A16Z: Sun Testimony, SpaceX worth 7.5 trillion
Original by:@mikemcg0 and @pmarca
Original language: Woo XI
Mask's pay program in SpaceX revolves around two objectives。
The first objective is: if the company ' s valuation is achieved$7.5 trillionand create at least one on Mars1 million populationHe will receive the first prize from the permanent human colony。
Second objective: if SpaceX operates data centres in space and makes them consume at least100 wattsIt's more than all data centres on Earth1000 times- He'll get the second prize。
If neither goal is achieved, Mask won't get anything except what he's got since 2019US$ 54,080 annual salaryI don't know。
The members of the Board of Directors who signed the scheme have been witnessing one thing for the past two decades:
Mask made the SpaceX predictions again and again, which seemed impossible, and then they became reality again。
He said SpaceX would put humans in orbit -- no private company had done it before. Now SpaceX has been regularly converted to NASA to transport astronauts。
He said SpaceX would land orbital rockets and repeat them. Until then, boosters were considered a one-off consumer throughout the industry. Now SpaceX has done hundreds of recyclings。
He had said that a satellite Internet operation could be worth tens of billions of dollars — until then the satellite Internet was almost a graveyard for bankrupt businesses. Now, Starlink's income has grown from zero in a few years$11.4 billionI don't know。
These projections, which are often radical on the timetable, have almost never been missed in the direction。
And SpaceX wrote in 2002 that the original mission was to make human beingsMultiplanetary speciesI don't know。
So the board tied his pay to the mission itself。
If this mission sounds like science fiction, it may be because it does come from science fiction。
Iain M. Banks and Civilization Blueprint
It took me 25 years to write a nameThe CultureThe world。
From most reasonable criteria, that may be the best utopian society ever imagined。
There, humans are calledMindsThe super smart AI to live together. Minds runs a huge orbital habitat like the small world. The human relationship with AI is not slavery or competition, but partnership。
No one is forced to work。
No one's hungry。
Minds carries the amazing load of calculations needed to run space cities。
And human beings are responsible for continuing to be human。
It was a full-time job。
SpaceX, three autonomous drones for the recovery of Falcon 9 boosters at sea, all named out of Banks' conscious starships:
-
Of Course I Still Love You -
Just Read the Industries -
A Shortfall of Gravitas
IN AN INTERVIEW AT THE UNITED KINGDOM ' S AI SECURITY SUMMIT IN 2023, MASK WAS ASKED: "A GOOD AI, WHAT SHOULD BE THE FUTURE?"
He replied:
"Banks' Civilization Series is the best idea of the future of AI so far. Nothing comes near it, and it makes you understand what a rather utopian, or progressive utopian, AI's future."
He's been telling us in the name of those landing platforms: what he wants to build。
Figure:"Of Course I Still Love You"Catch Falcon 9 booster on April 8, 2016. This is the first time in history that a successful unmanned ship has landed, and it is also a time when orbital stage space can be used again to move from theory to reality. The name of the ship came from a conscious starship in the Iain M. Banks Civilization novel series. (Photo: SpaceX)
But Civilization is not a paradise without friction。
Banks' novels are full of war, conspiracy and moral complexities. It is utopia because civilization has resolved the pre-conditions of survival to the point that hundreds of millions of human beings can finally take care of what Banks calls "the real importance of life":
Sports, games, love, studying dead languages, barbarous societies, impossible problems and climbing mountains without safety nets。
There are four premises for such a future。
First, to gain a meaningful part of a star's energy output — many orders of magnitude higher than the energy generated by human civilization today。
Secondly, large-scale physical intelligence: machines can build, mine, refine, repair anything without human intervention and can do it anywhere。
Thirdly, cheap digital intelligence, and that intelligence exceeds bio-intelligence。
Fourthly, the quality can be transported out of the Earth cheaply, frequently and reliably. Because of all the above, the Earth alone cannot expand。
From the future
Most analyses of SpaceX are being pushed forward from now:
Rockets, satellites, contracts, revenues。
But if you want to see what really happened, it would be more useful to start at the end and push back。
Mars City
At the operational level, the objectives are:
In the life cycle of today's living generation, build one on Mars1 million populationThe self-sufficient cities。
The real difficulty is “self-sufficiency”。
That means that if the planet stops sailing, the city must still live。
It must make everything itself:
Food, water, air, energy, medicine, machines and, ultimately, more human beings。
According to SpaceX ' s own calculations, sending 1 million people and millions of tons of cargo to Mars over decades would require thousands of Starship flights; more than 10 launches per day during each transfer window。
These windows are determined by Earth-Mars orbit mechanics, which is only a few weeks wide and every time26 monthsJust once。
Figure: SpaceX rendering of the city of Mars. (Photo: SpaceX)
Moon City
Moon cities are closer and easier to rehearse。
The permanent shading pits in the Antarctic of the Moon contain ice and some ridges have sustained solar exposure, so it is natural to establish bases there。
But Mask is talking about more than just a scientific outpost。
HE ENVISAGES BUILDING FACTORIES ON THE MOON, PRODUCING AI SATELLITES, AND USING THEMMass DriveLaunch them one by one into space。
The mass drive is also a concept that Mask borrowed from science fiction. It is an electromagnetic launch system that, using the gravity of only one sixth of the Earth and an atmosphere-free environment, throws solar satellites into deep space on an industrial scale。
These satellites can be made on the moon because the moon is about by weight20% siliconand10% aluminium— This is the two main inputs for solar cells and satellite structures。
"If you want to go beyond 1 watt a year, you have to go to the moon."
