Every Element in a SpaceX Rocket Existed a Billion Years Ago. SpaceX Did Not.

The inventory and the arrangement

Every element in a SpaceX rocket existed a billion years ago. SpaceX did not.

Let’s start with the bill of materials. Falcon 9 flies on an aluminum–lithium alloy. Starship is welded from stainless steel. The Raptor engine’s hot section is a nickel–chromium superalloy. The fuel tanks store methane, kerosene, and liquid oxygen. Iron, chromium, nickel, aluminum, lithium, carbon, oxygen, hydrogen. Every one of those elements was present on Earth before we were. The atoms in the rocket are older than the planet: the light ones were forged in the minutes following the Big Bang, the heavy ones spat out from dying stars going supernova. Nothing in a Starship is new matter, because mass is conserved in every chemical reaction that built it. What has been missing all this time was not substance. It was arrangement.

Technological advancement is a history of learning to arrange existing matter into configurations that do work. The difference between a riverbank and a rocket is the order of stuff, not stuff alone.

Aluminum makes this concrete. It is the most abundant metal in the Earth’s crust (accounting for roughly eight percent by mass) and the third most abundant element of any kind. In 10,000 BCE it was quite literally everywhere under our feet, locked away in clay and inaccessible. No one isolated it as a metal until 1825. Even decades later, it was so hard to extract that it traded as a precious metal: when the Washington Monument was capped in 1884, the builders set a hundred-ounce pyramid of cast aluminum at its apex, a metal then valued like silver, as a display of wealth. Two years later Charles Hall and Paul Héroult, working separately, found the electrolytic process that still carries their names, and the price of aluminum fell from over a thousand dollars a kilogram to under one.

Aluminum did not change. The clay under a modern smelter and the clay under a Neolithic camp are the same clay. In actuality, the knowledge changed. Aluminum was always there, and for all of human history until essentially yesterday, we did not possess the arrangement of steps that turns clay into metal. The rocket is that story compounded ten thousand times over: every alloy, every combustion cycle, every guidance law is a recipe we did not have and now we do. So the question that organizes everything below is simple. Where do those recipes come from, how soon do they arrive, and what happens to them if the only cookbook that holds them is lost forever?

Knowledge is the only thing that compounds

Recipes accumulate, and they accumulate in a particular way that almost nothing else in the physical world does.

We need to discuss the mechanism here. An idea is nonrival. If I use a theorem, a smelting process, or the equations that govern combustion in a Ford’s engine, my using it does not consume it or preclude you from using the same thing in the same moment. This is the structural fact Paul Romer built endogenous growth theory on, and that won him the Nobel Prize in 2018. Because ideas are nonrival, they generate increasing returns, and open-ended discovery of new recipes increases the size of our kitchen. While a single barrel of oil is burned a single time, the refining process is available forever, everywhere, by everyone, at no marginal cost.

Next, recipes are combinatorial. W. Brian Arthur’s account of how technology actually evolves is that new technologies “come into being as fresh combinations of what already exists.” A rocket engine is a combination of a turbopump, a combustion chamber, and a control loop, each of which is a combination of prior parts, down to the metallurgy, thermodynamics, and the calculus. Nothing is invented from nothing. Everything is assembled from the existing inventory of solved problems—the recipe book—which means every solved problem enlarges the space of problems solvable next. The toolkit builds on itself.

The toolkit does not lie solely with individuals. It is what Joseph Henrich calls the collective brain: “it’s our collective brains operating over generations, and not the innate inventive power of individual brains.” No one alive could rebuild all of chemistry from scratch. Leonard Read made the same point about a pencil in 1958: “Not a single person on the face of this earth knows how to make [a pencil].” A pencil. The wood, the graphite, the lacquer, the brass ferrule, the eraser, the global supply chains that converge on a six-inch object no one fully gets. A Raptor engine is just like the pencil, but with five more orders of magnitude of coordination behind it.

Nonrivality in ideas, combinatorial recipes, and our collective brain together are effectively a ratchet. Ideas are not consumed, they combine, and they are held in a collective memory that rarely forgets. The result is a forward-only arc whose rate of translation increases, because each new recipe dilates the set of accessible next recipes. This is a driving force in human evolution, and it has exactly one weakness. We will return to that.

