The Americans finally woke up from the SpaceXAI fever dream

Well, the joke wrote itself.

SpaceX floated at $135 per share on 12 June, valuing Elon Musk’s freshly assembled bundle of rockets, satellites, Twitter debris and artificial-intelligence expenditure at $1.77 trillion. Four days later, the shares reached $225.64 and briefly pushed the valuation to around $2.68 trillion. Apparently, the company had become nearly a trillion dollars more valuable over a long weekend, mirroring Nvidia’s silliness.

The shares are now trading at roughly $128, following a small rebound from Monday’s close of $119.85. That still leaves them below the IPO price, around 43% beneath the peak and roughly $1 trillion of imaginary wealth lighter than they were a month ago. Nothing comparable happened to the actual business during that period. SpaceX did not lose half its rockets, Starlink did not fall out of the sky and Grok did not suddenly become much more useless than it had been the previous Tuesday. The market merely began sobering up.

The IPO was a perfect collision of American billionaire worship, Wall Street opportunism, GenAI hysteria, retail fear of missing out and the stubborn belief that attaching Elon Musk’s name to an expensive promise makes arithmetic vulgar.

Investors were offered a capital-intensive aerospace company, a physically constrained satellite broadband network, the smouldering remains of Twitter and an AI division burning billions of dollars. They briefly valued the collection above companies making vastly more money because SpaceXAI was not being assessed as a business. It was being sold as a prophecy.

America had once again confused an enormous number with evidence of intelligence.

The valuation was always bloody stupid:

SpaceXAI generated $18.67 billion in revenue during 2025, while recording a $2.59 billion operating loss and a net loss approaching $5 billion. The IPO therefore valued a loss-making, infrastructure-heavy company at roughly 95 times annual revenue. The peak pushed that towards 144 times revenue, which is the sort of multiple normally produced when a spreadsheet has been replaced by a telescope pointed at Mars.

There was no useful price-to-earnings ratio because there were no earnings. This inconvenience was treated as another failure of conventional finance to understand Elon Musk rather than a fairly important piece of information.

The defence was predictable. Traditional valuation methods could not capture the scale of Musk’s vision. The company was building the future. Normal rules did not apply. Investors had to look decades ahead, preferably far enough ahead that revenue, profit and commercial viability became too blurry to inspect.

Whenever someone explains that a company cannot be valued using revenue, earnings or cash flow, there is a decent chance they are preparing to value it using vibes and prayers.

SpaceX’s promotional material helped by estimating that its total addressable market across space, connectivity and AI could eventually reach $28.5 trillion. The space market itself accounted for only a small part of that figure. Most of the fantasy came from adding enormous portions of AI infrastructure, enterprise software, advertising, communications and future industries that either barely exist or have yet to trouble reality.

SpaceX’s business plan was apparently “all of it”.

Start with rockets. Add satellite broadband. Fold in Twitter and xAI. Include digital advertising, enterprise software, global telecommunications, orbital computing, lunar industry, Mars and anything else capable of making a PowerPoint slide look more expensive. When the company’s actual markets cannot support the valuation, simply draw a larger circle around the global economy.

The $2.68 trillion peak required investors to assume that Starlink would keep growing almost indefinitely, Starship would become reliable, launch demand would explode, xAI would catch its better-established competitors, direct-to-mobile would conquer telecommunications and orbital data centres would become economically sensible.

They priced nearly every favourable outcome as though it had already happened. The possibility that several might fail together was left somewhere outside the valuation model, presumably in their parked Tesla trucks.

A rocket company ate Twitter and called it synergy:

SpaceX’s absorption of xAI transformed an already difficult company to value into a corporate turducken.

During 2025, the AI division generated $3.2 billion in revenue and lost $6.36 billion from operations. During the first three months of 2026, it produced another $818 million in revenue while losing $2.47 billion. It also consumed $7.72 billion in capital expenditure during that quarter, compared with $1.05 billion for Space and $1.33 billion for Connectivity.

It was a cash furnace with a chatbot attached.

Wall Street looked at a division losing around three dollars for every dollar of quarterly revenue and admired the size of the fire. This has become standard behaviour during the GenAI bubble. Spending billions on chips, data centres, electricity and vaguely described infrastructure is treated as proof that a company must be building something immensely valuable. The more money disappears, the more ambitious management appears.