Figure: SpaceX mass drive rendering in Moonbase Alpha for launch into orbit of the lunar-made AI satellite, the data centre. (Photo: SpaceX)
Orbital data centre
Mask bets:
IN A FEW YEARS, THE BEST PLACE TO PLACE AN AI DATA CENTRE, ECONOMICALLY SPEAKING, WILL BE SPACE。
AI'S BOTTLENECK IS ENERGY. WITH THE EXCEPTION OF CHINA, ENERGY SUPPLY GROWTH HAS BEEN VERY LIMITED, WHILE AI ' S COMPUTING DEMAND IS GROWING EXPONENTIALLY。
The solar panels in orbit, which provide the same power as the solar panels on the ground4 to 10 timesI don't know. The number of specific multipliers depends on how clear the ground position is。
The reason is simple:
There is no atmosphere in space, no cycle of day and night, no clouds and no season。
NASA MADE IT CLEAR DECADES AGO. NOW, THE ROCKET IS FINALLY CHEAP ENOUGH TO MAKE IT A REALITY。
Mask predicts that five years later, SpaceX's annual launch into orbit of AI algorithms will outpace the cumulative load on Earth。
That's why SpaceX merged with XAI in February。
Rockets and intelligence are becoming the same problem。
Starship: all means of transport upstream
Starship is the vehicle that makes everything possible upriver。
First flight this yearStarship V3It is the largest and most powerful rocket ever built. It is higher than a 40-storey building and pushes twice as much as Saturn V, who sent astronauts to the moon that year。
ACCORDING TO NASA, THE HISTORICAL COST OF GOING INTO ORBIT IS ABOUT PER KILOGRAMUSD 18,500I don't know。
In 2010, the first Falcon 9 reduced the cost by about 85% to about a kiloUSD 2,700I don't know。
In 2018, Falcon Heavy dropped it to about a kiloUSD 1,400I don't know。
Starship, the world's first fully and rapidly reusable spacecraft, aims to further reduce costs to 1 kgFrom $100 to $500I don't know。
A past launch of billions of dollars in space is becoming a tens of millions of dollars in volume business。
Starlink: Cash wheel
Starlink is helping to pay for everything else in cash。
According to SpaceX's IPO file, connect business segments -- almost all Starlink -- In 2025Income of $11.4 billion, same year growth50%EBITDA PROFIT MARGIN EXCEEDED60%I don't know。
As of March 2026, Starlink was164 countriesPossession10.3 million users, run in excess9600 satellitesUp。
Starlink was initially a side project to fill the company ' s own launch capacity; it is now becoming one of the great consumer-level businesses in history。
In 2019, when a16z did everything about SpaceX, several people told us that the economic model would never work。
The reason for this is that Starlink's terminal antenna requires antenna technology that was previously used only for F-22 fighters and naval destroyers and that was never oriented towards large-scale consumer production。
The first terminal manufacturing cost for SpaceX is about$3000It's only for sale$499I don't know。
However, they eventually hit down the manufacturing costs and proved the suspect wrong。
Falcon 9: Power to buy time for the future
Falcon 9 is the main rocket for all the other time。
It is the only orbital stage booster on Earth that is used repeatedly on a large scale. Prior to decommissioning, individual boosters usually fly more than 20 times。
In 2025, SpaceX launched the total global in-orbit mass83 per centI don't know。
Despite the pre-emptive advantage of half a century of everyone else, SpaceX is now launching into orbit payloads that exceed the sum of all the other forces in the world。
This is the stack from top to bottom。
After generations, the future of civilization lives at the top。
Falcon 9 and Starlink sit at the bottom and pay today's bill。
Each floor makes the next possible。
SpaceX CFO Bret Johnsen describes how it feels inside the company:
“[Maske] creates a culture: you set a goal that initially seemed extremely ambitious, and then, step by step, you find yourself moving towards something that is absolutely achievable..
Take Mars, for example. When I first arrived in 2011, when you talked about Mars and multiplanetary species, people turned their eyes. Now that we're talking about it, the reaction has been: "What year?"
I think the best thing about Elon is that he set these goals and built an excellent business model around every piece of intellectual property required to achieve the ultimate goal."
