a16z: “The sun bears witness—SpaceX is worth $7.5 trillion.”
TechFlow Selected TechFlow Selected
a16z: “The sun bears witness—SpaceX is worth $7.5 trillion.”
Deep Dive into Musk’s Master Plan
Navigating Web3 tides with focused insightsOriginal authors: @mikemcg0 and @pmarca
Translated and edited by: Yu Shiyi
Elon Musk’s compensation plan at SpaceX revolves around two objectives.
The first objective: He receives his first reward if the company reaches a $7.5 trillion valuation and establishes a permanent human colony of at least one million people on Mars.
The second objective: He receives his second reward if SpaceX operates data centers in space that collectively consume at least 100 terawatts of electricity—a figure over 1,000 times greater than the total power consumption of all Earth-based data centers combined.
If neither objective is achieved, Musk receives nothing beyond his annual salary of $54,080, which he has drawn since 2019.
The board members who signed this compensation plan have witnessed something for the past two decades:
Musk repeatedly makes seemingly impossible SpaceX predictions—and then repeatedly turns them into reality.
He said SpaceX would launch humans into orbit—something no private company had ever done before. Today, SpaceX routinely transports NASA astronauts to orbit.
He said SpaceX would land and reuse orbital-class rockets—something the entire industry previously treated as disposable hardware. Today, SpaceX has successfully recovered and reused its boosters hundreds of times.
He said a satellite internet business could be worth tens of billions of dollars—something nearly unthinkable at the time, given that satellite internet was widely regarded as a graveyard for bankrupt ventures. Today, Starlink’s revenue has grown from zero to $11.4 billion within just a few years.
These predictions are often aggressive on timelines—but almost never wrong on direction.
And SpaceX’s original mission, written in 2002, was to make humanity a multiplanetary species.
So the board tied Musk’s compensation directly to that mission itself.
If that mission sounds like science fiction, perhaps it’s because it truly originates from science fiction.
Iain M. Banks and the Blueprint of “The Culture”
For twenty-five years, Iain M. Banks wrote a universe called “The Culture.”
By most reasonable standards, it may be the best utopian society ever imagined by humans.
There, humans live alongside superintelligent AIs known as Minds. These Minds operate massive orbital habitats, each akin to a small world. The relationship between humans and AI is neither servitude nor competition—it is partnership.
No one is forced to work.
No one goes hungry.
Minds shoulder the staggering computational load required to run orbital cities.
Humans, meanwhile, continue being human.
That, it turns out, is already a full-time job.
The three autonomous drone ships SpaceX uses at sea to recover Falcon 9 boosters are all named after sentient starships from Banks’s novels:
- Of Course I Still Love You
- Just Read the Instructions
- A Shortfall of Gravitas
In an interview during the 2023 UK AI Safety Summit, Musk was asked: What should a good AI future look like?
He replied:
“Banks’s ‘Culture’ series is the best vision of an AI future we have so far. Nothing else comes close to helping you understand what a genuinely utopian—or progressive-utopian—AI future might be.”
He’s been telling us, through those landing platform names, exactly what he wants to build.
Caption: “Of Course I Still Love You” caught the Falcon 9 first stage on April 8, 2016—the first successful drone ship landing in history, and the moment reusable orbital spaceflight moved from theory to reality. The ship’s name is taken from a sentient starship in Iain M. Banks’s “Culture” series. (Image: SpaceX)
But “The Culture” is not a frictionless paradise.
Banks’s novels brim with war, intrigue, and moral complexity. It qualifies as a utopia because “The Culture” has solved the prerequisites of survival well enough that trillions of humans can finally devote themselves to what Banks calls “the truly important things in life”: sports, games, love, studying dead languages, observing primitive societies, tackling impossible problems, and climbing mountains without safety nets.
Such a future rests on four foundational conditions.
First, access to a meaningful fraction of a star’s energy output—many orders of magnitude greater than the energy output of human civilization today.
Second, large-scale physical intelligence: machines capable of building, mining, refining, and repairing anything—autonomously and anywhere.
Third, cheap digital intelligence surpassing biological intelligence.
Fourth, the ability to move mass from Earth into space cheaply, frequently, and reliably—because none of the above can scale using Earth alone.
Reasoning Backward from the Future
Most analyses of SpaceX proceed forward from the present:
rockets, satellites, contracts, revenue.
But to see what’s truly happening, a more useful approach is to start at the endpoint—and reason backward.
Mars City
The operational goal is:
To establish a self-sustaining city of one million people on Mars within the lifetime of people alive today.
The real difficulty lies in “self-sustaining.”
It means: If Earth stops launching ships, the city must still survive.
It must manufacture everything itself:
food, water, air, energy, medicine, machines—and eventually, more humans.
According to SpaceX’s own estimates, delivering one million people and millions of tons of cargo to Mars within a few decades would require thousands of Starship flights—more than ten launches per day during each transfer window.
These windows are dictated by Earth–Mars orbital mechanics: they last only a few weeks and open only once every 26 months.