A normal company losing billions might be badly run. An AI company losing billions is apparently investing in scale.

The distinction is wonderfully convenient because nobody has to explain when this astonishing scale will produce profits matching the cost. The money is being spent now; the returns live safely in a future quarter beyond the current PowerPoint deck.

Meanwhile, SpaceX’s Connectivity division, mostly Starlink, generated $1.19 billion in operating profit during the first quarter. That was less than half the AI division’s loss. The profitable business was feeding the furnace, and investors were encouraged to admire the heat.

SpaceX calls the arrangement vertical integration. Rockets launch the satellites, satellites provide communications, communications support the AI services and AI supposedly creates demand for millions more satellites. This sounds impressively strategic until translated into ordinary English: Musk shoved several companies he already controlled into the same box and announced that they were now helping one another.

SpaceX would supply SpaceX so SpaceX could perform more SpaceX activities for SpaceX.

Investors treated this as proof of an enormous external market rather than a company manufacturing a growing proportion of its own demand.

SpaceX sells rockets to SpaceX:

Space launch is technically impressive, strategically important and commercially limited. Humanity requires only so many government payloads, military satellites, scientific missions and expensive commercial spacecraft, regardless of how enthusiastically Musk talks about becoming a Kardashev Type II civilisation.

The Space division generated $4.09 billion in revenue during 2025 and lost $657 million from operations. During the first quarter of 2026, it managed $619 million in revenue and lost another $662 million. This was the supposedly sensible aerospace foundation supporting the $2.68 trillion valuation.

Government work remains crucial. Roughly one-fifth of SpaceX’s 2025 revenue came from US federal agencies, including NASA, defence bodies and intelligence organisations. These are legitimate customers purchasing useful services, but they also reveal the scale of the supposedly unlimited private space economy. A significant part of it remains the American government paying SpaceX to do things governments have always paid aerospace contractors to do.

SpaceX escapes some of that limitation by becoming its own largest customer. Its rockets launch thousands of Starlink satellites, which keep its launch cadence high and lower the deployment cost of its communications network. That integration is commercially useful, but internal activity is not independent proof of endless external demand.

Tesco could double its haulage operation by moving the same tins of beans between its own warehouses every afternoon. The additional lorry journeys would be real. Britain would not have discovered a trillion-pound market for transporting beans in circles.

SpaceX now wants to repeat the trick with orbital AI. The rocket company launches the broadband satellites. The broadband network helps finance the AI company. The AI company proposes millions of computing satellites, which would require more launches from the rocket company. Each new SpaceX business conveniently creates demand for the previous SpaceX business, and every internal expense is presented to investors as the birth of another vast industry.

Better still, several of these promises depend on Starship working reliably. SpaceX itself warns that Starship delays could obstruct its next-generation satellites, mobile connectivity expansion and orbital AI infrastructure. Investors therefore did not buy a collection of independent growth businesses. They bought several promises standing on one unfinished rocket.

Even Starlink has a bloody ceiling:

Starlink is SpaceXAI’s strongest business. It is real, useful and profitable, which is precisely why it gets wheeled out whenever someone asks why the rest of the corporate lasagne deserves trillions of dollars.

Connectivity generated $11.39 billion in revenue and $4.42 billion in operating profit during 2025. Starlink had 8.9 million subscribers at the end of that year and 10.3 million by March 2026. Those are impressive figures. They do not repeal physics or household budgets.

Satellite capacity is finite. Each spacecraft can process only so much traffic, and users within the same geographic area share the capacity passing overhead. Starlink therefore works best where subscribers are scattered. Put too many customers in one place and the network becomes congested; expand the network and SpaceX must manufacture, launch, operate and eventually replace more satellites.

This is expensive infrastructure travelling at several kilometres per second, not software that can be copied indefinitely at negligible cost.

Starlink’s ideal customer lives where fibre providers cannot be bothered to go but still has enough money for the equipment and monthly subscription. In practice, that means wealthier rural households, remote workers, ships, airlines, mines, construction projects, military units and government agencies.