Stupid index and algorithm
Mask didn't want to start a rocket company。
In 2001, 30-year-old Mask was thinking about what to do after selling PayPal。
HE'S ALWAYS BEEN INTERESTED IN SPACE. WHEN HE WENT LOOKING FOR NASA'S PLAN TO SEND HUMANS TO MARS, HE WAS SURPRISED TO FIND OUT THAT THERE WAS NO SUCH PLAN。
So he designed a plan:
Sending a small greenhouse to Mars and sending pictures back to Earth。
His idea was that if people saw a green bud appearing on a dead red planet, they might rekindle public interest in space and inspire political will to finance a real Mars project。
He just needs a rocket to send the greenhouse over。
Later that year, he went to Moscow to buy a refurbished ICBM. This was the first of two trips to Moscow。
The meetings were said to be full of vodka and gestures。
Mask's best friend, Adeo Resi, from the University of Pennsylvania. He told Esquire in 2012:
"We all go into a small room and everyone has a bottle of wine."
The Russians didn't take Mask seriously。
On one occasion, a general designer even spit on Mask and his team as a sign of contempt。
The second trip to Moscow was in February. Mask asked how much for a missile。
The other side said: each$8 millionI don't know。
Mask says: two eight million dollars。
Jim Cantrell, a space consultant in Musker, remembers that he said something like that:
"Young man, no."
And hinted he had no money at all。
Mask decided they weren't serious, so he got up and left。
Cantrell thought the trip was over。
On the return plane, he ordered wine with Mike Griffin, who later served as director of NASA and also as a consultant, and finally celebrated his departure from Moscow。
Mask sits in front of them, bows to the laptop。
Then he turned around:
"Hey, guys, I think we can build this rocket ourselves."
He showed them a spreadsheet listing the raw materials required for the rocket: aluminium, titanium, copper, carbon fibre, and the cost of each material。
These materials cost only the price2%I don't know。
As Mask later said:
“Evidently, you just need to come up with a smart way to combine these materials into rocket shapes.”
Within months, Mask decided to take it out$100 millionTo venture into a rocket company. That's more than half the $180 million he got from PayPal。
SpaceX was set up in a warehouse in El Segundo, California。
He extended a founding team invitation to five people。
Three people refused, including Cantrell and Griffin。
The two promised are:
-
Tom Mueller, later Vice-President of Propulsion Systems, also employee of Company 1 -
Chris Thompson, employee number two, in charge of operations and production。
Mask then joked:
"SpaceX in 2002 was basically a carpet and a Mexican homeless band. That's it. As you can see, I'm a dancing machine."
Many years later, Mask called the diagnostic tool behind the spreadsheet"idiot index"I don't know。
If the cost of a spare is high relative to the cost of raw materials, either you're a fool or you're working with a fool。
It sounds like a joke, but it's the basis of SpaceX strategy。
Each of the spare parts purchased by SpaceX was accompanied by a “wit index” calculation。
One of the most legendary stories of the early days of the company happened to Steve Davis。
Davis joined SpaceX directly after Stanford, the company's 14th employee. His task was to procure an implementer to control the upper stages of the Falcon 1 rocket。
When he reported, the traditional space supplier asked for the price$120,000When Mask laughed。
Mask told him that this part was no more complex than a garage door opener and gave it to him$5,000Budget, let him make it from scratch。
The biographer Ashlee Vance recorded that Davis had spent nine months grinding his design over and over and over and over and over and over and over and over and over and over again, and finally made an implementer that could only be used at a costUS$ 3900I don't know。
When Davis sent the technology for this victory to Mask, Mask only returned two letters:
“OK.”
To push the fool index below the theoretical threshold, you have to integrate vertically and control processes from end to end。
But vertical integration has fixed costs and can only be costed with high yields。
In order for the rocket business to achieve high yields, it must break the way the industry has been operating。
Traditional launch service providers, such as ULA and Arianespace, treat each mission as a custom project。
Clients specify orbits, payloads and integration requirements, and launch service providers design a customized set of tasks around satellites。
The model assumes that:
Only a few launches per year, each mission being extremely costly。
It makes scale manufacturing impossible。
SpaceX does the other way。
They issued a Falcon user guide defining the exact specifications of the rocket and informing the client that:
Please design your satellite to fit our rockets。
At the time, this was considered very radical and deprived SpaceX of some early business。
But it opens up the manufacturing wheel。
Standardization and reuse are mutually reinforcing。
Because each Falcon 9 is the same, the recovered booster can be transformed into a finished, qualified, ready to fly again。
The first two flights of the Falcon 9 booster were completed in 2017。
By 2020, a single booster can fly five times。
By 2021, we could fly ten times。
Today, the record keeper is flying35 timesI don't know。
This reuse changes space economics and it is difficult to see how competitors can catch up。
In 2021, Mask estimated that Falcon 9 would, under optimal circumstances, place 15 tons of payload in orbit at marginal launch costs (excluding indirect costs) of approximately$15 millionI don't know. He said it was about the cost of an alternativeOne in two to one in threeI don't know。
Today, SpaceX launches one rocket every two to three days with a re-use booster, while competitors can only launch a few customized rockets a year。
But SpaceX has advantages beyond economies of scale, vertical integration and better strategies。
It also has speed and culture。
Traditional space companies use analysis to remove uncertainty。
NASA USED A POLITE LANGUAGE TO DESCRIBE THE BOEING COMMERCIAL MANNED SPACE PROJECT:
“Advanced systems engineering methods are used to pre-empt engineering research and analysis to mature system design before manufacture and testing.”
