Post Content Musk’s plans for orbital data centres will start with a future ‘mini’ AI data centre satellite. (Photo: AP/ File)
Elon Musk has unveiled plans to build a massive chip fabrication facility in the United States, aimed at producing in-house semiconductors to power the artificial intelligence (AI), robotics, autonomy, and orbital data centre ambitions of his companies, including Tesla and xAI.
Musk’s so-called ‘Tera-fab’ facility will be located near Tesla’s existing headquarters and gigafactory in Austin, Texas. It will be jointly run by Tesla and SpaceX, both companies with Musk as CEO. The tech billionaire said that the “advanced technology fab” will have all the equipment necessary to manufacture and test chips of any kind, according to a report by Bloomberg.
The Tera-fab facility is expected to eventually lead to a terawatt of computing power per year even as Musk ramps up investments in AI and robotics.
While Musk’s newly unveiled projects have consistently been met with some skepticism, owing in part to his history of over-promising on goals and timelines, his chip production efforts seem to be grounded in industry practice. Despite chip partnerships and supply deals, Musk has said that the semiconductor industry is not moving fast enough to meet the surging demand majorly from Tesla which is doubling down on robotics, autonomous driving, and AI. “That rate is much less than we’d like. We either build the Terafab or we don’t have the chips, and we need the chips, so we build the Terafab,” Musk was quoted as saying.
Tech industry executives have similarly expressed concerns about a shortage of chips amid the race to build computing power for AI. However, not all companies can pursue semiconductor fabrication as such facilities typically require billions of dollars in investment and highly advanced equipment sourced from multiple providers. Whether Tesla, along with SpaceX, can execute such an undertaking remains the key question.
The EV giant’s existing suppliers include TSMC and Micron. It also has an agreement with Samsung for its chips to be produced at the South Korean giant’s facility near Austin.
Tera-fab chip plans
The upcoming Tera-fab will reportedly produce 2-nanometer chips. The facility is expected to support 100 to 200 gigawatts a year of computing power on Earth and a terawatt in space. Tera-fab will be used to make two types of chips: One that is optimised for edge and inference. These chips could be used in Tesla’s robotaxis and Optimus humanoid robots.
Story continues below this ad
Also Read | Leadership exits, Grok controversy: Why Elon Musk wants xAI to start over again
The other type will be high-power chips designed to be used in orbital data centres by SpaceX and xAI, which is now a wholly owned subsidiary of SpaceX. The vast majority of chips will be used by xAI, according to Musk, who did not provide timelines for when the facility will be built and begin production.
Orbital data centres
Musk’s plans for orbital data centres will start with a future ‘mini’ AI data centre satellite with the capacity to deliver 100 kilowatts of computing power. This mini data centre is just one piece of a much larger network of orbital data centres, each of which could be scaled to megawatt-range of computing power, according to the tech CEO.
He also reportedly has plans to potentially launch satellites from the surface of the moon as part of his loftier vision for a future filled with “amazing abundance”.
Also Read | Sam Altman vs Elon Musk: Why OpenAI’s CEO says space-based data centres won’t matter this decade
In January this year, Musk approached the US Federal Communications Commission (FCC) for a licence to launch one million data centre satellites into orbit around Earth. Orbital data centres are also a crucial part of SpaceX’s planned IPO, with the company expected to raise as much as $50 billion at a valuation of more than $1.75 trillion, as per Bloomberg.