The commercially attractive overlap is remote, underserved and solvent. It exists, but it is limited.

The billions of people with the worst internet access are disproportionately located in poorer countries and rural communities. The International Telecommunication Union estimates that 96% of the world’s offline population lives in low- and middle-income countries, while internet access remains unaffordable in around 60% of those countries. Only 23% of people in low-income countries were online in 2025.

Starlink can reach many of those people. Reaching them and selling to them are rather different achievements.

SpaceX’s own figures show the tension. Starlink’s subscriber count rose rapidly, but average monthly revenue per subscriber fell from $99 in 2023 to $81 in 2025 and $66 during the first quarter of 2026. The next customers are worth less than the earlier ones, because expanding beyond affluent early adopters requires cheaper plans and poorer markets.

That is hardly surprising. The person living in a remote Californian property and working for a technology company may happily pay for Starlink. A low-income family in a rural region without reliable electricity is less likely to regard another expensive subscription and a satellite terminal as the final missing pieces of the household budget.

Those people look marvellous in an addressable-market slide. They are less useful at checkout.

Starlink can continue growing through aviation, shipping, mobile services, enterprise contracts, military communications and subsidised access programmes. Those markets can be profitable, but they also expose the sentimental nonsense surrounding the service. The money increasingly comes from airlines, cruise ships, industrial sites, governments and comparatively wealthy rural customers rather than the world’s poorest disconnected households.

The honest pitch is simpler: Starlink brings broadband to places terrestrial networks cannot economically reach, provided the customer has money or arrives carrying a government cheque.

This is a good infrastructure business with a hard physical ceiling and a hard customer ceiling. Wall Street briefly valued it as though every human without fibre was a future subscriber with an American salary and unlimited orbital bandwidth.

Orbit is not Elon Musk’s attic:

Starlink’s expansion also requires treating low-Earth orbit as free industrial floor space.

By March 2026, SpaceX had around 9,600 broadband and mobile satellites in orbit, accounting for approximately three-quarters of all active manoeuvrable satellites. During 2025, the constellation performed more than 1,000 automated collision-avoidance manoeuvres every day.

Automation makes that system safer than launching thousands of uncontrolled metal boxes and hoping for the best. The scale remains ludicrous. One private company has filled a shared orbital environment with enough machinery that its fleet collectively has to dodge potential collisions more than a thousand times per day.

That is less a satellite network than a motorway junction being managed by a very nervous computer.

The European Space Agency describes Earth’s orbital environment as a finite resource and reports that congestion and debris are driving collision-avoidance events higher each year. SpaceX receives the subscription revenue; every other operator receives more traffic.

Retiring satellites do not simply disappear either. Low orbit allows them to re-enter rather than remain as permanent debris, but that still means continually manufacturing replacements, launching them and burning the old hardware through the atmosphere. Research published in Geophysical Research Letters has examined the potential ozone effects of aluminium oxides produced during satellite re-entry, although the long-term scale of the risk remains uncertain.

Most broadband companies occasionally post customers a new router. SpaceX maintains a moving civilisation of routers in orbit, replaces them with rockets and deposits the remains into the upper atmosphere.

Meanwhile, SpaceX’s plans for as many as one million orbital computing satellites have alarmed astronomers. A 2026 European Southern Observatory study estimated that such a constellation could cause dozens of trails in individual Very Large Telescope exposures and losses reaching 28% of the telescope’s field of view under some conditions.

Musk did not invent the atmosphere, low-Earth orbit or the night sky. His valuation nevertheless assumes all three will continue absorbing the consequences of his expansion without presenting an invoice.

America briefly suspended arithmetic:

The IPO brought together nearly every embarrassing feature of modern American finance.

There was the billionaire cult, in which previous engineering success became permission to believe every subsequent promise. SpaceX built excellent rockets, therefore Grok must become a leading AI model. Starlink became profitable, therefore orbital data centres must become economical. Musk accomplished difficult things before, therefore every idea emerging from his corporate digestive system deserved a trillion-dollar option value.

This was religious devotion with a brokerage app.