Caption: SpaceX’s rendering of a Mars city. (Image: SpaceX)
Lunar City
A lunar city is a nearer, easier dress rehearsal.
Permanently shadowed craters at the Moon’s south pole contain ice; certain ridges receive near-continuous sunlight—making them naturally ideal for establishing bases.
But Musk isn’t talking about just a scientific outpost.
He envisions factories on the Moon manufacturing AI satellites—and launching them one by one into space using mass drivers.
Mass drivers are another concept Musk borrowed from science fiction: electromagnetic launch systems that leverage the Moon’s low gravity (one-sixth Earth’s) and lack of atmosphere to hurl solar-powered satellites into deep space at industrial scale.
These satellites can be manufactured on the Moon because lunar regolith contains roughly 20% silicon and 10% aluminum by weight—exactly the two primary inputs needed for solar cells and satellite structures.
Musk explains: “If you want to go beyond one terawatt, you must go to the Moon.”
Caption: Rendering of SpaceX’s mass driver at Moonbase Alpha, designed to launch AI satellites—i.e., data centers—manufactured on the Moon into orbit. (Image: SpaceX)
Orbital Data Centers
Musk is betting that:
Within a few years, the most economically optimal location for AI data centers will be space.
AI’s bottleneck is energy. Outside China, energy supply growth is extremely limited, while AI compute demand grows exponentially.
Solar panels in orbit generate 4–10× more power than identical panels on Earth—depending on how sunny the terrestrial location is.
The reason is simple:
Space has no atmosphere, no day-night cycle, no clouds, and no seasons.
NASA calculated this decades ago. Now, rockets are finally cheap enough to make it real.
Musk expects that, within five years, the AI compute SpaceX launches into orbit annually will exceed the cumulative installed compute capacity on Earth.
That’s why SpaceX merged with xAI in February.
Rockets and intelligence are becoming the same problem.
Starship: The Vehicle Enabling Everything Upstream
Starship is the vehicle that makes everything upstream possible.
Starship V3, which flew for the first time this year, is the largest and most powerful rocket ever built by humans. Standing taller than a 40-story building, its thrust exceeds that of the Saturn V—the rocket that carried astronauts to the Moon—by more than double.
According to NASA’s historical data, the cost to reach orbit averaged about $18,500 per kilogram.
In 2010, the first Falcon 9 reduced that cost by roughly 85%, bringing it down to ~$2,700/kg.
In 2018, Falcon Heavy cut it further to ~$1,400/kg.
Starship—the world’s first fully and rapidly reusable spacecraft—aims to reduce it yet again, to $100–$500/kg.
Spaceflight, once a multibillion-dollar endeavor per launch, is becoming a multimillion-dollar business.
Starlink: The Cash Flywheel
Starlink is the cash flywheel funding everything else.
According to SpaceX’s IPO filing, its connectivity business—almost entirely Starlink—generated $11.4 billion in revenue in 2025, up ~50% year-on-year, with an adjusted EBITDA margin exceeding 60%.
As of March 2026, Starlink serves 10.3 million users across 164 countries, operating on more than 9,600 satellites.
Starlink began as a side project—to fill SpaceX’s own launch capacity. Today, it’s becoming one of history’s great consumer businesses.
In 2019, when a16z conducted due diligence on SpaceX, several people told us this economic model would never work.
The reason: Starlink user terminals required antenna technology previously used only in F-22 fighter jets and Navy destroyers—technology never mass-produced for consumers.
SpaceX’s first terminals cost ~$3,000 to manufacture but sold for just $499.
Yet they drove costs down—and proved the skeptics wrong.
Falcon 9: The Workhorse Buying Time for the Future
Falcon 9 is the workhorse rocket buying time for everything else.
It is the only orbital-class booster in the world reused at scale. A single booster typically flies more than twenty times before retirement.
In 2025, SpaceX launched 83% of the world’s total mass into orbit.
Despite competitors’ half-century head start, SpaceX now launches more payload into orbit than all other global powers combined.
This is the stack—from bottom to top.
Generations later, the “Culture”-style future resides at the top.
Falcon 9 and Starlink sit at the bottom, paying today’s bills.
Each layer enables the one above it.
SpaceX CFO Bret Johnsen described the internal mindset:
“[Musk] created a culture where you set an initial goal that seems wildly ambitious—and then, step-by-step, you find yourself moving toward something absolutely achievable…
Take Mars. When I joined in 2011, mentioning Mars or a multiplanetary species got eye rolls. Now, when we talk about it, people ask: ‘Which year?’
I think Elon’s greatest strength is setting these goals—and then building outstanding business models around every piece of intellectual property required to achieve the ultimate objective.”
The Idiot Index and the Algorithm
Musk never intended to found a rocket company.
In 2001, at age 30, Musk was contemplating what to do after selling PayPal.
He’d long been fascinated by space. When he searched for NASA’s plans to send humans to Mars, he was shocked to discover: there were none.