Count twice, cut once。
SpaceX reverses this order。
Companies create many cheap prototypes, push them towards failure, learn from failure and repeat them quickly。
The Starship test project, which produced a spectacular explosion, could exceed any rocket project in history。
But each failure is the data point where reality and model deviations occur。
This comparison is very clear to those who have worked simultaneously in two worlds。
Garrett Reisman is a NASA astronaut who has carried out two shuttle missions. In 2011, he left NASA and joined SpaceX as a senior engineer。
He described the mainstream view of NASA on SpaceX:
"They're cowboys; they're dangerous; they're killing people."
But what really changed his mind was seeing how SpaceX worked。
"THEY MAKE THINGS UP A MONTH, NASA COULD TAKE A YEAR. WE WERE SHOCKED."
The clearest example is the Falcon 1 project。
Between 2006 and 2008, SpaceX launched four Falcon 1 rockets on a small atoll in the Pacific called Kwajalein。
The first three failed。
But each failure is different and offers learning:
-
First, fuel leak -
Second, propellant shakes unusual -
Thirdly, the residual engine thrust resulted in a separation collision。
By September 2008, the company had only enough money to launch another launch。
And it's not the only company in Musker standing on the edge of the cliff。
He's also building an electric car company, Tesla, and only a few weeks from bankruptcy。
He had to decide whether to put the remaining PayPal cash in a single company or to divide it between two companies。
Mask remembers:
“That was a very difficult decision. Finally, I decided to separate my remaining money and try to keep both companies alive. But that could be a very bad decision, leading to the deaths of both companies。
I never thought I'd have a nervous breakdown, but I was really close."
He could not choose because, in his world view, both missions were indispensable:
Tesla wants to accelerate the world's transition to sustainable energy。
SpaceX wants to make humans multiplanetary species。
In the BBC documentary The Elon Musk Show:“All available resources must be invested in the company. He gave me the chance to quit. He said, "This is going to be the hardest part, and you don't have to stay with me."
Figure: In 2006, Elon Musk visited the remains of the first Falcon 1 on Omelek Island. Photo by Hans Koenigsmann
The fourth launch succeeded。
In December of that year, the weeks before SpaceX was about to run out of cash, NASA awarded the company a copy$1.6 billionfreight contract。
WHEN NASA CALLED MASK, HE WAS HIT BY A MASSIVE EMOTIONAL RELEASE, WHICH CAME OUT:
"I love you."
The model formed from a quick failure, a quick error correction, became the culture of SpaceX for every project。
That is why SpaceX today can quickly overlap between two test flights in Starship, while traditional space projects, ranging from one flight anomaly to a redesign of aircraft, often take years。
This approach is more effective than alternatives because:
On issues that you have not yet fully understood, you cannot be perfected by thinking alone。
Reality is the only sufficiently qualified certifying authority。
The key is to keep the cost of applying to reality sufficiently low to allow frequent attendance。
An algorithm for SpaceX
The above is an iterative cycle of SpaceX through storytelling。
But it also has a written version。
Over the past two decades, Musketeer has coded SpaceX into a five-step business process known as the companyAlgorithmI don't know。
Tim Berry has been working at SpaceX for 10 years, leading a production team of Falcon 9 and Falcon Heavy. He said that the method was "blowed into our heads."。
Walter Isaacson published his standard version in the Musk biography:
1. Challenge each request
Each request should be accompanied by the name of the person making the request。
You should not accept that “the request came from the legal authorities” or “the request came from the security services”。
You need to know who the real claimant is, and no matter how smart he is, question it。
The demands from wise people are the most dangerous, because people are the least willing to question them。
Then make these demands less stupid。
2. Delete all removable parts or processes
You might have to put them back later。
In fact, if you don't eventually add at least 10 percent of what was deleted back, it means you haven't done enough。
3. Simplification and optimization
This step should follow the second step。
A common mistake is to simplify and optimize a part or process that should not exist。
4. Speeding up the cycle
Each process can be accelerated。
But this should be done only after the first three steps have been completed。
Mask said he had made a mistake at the Tesla factory: it took a lot of time to speed up certain processes, and then realize that they should have been deleted。
Automation
Automation at the end。
Tesla's mistakes at the Nevarda and Fremont factories were to try to automate them from the beginning, not to question requests, remove parts and processes, and shake the bug out。
Most engineering organizations will jump directly to step 5。
They automate a process that should not have existed。
SpaceX runs these steps sequentially every time, in every part of the company。
When an algorithm runs enough times on a particular hardware, it begins to look like anything in the industry。
Figure: SpaceX Raptor engine 3 generation product, V1 to V3. (Photo: SpaceX)
Raptor 3, is the product of a team around the same engine 10 years later。
It produces more than Raptor 222% pushWeight reduction40%And don't need a heat shield。
THE REASON IS THAT THE PIPES AND BEAMS THAT WERE HANGING OUTSIDE THE ENGINE IN THE PAST HAVE BEEN INTEGRATED INTO THE ENGINE ' S METAL STRUCTURE THROUGH 3D PRINTING。
Mask said:
“Simplified Raptor engine, internalized secondary flow, and increased regenerative cooling of exposed components are alarming. It's near the limits of known physics."
