SpaceX的AI竞争对手正争相向这家人工智能和火箭公司租用算力，首席执行官埃隆·马斯克表示。
SpaceX's AI rivals are clamoring to rent computing power from the artificial intelligence and rocket company, CEO Elon Musk said.
“坦率地说，我们被每一家超大规模云服务商和数十家小公司不断催促，要求租用更多我们的算力，”马斯克周二在SpaceX第三季度财报电话会议上表示。
"We are also, frankly, hounded relentlessly by every hyperscaler and by dozens of smaller companies to rent more of our compute," Musk said during SpaceX's third-quarter earnings call Tuesday.
马斯克说，如果对所有人都说yes，SpaceX可以将公司的算力出租“多次”。
Musk said SpaceX could rent out the company's compute "several times over if we said yes to everyone."
然而，SpaceX正在保留一部分算力用于自身的人工智能雄心，包括扩展其AI聊天机器人Grok以及它最近收购的编码初创公司Cursor。
SpaceX, however, is reserving a chunk of that capacity for its own AI ambitions, including scaling its AI chatbot Grok and Cursor, the coding startup it recently acquired.
马斯克表示，SpaceX计划将大约一半的算力用于内部用途，另一半出租给第三方。这一比例可能每月都会变化，他说。“这将是动态的，”马斯克说。
Musk said SpaceX is seeking to devote roughly half of its compute to internal uses and rent out the other half to third parties. That balance could change month to month, he said. "It's going to be dynamic," said Musk.
“显然，我们的做法是按现货市场价格出租，而不是签订长期协议。”
"And obviously, our approach is to rent at the — effectively the spot market rate as opposed to sign these long-term deals."
该公司预计，随着AI模型不断改进，算力的价值将增加，并最终产生“每瓦更多有用智能”，马斯克说。
The company expects the value of compute to increase as AI models improve and ultimately generate "more useful intelligence per watt," said Musk.
“该算力的价值与每瓦有用智能成正比，”马斯克补充道，并指出较旧的英伟达H100芯片的租赁费率已从推出时上涨。“我认为这很可能会继续，”马斯克说。“我们将面临超级智能算力短缺。这种短缺将持续很长时间。”
"And the value of that compute is proportionate to the useful intelligence per watt," Musk said, adding that rental rates for older Nvidia H100 chips have climbed from when they were new. "I think that's likely to continue," said Musk. "I think we're going to face a shortage of super intelligence compute. And that shortage is going to remain for a long time."
马斯克将这种情况比作蒙蒂·派森著名的奶酪店小品，其中一位顾客反复询问不同种类的奶酪，却发现一种也没有。
Musk likened the situation to Monty Python's famed cheese shop sketch, in which a customer repeatedly asks for different cheeses only to find that none are available.
“我们实际上试图将部分推理算力卸载到其他云服务商，以便能用更多算力进行训练，”首席执行官说。“结果就是奶酪店小品。”
"We've actually tried to offload some of our inference compute to other cloud providers so that we can use more compute for training," the CEO said. "And we get the cheese shop sketch."