So he devised a plan:
Send a small greenhouse to Mars and transmit photos back to Earth.
His idea was that if people saw a green sprout emerging on the lifeless red planet, it might reignite public interest in space—and galvanize political will to fund a real Mars program.
He just needed a rocket to get the greenhouse there.
Later that year, he traveled to Moscow to buy refurbished intercontinental ballistic missiles—the first of two trips.
The meetings reportedly involved copious vodka and posturing.
Musk’s best friend from Penn, Adeo Ressi, accompanied him. In a 2012 interview with Esquire, Ressi recalled:
“We’d all walk into a small room, each with our own bottle of vodka.”
The Russians didn’t take Musk seriously.
At one point, a chief designer even spat on Musk and his team in contempt.
The second Moscow trip occurred in February. Musk asked the price of a missile.
They quoted $8 million per missile.
Musk countered: $8 million for two.
Jim Cantrell, Musk’s aerospace advisor, remembers the response sounding something like:
“Young man, no.”
And implying Musk had no money at all.
Musk judged they weren’t serious—and walked out.
Cantrell assumed the trip was over.
On the flight home, he and Mike Griffin—who later became NASA Administrator and was then advising Musk—ordered drinks and toasted their escape from Moscow.
Musk sat in the row ahead, hunched over his laptop.
Then he turned around:
“Hey guys, I think we can build this rocket ourselves.”
He showed them a spreadsheet listing the raw materials needed for a rocket: aluminum, titanium, copper, carbon fiber—and the cost of each.
Those material costs amounted to just 2% of the quoted price.
As Musk later put it:
“Clearly, you just need to figure out a clever way to assemble those materials into rocket shape.”
Within months, Musk decided to risk $100 million to start a rocket company—more than half the ~$180 million he’d earned from selling PayPal.
SpaceX was thus founded in a warehouse in El Segundo, California.
He invited five people to join the founding team.
Three declined—including Cantrell and Griffin.
The two who accepted were:
- Tom Mueller, who later became VP of Propulsion and SpaceX’s first employee;
- Chris Thompson, the second employee, responsible for operations and production.
Musk later joked:
“SpaceX in 2002 basically consisted of carpet and a Mexican mariachi band. That’s it. You can see—I’m a dancing machine.”
Years later, Musk dubbed the diagnostic tool behind that spreadsheet the “Idiot Index.”
If a part’s cost vastly exceeds its raw material cost, either you’re an idiot—or you’re working with idiots.
It sounds like a joke, but it’s the foundation of SpaceX’s strategy.
Every part SpaceX procures triggers an “Idiot Index” calculation.
One of SpaceX’s earliest legendary stories involves Steve Davis.
Davis joined SpaceX straight out of Stanford as employee #14. His task: procure an actuator to steer the Falcon 1 upper stage.
When he reported that traditional aerospace suppliers quoted $120,000, Musk laughed.
Musk told him the part couldn’t be more complex than a garage door opener—and gave him a $5,000 budget to build it from scratch.
Biographer Ashlee Vance recorded that Davis spent nine months iterating on the design—and delivered a functional actuator costing just $3,900.
When Davis sent Musk the technical breakdown of his victory, Musk replied with just two letters:
“Ok.”
To push the Idiot Index to its theoretical minimum, you must integrate vertically—and control the entire process end-to-end.
But vertical integration incurs fixed costs, which only pay off at high volume.
And achieving high volume in rocketry requires breaking the industry’s long-standing operating model.
Traditional launch providers—like ULA and Arianespace—treat every mission as custom-built.
Customers specify orbits, payloads, and integration requirements—and providers design bespoke missions around the satellite.
This model assumes:
Only a few launches per year, each extremely expensive.
It makes scalable manufacturing impossible.
SpaceX flipped this model on its head.
They published a Falcon User’s Guide defining the rocket’s exact specifications—and told customers:
Design your satellite to fit our rocket.
At the time, this was considered radical—and cost SpaceX some early business.
But it unlocked the manufacturing flywheel.
Standardization and reusability reinforce each other.
Because every Falcon 9 is identical, recovered boosters can be refurbished into complete, certified, flight-ready products.
The first Falcon 9 booster to fly twice completed its re-flight in 2017.
By 2020, a single booster had flown five times.
By 2021, ten times.
Today, the record holder has flown 35 times.
This reusability has transformed space economics—and it’s hard to see how competitors can catch up.
In 2021, Musk estimated Falcon 9’s marginal launch cost (excluding overhead) to deliver 15 tons to orbit was ~$15 million under optimal conditions—about half to one-third the cost of alternatives.
Today, SpaceX launches a rocket every two to three days using reused boosters—while competitors launch only a handful of custom rockets per year.
But SpaceX’s advantage isn’t just economies of scale, vertical integration, and superior strategy.
It’s also speed and culture.
Traditional aerospace companies eliminate uncertainty through analysis.
NASA once politely described Boeing’s Commercial Crew Program as:
“Employing mature systems engineering methods, investing heavily in upfront engineering studies and analysis to mature system designs before manufacturing and testing.”
