Post

Cato_KT
Cato_KT
Influential Creator
Show original
一个未经工信部公告的消息,直接吓坏了周一的A股科技股,中国人工智能之路任重道远,目前来看信心确实欠缺 这条消息工信部并未透露出一点相关消息,倒是海外媒体大肆报道,这让我有点质疑消息的真实性 不过对于国内科技股来说,临近国庆假期,这几天肯定是草木皆兵,风声鹤唳了 后续还是要关注工信部是否证实消息的真实性 前文提到的概念逻辑——国产芯片负责战略方向与大规模推理→英伟达高端GPU作为训练与高性能算力缺口的补充→通过审批数量与用途来控制对英伟达芯片的依赖性 想要实现这一可能难度不低,起码如果没有配套制度限制,很容易先打击到国内人工智能企业的股价#财报观察员:美光财报临近,AI存储需求成焦点
Cato_KT
Cato_KT
Is China planning to purchase the latest American chips?
China plans to purchase the latest US chips? This could be a substantial breakthrough for artificial intelligence following the China-US summit, and this time the consideration for procurement shows significant progress compared to before. Once executed, it could be the biggest breakthrough of this China-US summit! There are several details to note about this planned procurement: 1. According to reports, the purchase under consideration is not the previous H200, but the latest product of this month, the RTX Pro 5500, which is positioned as a high-end dedicated cloud computing product. 2. Part of this procurement involves some companies purchasing in large quantities with configurations of 8 cards per server, planned for AI model operation. Current reports show that ByteDance's purchase volume is very large, about 1 million cards. Nvidia's target supply for the Chinese market is 500,000 cards per quarter, so ByteDance's purchase amount alone already meets Nvidia's supply for two quarters. 3. The current price for a single RTX Pro 5500 card is about 85,000-90,000 RMB per card. Roughly calculating for 1 million cards, ByteDance's potential order is about 85-90 billion RMB, involving a huge amount of money. Therefore, the final outcome depends on the actual quantity approved by the government, the approval standards, and Nvidia's actual supply situation. #财报观察员:美光财报临近,AI存储需求成焦点 4. There is a shift in policy signals. Previously, the procurement logic for the H200 was limited approval for advanced training computing power + limited domestic inference. Now, the policy logic is very likely to become: domestic chips are responsible for strategic directions and large-scale inference → Nvidia high-end GPUs serve as training and high-performance components.

Disclaimer: OKX Orbit content is provided for informational purposes only. Learn more

Replies

No comments yet. Be the first to reply!