Wall Street supplied the machinery. SpaceX offered only a small portion of its shares for public trading, leaving buyers fighting over less than 5% of the company. The scarcity helped drive the first-day valuation above $2 trillion, because apparently placing five chairs in a room with fifty excitable investors proves each chair is worth hundreds of billions of dollars.

Retail buyers were deliberately invited into the circus. SpaceX reserved 30% of its IPO shares for them, an unusually large allocation, while commentators noted that many were likely to be less concerned about the valuation. The fan club had been given early access to the merchandise, except the merchandise cost $135 and came with one-tenth of the useful voting power.

The shares then entered the Nasdaq-100 on 7 July, placing SpaceX into an index tracked by more than 200 investment products holding over $800 billion. Funds following the index had to buy the shares because that is what index funds do, providing a fresh pool of demand shortly after the flotation. Ordinary savers could therefore acquire a small helping of orbital Grok whether they had asked for one or not.

Musk, meanwhile, retained around 82% of the voting power. The public supplied billions in new capital, accepted the volatility and obtained shares with inferior voting rights. Musk retained control over the company they had just funded.

That arrangement captures the American corporate bargain beautifully. The public may provide the money, government agencies may provide a dependable revenue floor and shared natural resources may absorb the external costs. The billionaire keeps control and receives personal credit for every success.

When the price rises, he is a visionary. When it falls, the believers are reminded that true innovation requires patience.

When the company loses billions, it is investing in the future. When the government supplies a fifth of its revenue, it proves strategic importance. When Musk keeps 82% of the votes, that protects the mission. Every possible outcome becomes another reason to give the billionaire more capital and less accountability.

The diamond hands have been selected:

SpaceX raised $75 billion through the IPO. The company received an unprecedented pile of fresh money to finance rockets, satellites and AI infrastructure, while Musk retained overwhelming control. Whatever happens to the share price next, that part of the operation worked perfectly.

Musk won.

SpaceX won.

The banks collecting fees won.

Investors allocated shares at $135 and clever enough to sell them above $200 won.

Everyone who watched the price climb from $135 to $225 and concluded that the expensive shares had become a bargain got the commemorative bag.

Nothing fundamental about SpaceX improved by 67% during those four days. The company did not announce a profitable Mars colony between Friday afternoon and Tuesday morning. Buyers paid more because other people were also paying more, which remains one of Wall Street’s favourite methods of converting mass panic into professional analysis.

Those who bought at the peak are now down around 43%. Investors who arrived during the first-day frenzy are also heavily underwater, while even the original IPO buyers are slightly down at the current price.

The response from Musk’s true believers was, of course, immediate:

Diamond hands.

Because nothing says financial intelligence quite like watching 40% of your money evaporate and deciding the real problem is that everyone else lacks courage.

Selling at $220 makes you a coward. Still holding at $128 makes you a visionary. Add a rocket emoji, change your profile picture to Musk looking thoughtful and suddenly you are no longer trapped in a collapsing stock. You are personally helping humanity reach Mars.

This is the beauty of “diamond hands”. It converts being wrong into a personality trait. The larger the loss, the stronger the conviction. If the shares fall another 20%, that merely proves the market does not understand the mission. If they fall another 40%, buy the dip. If the whole thing implodes, post a photograph of a Falcon launch and explain that you were never investing for money anyway.

Unfortunately, Mars does not accept unrealised losses as legal tender.

The situation may become even funnier once employees and early investors are allowed to sell. Reuters reported that as many as 911.5 million additional shares could become eligible for sale after SpaceX reports its first quarterly results. The original frenzy was helped by a tiny public float and far too many excitable buyers fighting over it. Soon, some of the people who received their shares much earlier and much cheaper may finally be allowed to hand them to the diamond-handed faithful.

The current bagholders may therefore be less the final victims than the welcoming committee.

SpaceX did not lose a trillion dollars of rockets, factories or contracts. Nothing physical disappeared. The market simply stopped agreeing that each share was worth whatever the last Musk fan with a brokerage account and a Mars wallpaper was willing to pay.

The trillion dollars was never sitting in a vault. It existed briefly as a shared hallucination, passed from buyer to buyer until somebody looked down, noticed there was no floor and discovered that “to the moon” was not supposed to describe the direction of the valuation.