Three days ago, Elon Musk announced Terafab, a $25 billion chip factory that would produce 70% of TSMC's entire global output from a single facility. In 2003, he gave a talk at Stanford with 30 employees and a $6 million rocket.
Musk is 32. SpaceX has 30 people. No lawyers. They've test-fired engines and built tank structures, but Falcon 1 hasn't launched yet. He's selling flights for $6 million each. His nearest competitor charges $25 million for less capability. He offers the Stanford audience a rocket if anyone's buying. Then adds: "There's not a lot of viral marketing that's going to happen with a rocket. I'm hoping, but I'm not counting on it."
He starts with Zip2. In 1995, he deferred his PhD at Stanford to start an internet company. VCs on Sand Hill Road hadn't heard of the internet. He had no money. Negative money, because of student debt. He couldn't afford both an apartment and an office, so he rented the office, slept on a futon, and showered at the YMCA on Page Mill and El Camino. "I was in the best shape I've ever been. You get a shower, a workout, and you're good to go." They drilled a hole through the floor to the ISP below them and ran a cable for internet. $100 a month. Six people total: him, his brother, a friend of his mom's, and three salespeople hired from a newspaper ad. Sold to Compaq in 1999 for over $300 million cash. "That's a currency I highly recommend."
Then PayPal. The email payments feature took a day to build. One day. It was supposed to be a side feature. The main product was an all-in-one financial portal that included banking, brokerage, and insurance. They'd demo the portal, and people would go "ho-hum." Demo the email transfer, and people would go "wow." So they dropped everything else. No VP of sales. No VP of marketing. Zero advertising spend. A million customers by year two. Sold to eBay for $1.5 billion.
Then he gets to space.
He says he started researching why space hadn't progressed since Apollo. In the 60s, humanity went from nothing to the moon. Then stopped. Every other technology sector improved by orders of magnitude. Computers went from filling rooms to fitting in pockets. Space went sideways. He wanted to know why.
He tried to buy a Russian ICBM. Three trips to Moscow. "On the range of interesting experiences, negotiating for a refurbished ICBM is pretty far out there." The deal fell apart. On the flight home from the third trip, he asked himself why America couldn't build a cheaper rocket. "It's not like we drive Russian cars or fly Russian planes."
So he started SpaceX. Thirty people. No lawyers. They outsource heavy machining and welding but do all design, analysis, testing, and launch operations internally. He describes the cost philosophy in one line: "There's no one silver bullet. It's really hundreds of small innovations." Ethernet inside the rocket instead of the copper cable bundles as thick as your arm running the full length of every other launch vehicle. Simpler structures, so fewer things break and fewer things to buy. Even if SpaceX did everything exactly the same way as Lockheed or Boeing, they'd still be dramatically cheaper just from having an order of magnitude less overhead.
He tells the room: "The fastest way to make a small fortune in the aerospace industry is to start with a large one."
Then, almost offhand, he describes SpaceX's endgame. The "Holy Grail objective" is to build the successor to the Saturn V to set up a moon base or conduct a Mars mission. He dismisses space mining and space solar power as economically unworkable. The real opportunity, he says, is a self-sustaining civilization on another planet. "Then you've got basically interplanetary commerce going on." Trillion-dollar level.
That was 30 employees and an untested rocket.
Last Saturday, Musk stood in a decommissioned power plant in Austin and announced that Tesla, SpaceX, and xAI would build a chip factory targeting one terawatt of computing output per year. He said all existing chip fabs on Earth produce about 2% of what his companies will need. Eighty percent of Terafab's output would go to space-based AI satellites. Estimated cost: $20 to $25 billion. No construction timeline was given. Tesla's CFO confirmed the cost isn't in the 2026 capital expenditure plan yet.
The SpaceX that showed up at Stanford in 2003 had 30 people and a rocket that cost less than a nice house in Palo Alto costs today. The SpaceX that showed up in Austin last Saturday is co-building the largest semiconductor fab ever attempted. Same guy. Same instinct: if nobody's building it fast enough, build it yourself.
Video: Elon Musk at Stanford's Entrepreneurial Thought Leaders Series, 2003. Original footage from Stanford University.
No comments:
Post a Comment