The engine that selects the next arrangement

A ratchet doesn’t turn on its own. One must provide the torque. The space of possible recipes is astronomically large and most arrangements of matter do no useful work, so something (or someone) must decide which combinations are worth trying, fund that work, and prune the failures. Humanity’s best tool for that job isn’t an ideology. It’s a price.

Hayek wrote in 1945 that the knowledge required to run an economy “never exists in concentrated or integrated form but solely as the dispersed bits of incomplete and frequently contradictory knowledge which all the separate individuals possess.” No single planner can gather it, because most of it is local, tacit, and changing by the hour. The price system is the only known system that compresses scattered knowledge into a single number one can act on, and it also routes capital and human effort toward configurations that produce the most work per price unit (dollar). Schumpeter, on the other hand, named “creative destruction,” the process that “incessantly revolutionizes the economic structure from within,” which kills arrangements that are not optimally productive. This ensures the capital behind them may be redeployed.

This system is evident in the field of spaceflight. For fifty years, the orthodoxy was simple: you throw rockets away after a flight. Could you imagine scrapping a 747 after a single trip from New York to London? Every space agency on Earth accepted this. Then, a private actor bet a ton of capital—took a risk—on this orthodoxy being a mere accounting habit and not a law of nature. That bet has decreased the cost of shooting a kilogram into low Earth orbit from roughly fifty-four thousand dollars in the Space Shuttle era to nearly twenty-seven hundred on a reused Falcon 9. A ninety-five percent reduction, with Starship engineered to take it lower by another order of magnitude. No government committee ordered this. No five-year “strategic plan” contained it. A price signal said there was work to be done and money to be made, and capital ran downhill toward it exactly the way water does.

I need to be precise about my claim because the objection is foreseeable. I am not articulating a defense of every market outcome, and I am not making a moral pronouncement. I am putting forth an observation about an economic system that has reproducibly arranged matter into new, high-work configurations at civilizational scale. Central planning has not done it. Goodwill has not done it. If you want these recipes to keep coming, we need to keep the engine running. That engine, however, is about to change gears.

The accelerant

If the rate of human progress is indeed proportional to how quickly we can sample the space of possible arrangements, it follows that we must optimize just how quickly we do it. That is happening as I write this article. We’ve built a thing that searches faster than we do, and anyone with access to the internet can use it.

The proof is already here. Consider the protein folding problem. Predicting a protein’s three-dimensional shape from its one-dimensional sequence sat unsolved for half a century. In 2022 DeepMind’s AlphaFold released predicted structures for 214 million proteins, covering almost every protein catalogued by science, and in 2024 its authors took the Nobel Prize in Chemistry. A fifty-year grand challenge retired, then handed to every biologist on Earth as a free asset. The same group’s materials model, GNoME, proposed 2.2 million new inorganic crystals in a single pass, 380,000 of them stable enough to be worth synthesizing, an expansion the team described as roughly eight hundred years of human materials discovery delivered at once. Reinforcement learning now holds the plasma inside a fusion reactor stable against instabilities no human hand could counter in real time.

But this is a double-edged sword and there’s no handle. The acceleration cuts both ways. The capability that can search the space of stable materials can search the space of pathogens. The systems that compress drug discovery compress bio-weapons discovery. Faster knowledge is faster everything, and the dangerous recipes are reachable from the same toolkit as the useful ones. So the accelerant does two things at once: it is the best tool we have ever held for building the backup, and it is the sharpest reason we have ever had to hurry. Which forces the question of what is at risk.

The single point of failure

Now we turn from economics to engineering, because the correct framing for what follows is based on reliability.

There is an important rule to follow when designing any system that matters: always have redundancy. Similarly to diversifying your investment portfolio, you wouldn’t base a metropolitan electrical grid system on one main transmission line. A component whose failure takes down the whole system is a single point of failure, and the discipline of reliability engineering is the practice of finding those points and mitigating them before disaster strikes. Applying this to consciousness is an interesting task. Consciousness, the only known instance of matter that has learned to observe and model itself, the only thing in the observed universe that asks why there is something rather than nothing, runs on exactly one node, Earth. There is no failover. There is no second site. If we get it wrong here, there is nowhere for the process to continue.

And the natural state of that node is not safe. We are comfortable in the present, but extinction events can be instantaneous or gradual. Extinction is not the exception in the history of life; it is the rule. More than ninety-nine percent of all species that have ever existed are gone. The average mammalian species persists for about a million years and then it doesn’t. The end-Permian extinction erased around ninety-six percent of marine species in a geological instant. Against that base rate, one intelligent species has built a single fragile node, on a planet with a documented history of mass death, and loaded the entire known inventory of consciousness onto it.