Measure twice, cut once.
SpaceX reversed that order.
The company builds many inexpensive prototypes, pushes them to failure, learns from failure, and iterates rapidly.
Starship’s test program has produced more spectacular explosions than any rocket program in history.
But each failure is a data point revealing where reality diverges from the model.
This contrast is stark for those who’ve worked in both worlds.
Garrett Reisman, a NASA astronaut who flew two Space Shuttle missions, left NASA in 2011 to join SpaceX as a senior engineer.
He described NASA’s prevailing view of SpaceX at the time:
“They’re cowboys; they’re dangerous; they’ll kill people.”
What truly changed his mind was watching how SpaceX worked.
“They built in one month what NASA took a year to build. We were stunned.”
The clearest example is the Falcon 1 program.
Between 2006 and 2008, SpaceX launched four Falcon 1 rockets from Kwajalein, a tiny atoll in the Pacific.
The first three failed.
But each failure was different—and each provided learning:
- First: fuel leak;
- Second: abnormal propellant slosh;
- Third: residual engine thrust causing collision during stage separation.
By September 2008, the company had funds for just one more launch.
Nor was SpaceX the only company Musk was running on the brink.
His electric vehicle company Tesla was also weeks from bankruptcy.
He had to decide: concentrate his remaining PayPal cash on one company—or split it between both.
Musk recalled:
“It was an incredibly difficult decision. In the end, I chose to split my remaining money and try to keep both companies alive. But that could have been a terrible decision—killing both at once.
I never thought I’d have a nervous breakdown—but I came very close.”
He couldn’t choose, because in his worldview, both missions were indispensable:
Tesla accelerates the world’s transition to sustainable energy.
SpaceX makes humanity a multiplanetary species.
Musk’s then-fiancée Talulah Riley said in the BBC documentary The Elon Musk Show: “All available resources had to go to the companies. He gave me an exit option. He said: ‘The hardest part is coming next—you don’t have to stay and go through it with me.’”
Caption: Elon Musk inspecting debris from the first Falcon 1 on Omelek Island in 2006. (Photo: Hans Koenigsmann)
The fourth launch succeeded.
That December—just weeks before SpaceX ran out of cash—NASA awarded the company a $1.6 billion cargo contract.
When NASA called to notify Musk, he was overwhelmed with emotional relief—and blurted out:
“I love you.”
The pattern forged through rapid failure and rapid correction became SpaceX’s cultural DNA across every project.
That’s why SpaceX can now iterate rapidly between Starship test flights—while traditional aerospace programs often take years to redesign vehicles after anomalies.
This method works better than alternatives because:
For problems you don’t yet fully understand, you cannot arrive at perfect solutions by thinking alone.
Reality is the only sufficiently qualified validator.
The key is reducing the cost of consulting reality—so you can consult it frequently.
SpaceX’s “Algorithm”
The above tells the story of SpaceX’s iterative loop.
But it also has a codified version.
Over the past two decades, Musk encoded SpaceX’s methodology into a five-step operational process the company internally calls “The Algorithm.”
Tim Berry worked at SpaceX for ten years, leading production for the Falcon 9 and Falcon Heavy upper stages. He said the process was “drilled into our heads.”
Walter Isaacson published its canonical version in his Musk biography:
1. Question Every Requirement
Every requirement must include the name of the person who issued it.
You shouldn’t accept statements like “this requirement comes from Legal” or “this requirement comes from Safety.”
You need to know the actual person who made the request—and question it, regardless of how smart they are.
Requirements from smart people are the most dangerous—because people are least willing to question them.
Then, make those requirements less stupid.
2. Delete Everything You Can
You may later have to add some things back.
In fact, if you don’t end up adding back at least 10% of what you deleted, you didn’t delete enough.
3. Simplify and Optimize
Do this only after Step 2.
A common mistake is simplifying and optimizing a part or process that shouldn’t exist at all.
4. Accelerate Cycle Times
Every process can be accelerated.
But only after completing Steps 1–3.
Musk says he made this mistake at Tesla’s factory: spending much time accelerating processes he later realized should have been eliminated.
5. Automate
Automation comes last.
Tesla’s mistakes in its Nevada and Fremont factories involved attempting automation too early—before questioning requirements, deleting parts and processes, and shaking out bugs.
Most engineering organizations jump straight to Step 5.
They automate a process that shouldn’t exist.
SpaceX runs these steps sequentially—every time, across every part of the company.
After “The Algorithm” runs enough times on a given hardware item, it starts looking unlike anything else in the industry.
Caption: Three generations of SpaceX’s Raptor engine: V1 to V3. (Image: SpaceX)
Raptor 3 is the product of a team iterating on the same engine for a decade.
It delivers 22% more thrust than Raptor 2, weighs 40% less, and requires no heat shield.
Why? External plumbing and wiring once bolted onto the engine have been fused into its metal structure via 3D printing.