In space history, no engine projects are known to be occurring so quickly。
In the last three decades, the main engine of the Space Shuttle was basically the same design。
The RD-180 driving Atlas V is a derivative version of an engine designed in the 1970s。
And SpaceX, in less than a decade, has done three completely new designs on Raptor, and each edition is much better than the last。
The same philosophy applies to people。
By mid-2018, Falcon 9 had returned to a reliable rhythm, and Mask turned its attention to a satellite Internet constellation that would eventually finance all upstream work。
The Starlink team is located in Redmond, Washington, and many senior engineers come from Microsoft, developing a slower pace than Mask wanted。
In June, he flew to Redmond, firing the senior leadership。
He then transferred young star engineers from the rocket sector and gave them one year to launch the first operating satellites。
This is a cruel way of managing companies. According to media reports of dismissals, the sector appears to be falling apart。
However, 11 months later, in May 2019, the first Starlink satellites were launched。
Mask cleared the bottlenecks and turned to the next problem。
He manages everything like this。
In 2018, Tesla was in Model 3 mass production of “the production hell”, and the rate of burning money was already threatening to survive. Mask really moved into the factory。
Years later, he remembered:
"I lived in Fremont and Nevarda factories for three years in a row. I slept on the floor under the desk so that the whole team could see me on the shift。
This is important because if the team thinks that their leaders are enjoying themselves elsewhere and drinking on the tropical islands, Mai Tai, they will be demoralized。
Because they saw me sleeping on the floor when they changed shifts, so they knew I was there. It's a huge difference, and they're doing their best.”
Later, he turned it into a corporate rule:
The higher the position, the greater the sense of presence。
TO FIND A PERSON WHO CAN OPERATE IN A WAY SIMILAR TO THE MASK CEO, IT IS NECESSARY TO RETURN TO THE LATE 19TH CENTURY AND EARLY 20TH CENTURY INDUSTRIAL TIMES:
Henry Ford, Andrew Carnegie, Thomas Watson, Andrew Mellon, Cornelius Vanderbilt。
The uniqueness of Musk's business style is his relationship to work。
He is said to appear weekly in each of his companies to identify the biggest problem and then repair it。
52 weeks a year, he does it every week。
So theoretically, every company solved 52 of the biggest problems in that year。
An engineer who joined SpaceX from another space company described this experience as:
"It's like being thrown into an alarming power zone. Everyone around you is absolutely competent."
constellations
SpaceX looks like a company。
But it is more useful to see it as the central node of a corporate constellation。
These companies are run by the same person, are built on the same long-term mission and are almost impossible to separate from one another。
Over the past 20 years, Mask has assembled a group of companies. Each company is addressing a constraint that otherwise would become a bottleneck for others。
Now they're getting back together。
SpaceX's merger with XAI in February is a microcosm of what SpaceX is becoming。
If the algorithm finally goes into orbit -- this is Mask's bet -- then SpaceX has the most credible path to deploy it at the size required by AI。
Putting quality on track, as well as large-scale production intelligence, is probably the two most critical capabilities in the coming decades。
Now they reinforce each other under the same roof。
xAI brings Grok, a cutting-edge model, and has a unique position in real-time information by accessing X's real-time data streams。
It also brought in engineers who built Colossus 1 and Colossus 2 supercomputers. These engineers are faster than many people in the industry think。
Figure: Colossus 1. (photo: xAI)
The building of Colossus is worth stopping and looking。
xAI took over an old factory in Memphis, and122 daysDelegate$100,000 GPUStart training。
It's all used when the shelf starts to arrive19 daysLet the whole cluster run。
YOUNG WEIDA, CEO, WONG IN-HOON SAYS THIS:
“From conception to the construction of a huge, liquid-cold, electric-powered, licensed plant, to be completed in that time, it is Superman。
As far as I know, only one person in the world can do this。
WHAT THEY DID WAS UNIQUE. NO ONE HAS EVER DONE THAT. 100,000 GPUS, AS A CLUSTER, COULD EASILY BE THE FASTEST SUPERCOMPUTER ON EARTH IN 2024。
This usually takes three years to plan, then to deliver the equipment, and one more year to make everything work.”
