据市场调查机构Factorial Funds最新报告显示，人工智能研究实验室OpenAI在部署其创新的文本转视频模型Sora时，可能面临巨大的硬件需求。该机构估计，在模型运行的峰值阶段，OpenAI可能需要高达72万张英伟达的顶级H100 AI加速卡。这一数字背后的价值预估达到了惊人的216亿美元，这在科技领域无疑是一笔巨额投资。
Factorial Funds进一步分析，为了训练Sora模型，OpenAI在一个月内的消耗量可能在4200到10500片H100加速卡之间。这一高强度的计算需求凸显了AI研发中对高性能硬件的依赖。据报告透露，使用单片H100加速卡生成1分钟的视频内容，需要实际渲染的时间长达12分钟，这在效率上提出了新的挑战。
这一消息引发了业界对AI技术进步成本的关注，同时也反映出OpenAI在推动人工智能边界拓展时所面临的硬件和计算力瓶颈。随着Sora模型的开发进程，如何在保证性能的同时，有效控制和优化成本，将成为OpenAI及其合作伙伴亟待解决的问题。此外，这也可能为高性能GPU市场带来新的需求增长点，影响相关产业的布局和发展。
英语如下：
News Title: “OpenAI’s Text-to-Video Model Sora Requires 720,000 Nvidia H100 Cards, Valued at $216 Billion for Deployment”
Keywords: OpenAI, Sora model, Nvidia H100
News Content: According to the latest report from market research firm Factorial Funds, artificial intelligence research lab OpenAI could face substantial hardware requirements in deploying its innovative text-to-video model, Sora. The firm estimates that during peak operation, OpenAI might need as many as 720,000 top-of-the-line Nvidia H100 AI accelerator cards. This figure translates to an astounding valuation of $216 billion, a massive investment in the tech sector.
Factorial Funds further analyzes that for training the Sora model, OpenAI’s consumption could range between 4,200 to 10,500 H100 cards within a month. This high computational demand underscores the reliance on high-performance hardware in AI development. The report reveals that using a single H100 card to generate one minute of video content requires actual rendering time of 12 minutes, posing new challenges in efficiency.
This news has drawn industry attention to the cost of advancing AI technology and reflects the hardware and computational power bottlenecks that OpenAI encounters in pushing the boundaries of AI. As the development of the Sora model progresses, the challenge for OpenAI and its partners will be to balance performance with effective cost control and optimization. Furthermore, this could stimulate new demand growth in the high-performance GPU market, potentially reshaping the industry landscape and development.
【来源】https://www.ithome.com/0/758/413.htm
Views: 1