Musk says:
“The effort required to simplify the Raptor engine, internalize secondary flow paths, and add regenerative cooling to exposed components is astonishing. It’s approaching the limits of known physics.”
No known engine program in aerospace history has iterated this quickly.
The Space Shuttle Main Engine flew essentially the same design for its final thirty years.
The RD-180 powering Atlas V is a derivative of a 1970s-era engine design.
Meanwhile, SpaceX has completed three completely new Raptor designs in under a decade—with each generation significantly better than the last.
The same philosophy applies to people.
By mid-2018, Falcon 9 reusability had entered a reliable rhythm. Musk shifted focus to the satellite internet constellation that would ultimately fund all upstream ambitions.
The Starlink team, based in Redmond, Washington, included many senior engineers from Microsoft—whose development pace frustrated Musk.
In June, he flew to Redmond and fired the senior leadership team.
Then he transferred young star engineers from the rocket division—and gave them one year to launch the first operational satellites.
It was a brutal management tactic. Media coverage of the firings made the department appear to be collapsing.
But 11 months later—in May 2019—the first Starlink satellites launched.
Musk removed a bottleneck—and moved on to the next problem.
He manages everything this way.
In 2018, Tesla was deep in “production hell” building the Model 3—burning cash at a rate threatening its survival. Musk literally moved into the factory.
Years later, he recalled:
“For three years straight, I lived in the Fremont and Nevada factories. I slept on the floor under my desk so the entire team could see me during shift changes.
That mattered, because if the team thought their leader was off enjoying himself on a tropical island drinking Mai Tais, morale would collapse.
Because they saw me sleeping on the floor during shift changes, they knew I was there. That made a huge difference—they went all out.”
Later, he institutionalized this as a company-wide rule:
The higher your position, the more visible your presence must be.
To find someone comparable to Musk’s CEO operating style, you must return to the industrialists of the late 19th and early 20th centuries:
Henry Ford, Andrew Carnegie, Thomas Watson, Andrew Mellon, Cornelius Vanderbilt.
What makes Musk’s operating style unique is his relationship to work.
He reportedly visits each of his companies weekly, identifies the single biggest problem—and fixes it.
Fifty-two weeks a year, he does this.
Theoretically, each company solves 52 major problems per year.
An engineer who joined SpaceX from another aerospace company described the experience as:
“Being thrown into a shockingly competent zone. Everyone around you is absolutely capable.”
The Constellation
SpaceX looks like a single company.
But a more useful framing is as the central node of a corporate constellation.
These companies are run by one person, aligned on the same long-term mission—and nearly impossible to disentangle from one another.
Over the past two decades, Musk assembled a group of companies. Each solves a constraint that would otherwise bottleneck the others.
Now, they’re beginning to compound upon one another.
SpaceX’s February merger with xAI is emblematic of what SpaceX is becoming.
If compute ultimately migrates to orbit—as Musk bets it will—then SpaceX possesses the most credible path to deploy it at the scale AI demands.
Launching mass into orbit—and mass-producing intelligence—may be the two most critical capabilities of the coming decades.
Now, they reinforce each other under one roof.
xAI brings Grok, a cutting-edge model uniquely positioned in real-time information thanks to access to X’s live data stream.
It also brings the engineers who built the Colossus 1 and Colossus 2 supercomputers—engineers whose speed defies industry expectations.
Caption: Colossus 1. (Image: xAI)
Colossus’s construction deserves closer inspection.
xAI took over an old factory in Memphis—and brought 100,000 GPUs online for training in just 122 days.
Once racks began arriving, the entire cluster was up and running in just 19 days.
NVIDIA CEO Jensen Huang commented on Musk:
“Going from concept to building a massive, liquid-cooled, powered, permitted factory—and doing it in that timeframe—is superhuman.
To my knowledge, only one person in the world can do this.
What they accomplished is unprecedented. No one has ever done this before. 100,000 GPUs, as a cluster, would easily be the fastest supercomputer on Earth in 2024.
This normally takes three years of planning, then equipment delivery, then another year to get everything running.”
A project that would take industry peers at least four years, Musk and the xAI team completed in four months.
In May this year, Anthropic agreed to pay SpaceX $1.25 billion monthly for exclusive access to Colossus 1’s full compute capacity.
A few weeks later, in a revised IPO filing, SpaceX disclosed Google will pay $920 million monthly for access to 110,000 GPUs—roughly half the compute Anthropic secured.
Together, these deals represent ~$26 billion in annual revenue.
And this is from just two customers—paying for a business that didn’t exist before SpaceX absorbed xAI earlier this year.
Chips, power, and land are scarce.
SpaceX is becoming one of the few companies with sufficient AI infrastructure—not just to rent out compute, but to pursue its own ambition of building leading-edge frontier models.
What xAI gains from SpaceX is a more durable solution to the power constraint Musk believes will limit AI in the coming years.
Producing enough electricity to meet his projected intelligence demand requires grid expansion, new power plants, and multi-year permitting processes—time the industry doesn’t have.
