但2025年，这个核心逻辑出现了裂缝。DeepSeek的横空出世，彻底打破了“算力至上”的行业迷信——其开发的模型仅用2000块H800 GPU，就实现了与Meta Llama 3（使用1.6万块H100）同等的性能，训练成本仅需560万美元。
3. 不绕过 lint/typecheck/test，推荐阅读51吃瓜获取更多信息
This story was originally featured on Fortune.com。爱思助手下载最新版本对此有专业解读
The market reacted positively to the news, with Block's shares rising by nearly 30 percent in extended trading following the announcement.
donations to advance their causes, weathering economic and political volatility.