据市场调查机构Factorial Funds最新发布的报告显示，OpenAI在推进其文本转视频模型Sora的部署过程中，将面临巨大的硬件需求。该机构估计，在Sora的峰值运行期间，可能需要高达72万张英伟达的顶级H100 AI加速卡，这相当于约216亿美元的硬件成本。
Factorial Funds的报告指出，为了训练Sora模型，OpenAI在一个月内可能需要使用4200到10500片H100加速卡，具体数量取决于训练的复杂度和效率。这一数字凸显了在人工智能领域，尤其是高精度的视频生成技术上，对高性能计算资源的极度依赖。
据透露，每片英伟达H100加速卡用于生成1分钟的视频，需要实际渲染的时间长达12分钟。这意味着Sora的运行不仅在硬件投资上存在巨大挑战，同时在时间效率上也面临考验。这一计算效率问题可能影响到Sora在实际应用中的普及和商业化进程。
OpenAI的Sora模型，如果成功部署，将有可能彻底改变内容创作和媒体行业，将文本直接转化为栩栩如生的视频。然而，目前的技术瓶颈和高昂的成本无疑为这一前景蒙上了一层阴影。随着人工智能技术的快速发展，如何平衡技术进步与资源消耗，将是业界需要共同面对的重要议题。
英语如下：
**News Title:** “OpenAI’s Text-to-Video Model Sora: Estimated Deployment Requires 720,000 Nvidia H100 Cards, Amounting to a Stunning $216 Billion”
**Keywords:** OpenAI, Sora model, Nvidia H100
**News Content:** According to a newly released report by market research firm Factorial Funds, OpenAI faces substantial hardware demands in the deployment of its text-to-video model, Sora. The agency estimates that during peak operation, as many as 720,000 top-of-the-line Nvidia H100 AI accelerator cards may be required, translating to an approximate hardware cost of $216 billion.
The Factorial Funds report reveals that for training the Sora model, OpenAI might need 4,200 to 10,500 H100 cards within a month, depending on the complexity and efficiency of the training process. This figure underscores the extreme reliance on high-performance computing resources in the AI domain, particularly in advanced video generation technologies.
It has been disclosed that each Nvidia H100 card takes 12 minutes to render one minute of video. This indicates not only significant challenges in hardware investment for Sora’s operation but also tests its efficiency in terms of time. This computational efficiency issue might impact the普及 and commercialization of Sora in practical applications.
If successfully deployed, OpenAI’s Sora model has the potential to revolutionize content creation and the media industry by transforming text into lifelike videos. However, the current technological bottlenecks and exorbitant costs cast a shadow over this promising prospect. As AI technology rapidly progresses, striking a balance between technological advancements and resource consumption emerges as a critical issue for the industry to address collectively.
【来源】https://www.ithome.com/0/758/413.htm
Views: 2