Elon Musk and Sam Altman are not only battling it out in the courts over Musk’s stake in OpenAI, but they’re also battling it out in their views around how AI will be scaled in the coming years
Sam Altman has appeared to dismiss the idea of data centers in space being a reality in the foreseeable future. Musk, on the other hand, is a big proponent of the idea — he’s merged xAI and SpaceX to help accelerate their development, and believes they’ll be a reality in less than 3 years. Altman, however, doesn’t seem as convinced.
“Do you think space datacenters will provide a meaningful amount of compute for OpenAI in the new two to three years?” Altman was asked on the TBPN podcast. “No,” Altman decisively replied. “5 years?” the host countered. “No,” said Altman, just as decisively. “10 years?” the host offered. Altman laughed, before seemingly sarcastically saying “I wish Elon luck”.
Altman seemed to be saying that he didn’t believe that AI datacenters needed to move to space in the near future. Musk, on the other hand, strongly believes that the most viable AI datacenters will be in space, thanks to the ready supply of solar energy, lower cooling requirements, and government regulations slowing down the pace of energy development in western countries.
Musk however took notice of Altman’s comments. “He’s right…for OpenAI,” Musk said on X, hinting that OpenAI wouldn’t be able to put any datacenters in space, which would be a disadvantage for the company in the coming years.
Elon Musk isn’t the only one who believes that space datacenters are the future. A startup named Starcloud, backed by Google and NVIDIA, has been working on the idea for a few years, and has even trained an AI model in space using an NVIDIA GPU. Google has launcher project Suncatcher, through which it aims to send TPUs to space by early next year to scale its AI compute. And Jeff Bezos, who runs space company just like Musk, has said that space datacenters will be a reality in 20 years. It remains to be seen if space datacenters take off (pun intended), but two of the biggest AI companies clearly seem to have extremely diverging views on their viability, and this could have big implications for the AI race in the years to come.