The contemporary risk is not an asteroid. In the most careful published accounting we have, Toby Ord puts the odds of an existential catastrophe this century at about one in six, and the leading causes are not natural. Unaligned artificial intelligence he rates near one in ten, engineered pandemics near one in thirty, nuclear war and climate change near one in a thousand each, and that comet from “Don’t Look Up” is near one in a million. The dominant risks are anthropogenic, recent, and rising, and they are downstream of the technological acceleration discussed in the previous section. Martin Rees gives civilization even odds of reaching 2100. The Doomsday Clock reads eighty-nine seconds to midnight, the closest it has stood in its history. My own read compresses the window further: the dangerous interval is the next century and a half, because the technologies that finally make us powerful enough to leave are the same technologies that make us powerful enough to kill ourselves entirely. If we get this wrong here, it is over. There is no redundancy.

This is not alarmism. Alarms are a poor instrument in this case; they degrade judgment when judgment is scarce. This is the single most consequential reliability problem any engineer has been handed, stated without drama: one node, all of consciousness, zero redundancy. You do not solve that problem by being careful on the one node. You solve it by building a second.

Redundancy is the whole strategy

The fix for a single point of failure is not to make the one node perfect. This is an insufficient strategy because it is inherently not redundant. The fix is a second node that lies sufficiently outside the blast radius of the first. For consciousness, that means a self-sustaining settlement off of Earth, far enough and independent enough that the loss of Earth would not be the loss of everything. That destination is Mars, not the Moon or low orbit, which could not survive the loss of Earth. The architecture that gets there is specific, staged, and under way.

Stage one is energy and knowledge, and the form it is taking is data centers in space. The logic is physical, we don’t have to treat it as sci-fi or poetry. In the right orbit, solar panels never enter night, never sit under weather, never require energy-intensive cooling, and benefit from up to eight times the energy they would in the troposphere. The power is continuous and the cold of deep space is the largest heat sink available. So it logically follows that the solutions architect here would put the compute where the power is. This is not a thought experiment. In November 2025 Starcloud placed a data-center-class NVIDIA H100 in orbit and, weeks later, trained a model there; Google’s Project Suncatcher will fly prototype satellites carrying its tensor chips by 2027; Axiom and NVIDIA are putting compute nodes in low Earth orbit now. Stage two follows from the accelerant: that off-world compute, running continuously on free power, manufactures the knowledge that closes the thousand open problems of settlement, life support, in-situ resource use, radiation, agriculture. Stage three is the redundancy itself, a second node that finally takes consciousness off the planet—the current single point of failure. This is also the next rung of the energy ladder. A civilization advances the Kardashev scale by capturing more of its star’s output, and the next increment of solar energy is not on a rooftop or in a desert. It is in orbit.

If we’re being honest, we must acknowledge the hard problem: heat. In a vacuum there is no air to carry warmth away; convection does not exist; the only way to shed the heat of a gigawatt of computers is to radiate photons into the 2.7-kelvin space in the background, which requires radiator panels on the order of twelve hundred square meters for every megawatt dissipated. Now add radiation-hardening of the chips, launch mass, latency, and serviceability. None of this is small. And the company having made the most progress says so itself: SpaceX’s own public filing describes its orbital-compute ambitions as relying on “unproven technologies, or technologies that do not exist,” that “may not achieve commercial viability.” Read that as the strength it is. That sentence is what a genuine frontier looks like written into an SEC filing. The objection is the plan’s to-do list.

It takes an actor, and the actor is funded by what you resent

Achieving goals like this isn’t easy, and executing on the decades of work required isn’t done in a vacuum. Someone must turn the ratchet, run the engine, and take on the risk, and moving from an eye-roll-inducing proposition to a non-trivial probability is almost always path-dependent, mediated by the right actor at the right time. The probability that consciousness acquires a second node this century is not spread evenly across humanity. It is empirically carried by one program, because reusable heavy lift at this cadence requires a capital base, vertical integration, and a tolerance for public failure that no government agency will attempt and no committee will approve. Others have entered the field; none yet operates at the cadence or capital scale this demands. You are free to dislike the man. The algebra showing who is actually moving the number does not care whether you like him.