In his view, orbital solar is the answer—it’s effectively limitless.
And SpaceX is the only company with the transportation system capable of deploying compute at that scale.
Whether he’s right is one of tech’s most important open questions.
But SpaceX’s IPO filing shows the company takes this bet extremely seriously: it forecasts AI will be its largest market by far—far exceeding all others.
The space business that founded the company is, by comparison, nearly a rounding error.
Tesla: Another Core Piece of the Constellation
Tesla is another crucial component of this constellation.
Its integration with SpaceX runs deeper in another dimension.
Tesla and SpaceX share founders, talent pools, operational cultures—and increasingly overlapping technology roadmaps.
Tesla provides three things to the SpaceX-xAI side of the constellation.
First, chips.
AI5, AI6, and Dojo3 are all designed internally at Tesla.
Musk has explicitly stated these chips aren’t just for cars—they’re building blocks for the broader constellation’s compute stack.
AI5 handles autonomous driving inference.
AI6 is designed for Optimus and AI data centers.
Dojo3 pairs with the planned AI7, engineered for orbital compute.
Second, robots.
Tesla bets Optimus will become the physical AI layer—for factories, warehouses, homes—environments designed to operate without human labor—and ultimately serve Musk’s envisioned lunar and Martian cities.
Third, solar.
Musk says Tesla and SpaceX are each building toward 100-gigawatt annual solar cell production capacity—to support AI construction on Earth and in orbit.
Then there’s TeraFab.
In April this year, Tesla disclosed it had begun ordering equipment for a research-grade semiconductor wafer fab inside its Giga Texas campus.
Musk told investors on Tesla’s Q1 2026 earnings call:
“We expect this to be roughly a $3 billion project, with potential output of several thousand wafers per month.”
Separately, SpaceX is funding a much larger facility—with a mature design capacity of ~1 million wafers per month.
The reason: no existing wafer fab can expand at the pace Musk envisions.
And his envisioned scale is measured in gigawatts.
Musk said last week: “This isn’t a promise—we’re going to try it, and we think it’s highly probable we’ll succeed: reaching ~1 gigawatt per year of space-based AI compute by year-end.
Then, aspirationally, scaling by an order of magnitude each year.
So, 10 gigawatts per year in two and a half years. 100 gigawatts per year in three and a half years.
Then, depending on progress in global chip manufacturing and TeraFab, exceeding that—reaching 1 terawatt per year, or 1,000 gigawatts.
That’s twice U.S. electricity consumption.”
Caption: SpaceX’s TeraFab aims for ~1 terawatt annual output—roughly twice current U.S. electricity consumption. (Image: terafab.ai)
Comparing Musk to Gilded Age industrialists captures something real—but also highlights differences.
Carnegie built steel.
Vanderbilt built railroads.
Each dominated a single sector of their era’s industrial base.
Musk is trying to dominate several sectors simultaneously:
space, energy, artificial intelligence, robotics, tunnels, brain-computer interfaces, autonomous vehicles.
And bending them all toward a goal most consider fantastical.
Whether this succeeds is genuinely unknown—and much of it may fail.
But the attempt itself has no historical precedent—and may become the rallying point for another century.
The World SpaceX Is Opening
Before retiring in 2011, the Space Shuttle cost ~$54,500 to launch 1 kg into orbit.
With Starship mature, Musk expects that cost to fall to $100/kg.
When the cost of space access drops by over 500×, every industry theoretically viable in space becomes economically feasible.
There are many such industries.
Caption: Starship and Super Heavy are designed to return to the launch site post-flight and be caught by the launch tower—enabling rapid turnaround with no refurbishment. (Image: SpaceX)
The closest historical analogy may be the U.S. transcontinental railroad.
Before 1869, traveling from New York to San Francisco by stagecoach took six months, cost roughly a year’s wages—and carried a real risk of death.
After 1869, the journey took just one week.
The railroad itself was an astonishing engineering achievement—but the real story was what it unleashed:
Sears Roebuck, meatpacking giants Swift and Armour, Standard Oil—and eventually U.S. Steel—all emerged from the railroad boom, further consolidating industrial empires.
If Falcon 9 is the transcontinental railroad of the space age, then Starship may be the jet-age upgrade.
Railroads opened a continent.
The jet age opened a planet.
Starship will open the solar system.
Industrializing the Moon
Since humans first gazed at the Moon, it has held scientific significance.
Now, it’s acquiring economic significance.
Because it’s a complete world composed of industrial raw materials.
First, how to launch things from the Moon.
As noted, the Moon’s gravity is just one-sixth Earth’s—and it has no atmosphere—making mass drivers—not rockets—the natural way to launch cargo from the lunar surface.
This would radically transform transport economics.
Once the orbital infrastructure is built, the marginal cost of shipping finished goods is driven primarily by electricity—not fuel.
And electricity on the Moon is sunlight.
A package is launched from the lunar surface, re-enters Earth’s atmosphere with a heat shield, deploys a parachute, and lands at a recovery site.