A project that takes at least four years for the rest of the industry took four months for the Mask and XAI teams。
This May, Anthropic agreed to pay SpaceX monthly$1.25 billionBuy the full calculus of Colossus 1。
In a few weeks, in an IPO file revision, SpaceX discloses Google will pay monthly$920 millionAnd getting $110,000 GPU access is about half of what Anthropic gets。
The sum of the two transactions represents approximately each year$26 billion in incomeI don't know。
And it's just two customers, a SpaceX payoff for a business that didn't exist before the absorption of xAI earlier this year。
Chips, electricity and land are scarce。
SpaceX is becoming one of a few companies with adequate AI infrastructure that can both rent off-the-book and pursue its own ambition to build a front-line model。
xAI obtained from SpaceX is a more durable solution that Mask believes will limit the electricity constraints of AI in the coming years。
To produce enough electricity to meet his expected intellectual needs, it requires grid expansion, new power plants and multi-year licensing processes, an industry that does not have that much time。
In his view, orbit solar energy was the way out, as it was almost unlimited。
And SpaceX is the only company with the means of transportation that can be sent on a large scale。
His correctness is one of the most important open questions in the field of science and technology。
But the IPO file for SpaceX shows that the company takes this bet very seriously: It expects AI to be the largest market for companies in the future and far larger than other markets。
The company's space business is almost like a round error compared to these ambitions。
Tesla: Another core puzzle in the constellation
Tesla is another important part of this constellation。
It's integrated with SpaceX in another way。
Tesla and SpaceX share founders, talent pools, operating culture, and increasingly overlapping technology road maps。
Tesla offers three things to SpaceX-XAI this side of the constellation。
First, the chip。
AI5, AI6 and Dojo3 are all Tesla internal designs。
Mask has made it clear that these chips are not just designed for cars, but are components of a broader constellation computing stack。
A5 HANDLES AUTO-DRIVING REASONING。
AI6 Designed for Optimus and AI data centres。
Dojo3 complements the planned AI7 and works for orbital algorithms。
Number two, robot。
Tesla's bet is that Optimus will become an AI layer of physics for factories, warehouses, families, for environments that want to operate without human labour, and eventually for the cities of Mars and the Moon as envisaged in Mask。
Third, solar energy。
Mask says Tesla and SpaceX are heading towards each other each year100 GiwaSOLAR CELL CAPACITY-BUILDING TO SUPPORT THE CONSTRUCTION OF THE EARTH AND AI IN ORBIT。
And then..TerraFabI don't know。
In April this year, Tesla disclosure began ordering equipment for a research semiconductor crystal round factory in the company Giga Texas。
Mask told investors at Tesla's first quarter of 2026:
"We expect it to be about $3 billion, possibly thousands of crystals a month."
SpaceX provides an additional, much larger facility with a monthly design capacity when matureOne million crystalsI don't know。
The reason for this is that no existing round mill can expand at the speed envisaged by Mask。
And the measure which he conceived was measured in Giwa。
"THIS IS NOT WHAT WE PROMISED TO DO. THAT'S WHAT WE'RE TRYING TO DO, AND WE THINK IT'S POSSIBLE: BY THE END OF NEXT YEAR, WE'LL HAVE AN ANNUALIZATION RATE OF ABOUT 1 GIVA PER YEAR IN SPACE AI ALGORITHMS。
Then, from the perspective of the vision, one order of magnitude is expanded each year。
In other words, two and a half years later, the space age is 10 giva. Three and a half years later, maybe 100 giva。
And then, depending on global chip manufacturing and the progress of TerraFab, it continues to go beyond this scale to 1 watt a year, or 1,000 gueva。
That's twice as much of America's electricity consumption."
Figure: TerraFab design target for SpaceX is to reach 1 watt-out per year, approximately twice the current electricity consumption in the United States. (photo: terafab.ai)
The comparison of Mask with gold-plated industrialists did capture something real, but also pointed to differences。
Carnegie built steel。
Vanderbilt builds railways。
Each of them dominated a sector of the industrial base of that era。
Mask tried to work in several sectors simultaneously:
Space, energy, artificial intelligence, robotics, tunnels, brain interfaces, auto-driving cars。
And turn them all to a goal that most people think is fantasies。
It is indeed unknown whether all this will work; many of them may not succeed。
But this attempt has no historical precedent in itself and may become a staging ground for another century。
SpaceX opened the world
Prior to decommissioning in 2011, the cost of putting 1 kg of cargo into orbit was approximatelyUSD 54,500I don't know。
And when Starship matures, Mask expects to fall to a kilo$100I don't know。
When the cost of entering space has fallen more than500 timesEvery industry that can theoretically exist in space is beginning to be economically viable。
Such industries are a lot。
Figure: Starship and Super Heavy were designed to return to the launch site after flight and were taken over by the launching tower, thereby enabling quick re-issuance without renovation. (Photo: SpaceX)
Perhaps the closest historical analogy is the United States Transcontinental Railway。
Prior to 1869, a six-month ride from New York to San Francisco was required, at a cost of approximately one year's wages and with a fairly real risk of death。
After 1869, this journey takes only a week。
The railway itself is an amazing engineering achievement, but the real story is what it opens:
Sears Roebuck, Swift, Armour, Standard Oil, and eventually U.S. Steel... These were born out of the railway boom and further integrated the industrial empire。
If Falcon 9 is the transcontinental rail of the space age, then Starship could be equivalent to an aircraft-level upgrade。
The railroad opened a continent。
The jet age opened a planet。
Starship will open the solar system。
The Industrial Moon
The Moon has been of scientific significance since the very beginning of mankind's vision of the moon。