And here is the part the public has exactly backwards. Elon Musk’s 2025 Tesla compensation, the package an entire commentariat spent a season resenting, is the largest in corporate history and it pays him nothing up front. No salary. No guaranteed equity. It vests only if Tesla’s market value climbs toward eight and a half trillion dollars and the company hits a set of operational milestones that sound, on their face, absurd: twenty million cumulative vehicles, ten million self-driving subscriptions, one million humanoid robots delivered, one million robotaxis in commercial operation, and four hundred billion dollars of annual profit. If those things do not happen, no cigar for Elon. The compensation isn’t a grift. It is best described as a bet structured such that he is paid in proportion to outcomes most of us consider impossible.

Before you object to the size of it, look at what the businesses behind it are aimed at. The same man has staked his life’s work on two goals that make the robots look modest. One: a self-sustaining city of a million people on Mars, which he has stated flatly as a target within the next few decades. Two: terawatts of solar-powered intelligence manufactured off the Earth, a lunar base he describes as producing a hundred terawatts a year and building solar-powered AI satellites on site, hundreds of gigawatts launched to orbit annually, on the order of a thousand times the entire electrical draw of every data center operating today. Both sound preposterous, but they’re also stated, on the record, as operational objectives by the one person who owns the physical capacity to attempt them. The compensation everyone resents is not a scandal. The compensation is the fuel for the only serious attempt at the redundancy this essay is about.

So I propose that some of the critics sit with a question, because it is a question about them, not about him. Some people create. Some people build. Some people only consume. That’s fine; most of the major increments in any era are delivered by a small number of people and the rest of us enjoy living on the surplus. But we should not confuse our comfortable existence with the natural order of things. The natural state of humanity is poverty and suffering. The size of a builder’s reward as incentivized in a capitalist system is hardly evidence that building is not worth doing. Are we so uninspired that we cannot believe humanity is capable of this? Because the people who are going to build the backup are the ones who believe it can be done.

The one disposition God told you not to feed

Look honestly at the dominant public response to the man building the backup. It is not reasoned disagreement about heat rejection or capital structure. It is resentment. Resentment that he has what he has, plain and unadorned, and it is worth asking why that particular feeling, of all the feelings available, is the one we reach for.

Consider something about the oldest moral code in Western civilization. In reviewing the commandments that govern how people treat one another, notice what kind of thing each one regulates. Do not kill. Do not steal. Do not bear false witness. Do not commit adultery. Every one of those is an act, a behavior carried by hand. Something external. And then there is the last one, and it is different in kind: you shall not covet. Not “do not take.” Do not even want. It is the only commandment that reaches beyond the hand to the heart, the only one that prohibits not a deed, but a disposition. A feeling you are forbidden to entertain regardless of whether you ever act on it. Nine of them govern the hand. One governs the heart. Alone among everything a human can feel, envy is the one feeling we’re commanded to dispel.

There is a reason that feeling and no other earned its prohibition, and you can see it once you notice what envy is. Of the seven deadly sins, it is the singular one that provides no pleasure. Gluttony, greed, lust, pride—each enjoyable in the moment. Envy is empty: it produces nothing, builds nothing, lights no city, and won’t settle another planet. It is pure friction, a disposition that consumes us and leaves the world as poor as we found it. Of course it is the one feeling named in the commandments; it is the only one with no upside, even a transient one.

Envy refuses to look at one thing. The natural state of humanity is not the air-conditioned, well-fed, vaccinated present you are reading this from. For all of human history but the last few generations, roughly four in five people who ever drew breath lived in extreme poverty, and the average person died around the age of thirty. Hobbes named the baseline three and a half centuries ago and nothing in the data has refuted him: the natural condition of mankind is “solitary, poor, nasty, brutish, and short.” Everything above that line, the surplus food, the antibiotics, the clean water, the rocket, was built. It was arranged, recipe by recipe, by people who turned existing matter into configurations that do work. Work that struggles against the resistance of a universe that defaults to entropy and a species that defaults to envy. Prosperity is the anomaly. Poverty is the rule. The builders are the only reason that rule is broken.

So the choice is yours, and it is not complicated. You can resent the people arranging the next configuration, or you can join them. Consciousness is running on a single point of failure. The backup is being built by the people who believe it possible. The response observed by the public is repugnant. Resentment and envy are specifically prohibited by the oldest code we have, and now in the midst of the single most consequential redundancy project in the history of the world, you might finally see why.