Nvidia's stock bucked broader chip-sector weakness on Wednesday as it secured its fifth session in a row of gains.
The company's artificial-intelligence chips got a major endorsement from Elon Musk on SpaceX's first-ever earnings call late Tuesday. He touted Nvidia's upcoming Vera Rubin chip architecture as "the best" and said that SpaceX $(SPCX)$ has "decided to build exclusively on Nvidia" from now on.
"We think it's the best AI computer, and we greatly value our close cooperation and partnership on many levels with Nvidia," Musk said.
Later in the call, Musk said SpaceX's "understanding with Nvidia is that we will receive a very significant percentage of their GPUs next year," referring to graphics processing units. Nvidia currently dominates the market for GPUs, where it mostly competes with Advanced Micro Devices $(AMD)$.
Nvidia's stock gained 3.4% Wednesday and has risen for five consecutive sessions, gaining 15.4% over this period, according to Dow Jones Market Data.
SpaceX is expecting to install more than 2 gigawatts of compute capacity by the end of the year, Musk said, and has planned for close to 10 gigawatts by the end of 2027.
Given that Nvidia could see $35 billion in revenue per installed gigawatt of its chips, SpaceX on its own would make up more than the $160 billion in incremental growth Wall Street expects for the chip maker's data-center business in 2027, Wolfe Research analyst Chris Caso said.
"Amid market concerns regarding the durability of AI spending, these comments provide some evidence that spending growth is likely to continue," Caso said in a note to clients.
He added that Nvidia's deals with OpenAI, Anthropic and other cloud service providers make "consensus numbers appear very conservative."
SpaceX's additional compute capacity will go toward developing its Grok chatbot and AI software tools as it prepares to close its acquisition of AI coding platform Cursor, Musk added. He said SpaceX's Starmind AI satellite "will be essentially an optimized Vera Rubin NVL72 computer," and that the company expects to launch it next year. The rack-scale system connects 72 of Nvidia's Rubin GPUs with 36 of its Vera central processing units.
Comments