Now it begins to have economic significance。
Because it is an entire world of industrial raw materials。
Let's see how we get things off the moon。
As noted earlier, the Moon has only one sixth of the gravity of the Earth and has no atmosphere, making mass drives — not rockets — natural means of transporting goods from the surface of the Moon。
This will revolutionize transport economics。
Once the orbit is completed, the marginal cost of transporting manufactured goods is determined mainly by electricity rather than fuel。
And the power on the moon is sunlight。
A parcel was thrown from the surface of the moon, re-entered the Earth ' s atmosphere with an insulation shield, opened its parachute and landed at the recovery site。
When the amount swallowed is large enough, the marginal costs no longer look like space, but more like freight。
And then, what do you make there。
The same moon is a source of silicon and aluminium for solar cells and satellites, as well as for the entire industrial base。
In the 1930s and the 1940s, the space revolutions could be like this:
Automatic mining vehicles operate 24 hours a day on the border
Refineries produce aluminium and silicon
Plants assemble satellites, solar panels and the chips needed to run them。
Most industries on Earth have a version of the Moon that awaits construction。
SpaceX can't build it all alone。
Those who built the Moon, Alcoa, the Moon Caterpillar and the Moon Union Pacific will be the giants of the twenty-first century。
Figure: Starship HLS is a Moon Launcher built by SpaceX for the NASA Artemis project, designed to allow humans to re-enter the surface of the Moon after more than 50 years and to deliver the basic modules needed for permanent presence to the vicinity of the Antarctic of the Moon. (Photo: SpaceX)
The strength of the sky
In 2030, artificial intelligence bottlenecks may not be chips but electricity。
The obvious response is to build more solar energy in Texas or Nevada。
But it hit the wall faster than people thought。
1 Tawa continuous solar power, approximately US land area1%I don't know。
The new utilities are usually licensed for one year or more。
xAI built Colossus in Memphis, which required the deployment of a fleet of temporary gas tankers to be drawn into a state-sanctioned process and the establishment of an independent power hub at Mississippi, on the other side of the state border, to get 1 Giva online。
IT'S IMPOSSIBLE TO EXTEND THIS TO THE HUNDREDS OF GIVA NEEDED FOR AI CONSTRUCTION。
Even providing back-up gas turbines for solar energy, there is a backlog of internal and directional leaves in 2030。
Figure: Baker Hughes Frame 5/2C gas turbine generators. Casting and directional blades within such combustion machines are produced by a small number of professional foundries, all of which accumulated in 2030. A mega-data centre requires dozens of such equipment. (Photo by Baker Hughes)
The solution is to move the math to where the sun is。
Once Starship flies daily, orbital deployment becomes routine, it will be easier。
The economy, on the other hand, will continue to improve with rocket launches, solar panels and the cost curve of chips。
SpaceX CFO Bret Johnsen explains:
“We are upgrading the capacity of the plant and benefiting from a decrease in the cost of silicon, so our cost will decline in the coming years。
If you look at ground solutions, the curve is going in the opposite direction. Everything is becoming expensive: cooling methods, electricity costs are not falling, land and regulation are becoming more challenging.”
A common objection comes from those who heard the “space data centre” and thought that a building of the size of Colossus would be launched into orbit。
But that is not the case。
SpaceX early investor Gavin Baker said, "It's about as big as a Blackwell hanger, with solar wings, and each side may be 500 feet long. You put it in the sun-synchronous orbit, so the solar panels are always in the sun。
I spent a lot of time in Starbase over the years, and I talked to a lot of SpaceX engineers. I do believe that this is the most talented group of engineers on Earth who are very sure that they have solved this problem.”
Figure: AI Sat Mini is designed to capture the energy of the sun. (photo: terafab.ai)
In fact, Mask thinks that AI Sat Mini is easier to make than Starlink。
He explains:
“You still have some laser links, but you don't need those extremely complex antennas on Starlink。
IT IS EASIER TO DESIGN AN AI SATELLITE THAN THE TWO。
AI satellites don't need magic. A lot of the technology we've made for Starlink V3. We do not think this is a particularly difficult issue than what we are already doing.”
He predicted that within five years, SpaceX would launch to orbit an AI calculator per year, which would exceed the cumulative calculus of installation on Earth。
The math here is probably:
Annual10,000 Starship launchesThat is, more than one launch every hour around the clock。
By the late 2030s, as the Moon mass drive went onlineSquatThe threshold will enter the horizon:
That was 1,000 times the calculus of deployment in 2030, and it was launched into deep space at the rhythm of one satellite every few minutes。
Mars
The Mars trajectory should have started this year。
Mask announced in September 2024 that SpaceX would launch five unmanned Starships with Optimus robots on Mars at the transfer window in November 2026 to test the landing system, find ice and start building infrastructure for future manned missions。
He said in May 2025 that the probability of achieving this goal was five or five。
But earlier this year, things changed。
On 8 February, Mask posted on X that SpaceX would postpone the Mars schedule and shift its immediate focus to self-sufficient cities on the Moon。
The reasons are:
The Mars launch window appears only once every 26 months and takes six months to fly; the moon can arrive once every 10 days, with only two days to fly。
He writes:
“This means that we can finish the Moon city faster and faster than Mars。
In other words, SpaceX will also try to build the city of Mars and start doing so in about five to seven years' time, but the overriding priority is to guarantee the future of civilization and the moon is faster.”