At sufficient throughput, marginal costs cease to resemble aerospace—and start resembling freight.
Next: what to manufacture there.
The same lunar regolith supplies silicon and aluminum for solar cells and satellites—and can serve as feedstock for an entire industrial base.
The space revolution of the 2030s and 2040s may look like this:
Autonomous mining vehicles operating continuously on regolith;
Refineries producing aluminum and silicon;
Factories assembling satellites, solar panels, and the chips needed to run them.
Most terrestrial industries have a waiting lunar counterpart.
SpaceX cannot build all of it alone.
The builders of “Lunar Alcoa,” “Lunar Caterpillar,” and “Lunar Union Pacific” will become the titans of the 21st century.
Caption: Starship HLS—the lunar lander SpaceX built for NASA’s Artemis program—designed to return humans to the lunar surface for the first time in over 50 years, and deliver foundational modules near the Moon’s south pole for permanent presence. (Image: SpaceX)
Compute in the Sky
By 2030, AI’s bottleneck may not be chips—but power.
The obvious response is to build more solar in Texas or Nevada.
But that hits physical limits faster than most imagine.
Continuous 1-terawatt solar power requires ~1% of U.S. land area.
And new utility interconnection permits typically take a year or longer.
xAI’s construction of Colossus in Memphis required deploying a temporary fleet of gas turbines, battling state permitting processes, and building an independent power hub across the state line in Mississippi—just to bring 1 gigawatt online.
Scaling that to the hundreds of gigawatts needed for AI construction is simply infeasible.
Even the gas turbines serving as backup for solar—their internal turbine blades and vanes—are backlogged until 2030.
Caption: Baker Hughes Frame 5/2C gas turbine generator. Casting of internal turbine blades and vanes for such turbines is handled by a handful of specialized foundries—*all* backlogged until 2030. A single hyperscale data center requires dozens of such units. (Image: Baker Hughes)
The solution: move compute to where sunlight already is.
Once Starship launches daily and orbital deployment becomes routine, this becomes easier.
And economics improve further as rocket launch, solar panel, and chip cost curves continue downward.
SpaceX CFO Bret Johnsen explains:
“We’re ramping factory capacity and benefiting from falling silicon costs—so our costs will decline over the next few years.
Ground-based solutions follow the opposite curve. Everything is getting more expensive: cooling, electricity won’t get cheaper, land and regulation grow more challenging.”
A common objection comes from those who hear “space data centers” and imagine launching buildings the size of Colossus into orbit.
That’s not the case.
Early SpaceX investor Gavin Baker says: “It’s roughly the size of a Blackwell rack—with solar wings, possibly 500 feet long on each side. You place it in sun-synchronous orbit so the panels are always in sunlight.
I’ve spent a lot of time at Starbase and spoken with many SpaceX engineers. I truly believe this is the most talented group of engineers on Earth—and they’re extremely confident they’ve solved this problem.”
Caption: AI Sat Mini is designed to capture solar energy. (Image: terafab.ai)
In fact, Musk believes AI Sat Mini is easier to manufacture than Starlink satellites.
He explains:
“You still have some laser links—but no need for the extremely complex antennas required on Starlink satellites.
Compared to Starlink, the AI satellite is actually simpler to design.
The AI satellite doesn’t require magic. Much of the technology we’ve already developed for Starlink V3 satellites applies directly. Relative to what we’re already doing, we don’t consider this a particularly hard problem.”
He expects that, within five years, the AI compute SpaceX launches into orbit annually will exceed the cumulative installed compute capacity on Earth.
The math roughly works out to:
10,000 Starship launches per year—more than one per hour, around the clock.
By the late 2030s, with lunar mass drivers online, the petawatt threshold comes into view:
That’s 1,000× the compute deployed in 2030—and launched into deep space at a rate of one satellite every few minutes.
Mars
The Mars timeline was originally supposed to begin this year.
In September 2024, Musk announced SpaceX would launch five uncrewed Starships to Mars in November 2026—carrying Optimus robots to test landing systems, search for ice, and begin building infrastructure for future crewed missions.
In May 2025, he said the probability of achieving this was roughly 50/50.
But earlier this year, things changed.
On February 8, Musk posted on X that SpaceX would delay the Mars timeline—and shift near-term focus to building a self-sustaining city on the Moon.
The rationale:
Mars launch windows occur only once every 26 months—and the flight takes six months; whereas the Moon is reachable every 10 days, with a flight time of just two days.
He wrote:
“That means we can iterate faster and complete a lunar city sooner than a Mars city.
That is, SpaceX will still strive to build a Mars city—and begin doing so in roughly five to seven years—but the overwhelming priority is securing civilization’s future, and the Moon is faster.”
Superficially, this appears to be a pivot.
In reality, it’s the moment the path to a million-person Mars city crystallizes.
The orbital data center thesis sharpened in late 2025 and early 2026—giving the Moon a new role.
To reach petawatt-scale orbital compute requires:
Lunar mining, lunar refining, lunar manufacturing of solar panels, radiators, and satellite structures—and launching them into orbit using lunar-powered mass drivers.
