Tech • AI • Robotics • Game
SpaceX Is About to Do Something Impossible
9.4/10TL;DR
Elon Musk outlined aggressive projections for SpaceX, Tesla, and U.S. chip manufacturing, including SpaceX revenue above $100 billion annualized by December, possible $500 billion ARR next year, and a long-term push to cut launch costs and massively expand AI chip production.
KEY POINTS
SpaceX revenue targets
Musk said SpaceX is on track to reach an annualized revenue run rate above $100 billion by December, barring a major disruption. He added that annual recurring revenue could exceed $500 billion by December next year, while stressing that this was a forecast rather than a promise. He also said $1 trillion in ARR is no longer impossible, though he put the odds at only about 2% for now.
A jump from last year’s scale
The company’s revenue was described as roughly $15 billion to $16 billion last year, making the projected move to more than $100 billion unusually steep even by high-growth standards. A rise to $500 billion would place SpaceX among the world’s very largest companies by revenue, near the scale of recent leaders such as Amazon, Walmart, and State Grid. The comparison underlines how extraordinary the forecast is relative to current corporate rankings.
Dependence on Asian chip manufacturing
Musk warned that advanced chip production remains heavily concentrated in Asia, especially at TSMC in Taiwan and Samsung. He said even optimistic assumptions about new U.S. fabs suggest American capacity would amount to only about 10% of Asian output. That leaves U.S. industry exposed both to geopolitical risk and to broader supply constraints in AI hardware.
Tesla’s U.S. chip ambitions
He said TSMC is expanding in Arizona, while Samsung has operating fabs in Texas and is building or has built a fab dedicated to Tesla’s AI5 and AI6 chip designs. Tesla is also building a research fabrication lab at Gigafactory Texas in Austin to test multiple chip-manufacturing approaches before scaling them at a larger production site in Grimes County near Texas A&M. The goal is tighter control over design, packaging, lithography, and testing under one roof.
The “terafab” goal
Tesla is targeting what Musk called a “terafab,” aimed at eventually producing 1 terawatt per year of AI chips. He said global AI chip production this year is around 20 gigawatts to 25 gigawatts, with next year estimated at 35 gigawatts to 45 gigawatts. Reaching 1 terawatt would therefore require around 30 times projected global output for next year in a single system.
Faster chip iteration
A central claim was that vertically integrated manufacturing could radically shorten development cycles. Work that normally takes a year could potentially be reduced to a month, allowing much faster chip design, fabrication, testing, and revision. If achieved, that would represent a major shift in semiconductor iteration speed.
Why space-based computing is being considered
Musk said AI growth at the 200-gigawatt to 1-terawatt scale may be impractical on Earth because of power limits. He noted average U.S. electricity consumption is just over 500 gigawatts, making annual additions of hundreds of gigawatts for AI alone unrealistic. That is driving interest in moving large-scale compute into orbit, where solar generation is constant and stronger due to the lack of atmospheric losses.
Early tests in orbit
SpaceX has already launched satellites carrying an NVIDIA GPU and a Google TPU, and both were described as operating successfully in space. Musk said the first StarMind computers, co-designed with NVIDIA, could begin launching in the third quarter of next year. He expects slow initial deployment, followed by significant volume in 2028 and very large volume in 2029.
Starship and launch economics
He argued that Starship, if it becomes the first fully reusable rocket, could lower the cost of access to space by a factor of 100 or more. He said NASA estimated orbital launch costs at about $18,500 per kilogram in 2010, while reusability with Falcon 9 cut that to around $2,500, then $1,500. His target is eventually under $100 per kilogram to orbit.
Robotics and economic output
On humanoid robots, Musk argued that a robot working 168 hours per week could deliver roughly five times the weekly productivity of a person working around 40 hours. Using that logic, he estimated the U.S. goods-and-services output of roughly 120 million productive workers could be matched by 20 million to 30 million robots. He said a robot may initially cost about the same as a car, around $30,000, before becoming cheaper over time, with rental as a likely business model.
CONCLUSION
The projections combine extraordinary revenue expectations with a broader strategy linking reusable rockets, domestic chip fabrication, orbital computing, and humanoid robotics. If even part of that agenda materializes, it would imply a major shift in the scale of both SpaceX and Tesla within a very short period.
Ask a question