On the surface, it's a turn。
But in fact, this is the moment when the path to millions of people in Mars becomes clearer。
The orbital data centre issue became clearer in late 2025 and early 2026, giving the moon a new character。
In order to achieve the calibration of the balconies, it is necessary to:
Moon mining, Moon refining, Moon manufacturing solar panels, radiators and satellite structures, and launching them into orbit with a mass drive of lunar energy supply。
The industrial base of this size requires a permanent population, which requires cities。
The city can be funded entirely by the orbital computing industry and become a rehearsal for Mars。
SpaceX addresses every problem that has to be solved in order to build a self-sufficient city on Mars
-
Radiation shielding -
Life support -
In situ resource utilization -
Permanent governance of the off-site population -
Supply chains across gravity wells。
The building of the Moon City will be a much faster iterative cycle, teaching SpaceX how to build Mars City。
First no-man moon landing demonstration, 2027 at the earliest。
According to the Mask public timetable, the city of the Moon will arrive later in less than 10 years。
Quality drives, lunar industry construction, orbital computing infrastructure lunar manufacturing will be launched in parallel。
Then Mars。
But the hardest part is not the transporter。
Instead, it is to build the Mars-end infrastructure so that it can absorb these people。
The Moon rehearsal will help。
Optimus will help。
In May 2025, in the Starbase Mars speech, Mask repeatedly stated that the early days of the Starship would carry Optimus robots, look for resources and start building infrastructure for humans to arrive。
The company is building an annual product at Fremont1 millionAnd build an annual production in Giga TexasTen millionThe line of production。
These robots are still in their early stages of production and have yet to complete meaningful practical work at the Tesla plant。
But the capacity to go online for the next two to three years is critical to the self-reliance of the initial Mars base。
Note: SpaceX rendering: Optimus robots work on Mars, recapitulating the famous photo, Lunch on the skyscrapers, during the construction of Rockefeller Center in 1932. (Photo: SpaceX)
The conscious sun
SpaceX adopted the mission statement after taking in XAI in February:
Scaled, a conscious sun was born to understand the universe and to extend the light of consciousness to the stars。
That depends on how you understand it。
It's either the most ridiculous thing a serious company has ever written on mission pages
Or the most honest words。
In our view, it is the latter。
If you look at the structure of the organization, SpaceX is a launch service provider with an Internet subsidiary and an AI laboratory recently acquired。
If you look at the technology road map, it's the only company on the planet that is in the process of completing its scarce transformation。
If you look at the Declaration of Mission, it is a founder of the most operational capacity of our time, who has seriously tried to push humanity through that bottleneck。
There are two possibilities at the other end of the bottleneck:
One is that we become an interstellar species that share the universe with the smart machines that we build
Another is that we are just a footnote on a rocky planet that has not made a leap。
When the first child born on Mars asked his parents why our family was here, Starship had probably been flying for 30 years。
In the factory around the corner, Optimus robots are working, running the next generation of Grok, who has improved themselves for 20 years。
Maintaining the computing power of her city, from space data centres。
These data centres are made of other robots using the Moon Boundary and are launched by a mass drive. That mass drive has been placing satellites deep into space for almost a generation, at a single pace every few minutes。
Her parents came to Mars and were riding an aircraft named after a starship in a novel by Iain M. Banks。
Because at some point in the early 21st century, a man who had read those books in his youth decided to spend his life making them a reality。
Banks understands those who choose to go to Mars。
Civilization is paradise, but his most interesting people are often those who leave it。
Civilization solves the scarcity, and the rest is humanity's desire for a difficult journey。
Even though heaven is next door, borders remain meaningful habitats。
Mask said that the words of early Mars colonists would be Shackleton。
It comes from the famous recruitment ad for the 1914 Trans-Antarctic expedition:
“Recruit men on dangerous journeys. Low wages, harsh colds, long months of darkness and constant danger call into question the safe return. Those who succeed will be honoured and recognized.”
The ad was almost certainly invented by a later generation。
But the story has been glorified for 100 years because it captures some kind of truth about who chooses to leave。
Why would anyone find it attractive
Mask said, "Oh, my GodLife cannot be just another tragic solution. Something must inspire you to wake up in the morning glad you're part of humanity。The Earth is the cradle of mankind, but you cannot stay in it forever. The time has come to go out, to become a civilization headed towards stars, to reach the stars and to expand the scope and scale of human consciousness. I think it's incredibly exciting. It made me happy to be alive. I hope you feel the same way。"
Figure: Starman, a human model in a SpaceX suit, sits in front of Elon Musk private Tesla Roadster and orbits the sun. This vehicle was launched on 8 February 2018 when Falcon Heavy first tested the payload. Its current orbit will pass around Mars about once a year for about a million years. (Photo: SpaceX)
Disclaimer
This material is used only for educational purposes and does not constitute an investment proposal or an offer for the services of an investment adviser。
This material shall not serve as a basis for any investment decision-making。
a16z has invested SpaceX through its administered funds and thus has a financial interest in the company ' s performance and future prospects。
In particular, if SpaceX ' s value increases, a16z will benefit; as a shareholder in the company, a16z fund will also receive any usual dividends payments。
However, SpaceX did not pay a16z for this material。