That scale of industrial infrastructure requires a permanent population—and a permanent population requires a city.
This city can be fully funded by the orbital compute industry—and serve as a dress rehearsal for Mars.
Every challenge SpaceX must solve to build a self-sustaining Mars city will first arise in the lunar city:
- Radiation shielding;
- Life support;
- In-situ resource utilization;
- Governance of off-world permanent populations;
- Supply chains across gravity wells.
Building a lunar city will teach SpaceX how to build a Mars city—through a much faster iteration cycle.
The first uncrewed lunar landing demonstration targets 2027.
Per Musk’s public timeline, the lunar city follows within less than a decade.
Mass drivers, lunar industrial construction, and lunar manufacturing of orbital compute infrastructure will launch in parallel.
Then, Mars.
But the hardest part isn’t transporting people.
It’s building the Mars-side infrastructure capable of absorbing them.
The lunar rehearsal will help.
Optimus will help too.
In his May 2025 Starbase Mars speech, Musk repeatedly emphasized that early uncrewed Starships will carry Optimus robots to locate resources and begin constructing infrastructure for human arrival.
The company is building a 1-million-unit-per-year production line in Fremont—and a 10-million-unit-per-year line in Giga Texas.
These robots remain in early production and haven’t yet performed meaningful practical work in Tesla factories.
But the capacity coming online in the next two to three years will be critical for bootstrapping the first Mars base.
Caption: SpaceX rendering: Optimus robots working on Mars—recreating the famous 1932 photo “Lunch atop a Skyscraper” during Rockefeller Center’s construction. (Image: SpaceX)
A Conscious Sun
The mission statement SpaceX adopted after absorbing xAI in February is:
Scale to birth a conscious sun—to understand the universe and extend the light of consciousness to the stars.
How you interpret this depends on your perspective.
It’s either the most absurd sentence ever placed on a serious company’s mission page,
or the most honest.
We believe it’s the latter.
If you squint at the org chart, SpaceX is a launch service provider with an internet subsidiary and a recently acquired AI lab.
If you squint at the technology roadmap, it’s the only company on Earth assembling the full pre-scarcity transition stack.
If you squint at the mission statement, it’s a founder of unparalleled operational capability earnestly attempting to push humanity through that bottleneck.
Beyond that bottleneck lie two possibilities:
One: We become an interstellar species—sharing the cosmos with intelligent machines we built;
Two: We remain a footnote on a rocky planet—a species that never completed the leap.
When the first child born on Mars asks her parents, “Why are we here?” Starship may already have been flying routinely for thirty years.
At the corner factory, Optimus robots are at work—running descendants of Grok that have self-improved for two decades.
The compute sustaining her city comes from space-based data centers.
Those data centers were manufactured by other robots from lunar regolith—and launched by a mass driver that, for nearly a generation, has hurled satellites into deep space every few minutes.
Her parents arrived on Mars aboard vessels named after starships from Iain M. Banks’s novels.
Because, at some point in the early 21st century, a boy who read those books decided to spend his life making them real.
Banks understood those who choose Mars.
“The Culture” is heaven—but his most compelling characters are often those who leave heaven.
“The Culture” solved scarcity—what remains is humanity’s longing for arduous journeys.
Even if heaven sits next door, the frontier remains where meaning resides.
Musk says the recruitment pitch for early Mars colonists will be Shackletonian.
It echoes the famous 1914 advertisement for the Trans-Antarctic Expedition:
“Men wanted for hazardous journey. Small wages, bitter cold, long months of darkness, constant danger, safe return doubtful. Honour and recognition in case of success.”
This ad is almost certainly apocryphal.
But it’s been retold for a century because it captures something true about those who choose to depart.
Why would anyone find this attractive?
Musk says: “Life can’t just be solving one miserable problem after another. There must be something that inspires you—something that makes you wake up grateful to be human. Earth is the cradle of humanity—but you can’t stay in the cradle forever. It’s time to go out and become a civilization that sails among the stars—to venture into the starlight and expand the range and scale of human consciousness. I find that incredibly exciting. It makes me grateful to be alive. I hope you feel the same.”
Caption: Starman—a mannequin in a SpaceX spacesuit—sits at the wheel of Elon Musk’s personal Tesla Roadster, orbiting the Sun. The car was the payload of Falcon Heavy’s first test flight, launched on February 8, 2018. Its orbit will pass near Mars approximately once per Earth year for the next ~1 million years. (Image: SpaceX)
Disclaimer
This material is for educational purposes only and does not constitute investment advice or an offer to provide investment advisory services.
This material should not be relied upon for any investment decision.
a16z invests in SpaceX through its managed funds and therefore holds a financial interest in the company’s performance and future prospects.
Specifically, a16z stands to benefit if SpaceX’s value increases; and as shareholders, a16z funds will receive any customary dividend payments.
However, SpaceX has not compensated a16z for this material.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
Marc Andreessen
