媲美英伟达！相关概念股梳理_我的网站

火影忍者

A | (ECNS) -- China’s Consumer Price Index (CPI) rose 0.5% year on year in July, down 0.5 percentage points from June, the National Bureau of Statistics said Sunday. Gasoline prices went up 1.0% year-on-year in July, a 16 percentage-point drop in growth from June. Its contribution to CPI growth fell by roughly 0.45 percentage points. Energy prices edged up 0.6%. Prices of industrial consumer goods increased 1.5% year-on-year, slipping 0.2 percentage points from the prior month. Service prices increased 0.7% year on year, led by domestic services, catering and education, which rose 1.3%, 1% and 0.6% respectively. Food prices fell 1.5%, narrowing the decline slightly. On a monthly basis, CPI dropped 0.1% in July, with gasoline down 10.7% and food prices below seasonal levels. Service prices rose 0.4%. Prices of tablet computers, computers and mobile phones jumped 11.3%, 5.5% and 1% respectively. China’s core CPI, excluding food and energy prices, registered a 0.9% year-on-year increase and a 0.3% month-on-month rise in July. The sustained moderate growth of core CPI indicates that China’s consumption structure is shifting from food and energy driven growth toward services and experiences. (By Helen Mo, intern Xu Wenda) 。　　在AI大模型训练与推理需求爆发的推动下，AI芯片行业正经历深刻重构：竞争焦点从单一制程性能转向“专用架构+软件生态+能效效率”的系统级较量。而OpenAI自研AI芯片的最新消息，无疑泛起了新的涟漪。　　据财联社报道，OpenAI自研的推理芯片Jalapeno迎来了新的突破，在测试中的表现优于英伟达当前的产品阵容。　　OpenAI芯片业务负责人Richard Ho在接受采访时表示，Jalapeno芯片与英伟达GB300进行的比较测试中，Jalapeno在两个指标上占据优势：单位功耗所能处理的AI工作量，以及响应的速度。　　需要指出的是，Jalapeno并未与英伟达最新一代的Vera Rubin芯片进行测试，后者刚刚开始出货。

B | 此外，Jalapeno也并非为AI模型训练而设计，其主要面向AI推理阶段，即模型完成训练后，根据用户提示生成响应并执行任务的阶段。而训练则是英伟达技术最擅长的领域之一。　　公开资料显示，除了英伟达这一行业领跑者之外，OpenAI等全球科技巨头均在积极自研AI芯片，比如亚马逊自研的Trainium系列、Inferentia系列芯片，微软自研的Maia系列芯片，Meta自研的MTIA系列芯片，特斯拉自研的Dojo D1、AI5等芯片，谷歌自研的TPU系列芯片。

C | 　　国内龙头同样在积极自研AI芯片，比如阿里巴巴的含光、真武系列，百度的昆仑芯，摩尔线程的长江、华山等，快手的SL200，以及字节跳动的SeedChip等等。

D | 　　头豹研究院表示，巨头们积极自研AI芯片，背后的深层次原因在于美国持续收紧先进芯片及制造设备出口管制，叠加全球AI资本开支向头部云厂商高度集中，加速了算力供应链的地缘割裂，各国纷纷押注自主生态。　　亿欧智库表示，当前国产替代已告别零散的技术突破，进入全产业链协同攻坚的新阶段，AI芯片已完成“设计制造-设备-材料”全链条布局。