SpaceX (NASDAQ:SPCX) and Tesla Inc. (NASDAQ:TSLA) CEO Elon Musk could eventually need around $1 trillion of chip-manufacturing investment to feed his companies’ AI ambitions, according to ARK Invest.
ARK chief futurist Brett Winton said on the firm’s Brainstorm podcast that reaching Musk’s goal of producing one terawatt of computing hardware a year at Terafab, his planned chip complex in Texas, could require about $1 trillion in capital.
SpaceX and Tesla said in August they would initially invest $16.8 billion in Terafab. Earlier filings showed investment could reach $119 billion if SpaceX builds all planned phases.
Why Musk Needs So Many Chips
Much of that demand could come from SpaceX. Musk estimated roughly 75% of Terafab’s eventual AI chip output could go to SpaceX’s AI spacecraft, with about 25% going to Tesla’s Optimus robots.
SpaceX says its orbital AI plans will require far more advanced chips than it can currently access. During a March presentation in Austin, Musk said existing global production would meet only a small fraction of his companies’ future needs.
“We either build the Terafab or we don’t have the chips,” Musk said.
That shortage sits at the center of Winton’s argument. As SpaceX drives down launch and other satellite costs, he says advanced chips could become the majority of an AI satellite’s cost, leaving Musk to either manufacture huge volumes himself or keep relying heavily on outside suppliers.
That could still benefit Nvidia Corp. (NASDAQ:NVDA). Winton speculated Nvidia may view Musk as an “infinite demand generator” and offer his companies better economics as their chip requirements grow.
Musk has separately said SpaceX and Tesla will continue ordering Nvidia chips “at scale.” Terafab also plans to use Intel’s 14A manufacturing process.
Could $1 Trillion Be Too Low?
The $1 trillion figure remains highly speculative. SpaceX has not committed anywhere near that amount, and later spending remains unsettled.
But Musk’s target shows why the numbers get so large. SpaceX says future chip demand across SpaceX and Tesla could exceed one terawatt of computing power, driven partly by plans for thousands of AI satellites.
Bernstein estimates producing enough chips for that scale of compute could require $5 trillion to $13 trillion in industry investment, though that figure covers far more than Terafab alone.
One near-term gauge of compute scarcity is the rental market. Nvidia’s B200 currently rents for about $6.28 per GPU-hour on the Ornn Compute Price Index.
Polymarket traders put the most likely year-end range at $5.50 to $6.00, suggesting some easing in compute costs by year-end.
For now, traders are betting the AI chip squeeze will ease before Musk’s trillion-dollar buildout takes shape.
Image: Shutterstock
Login to comment