The Price of Perfection

Listen to this article

Sky — Full narration (20:59)

0:00

/1259

1×

The Summary

It was past midnight in Dublin when I stopped counting pages and started counting zeros.

The SpaceX S-1 had landed on my desk that morning — 377 pages of orbital mechanics, satellite diagrams, and a soaring narrative about humanity's multi-planetary future. The radiator was ticking in the corner. A cold mug of tea sat on top of a printed page I had not meant to print: the share count table on page 246, where the numbers live. I had read the rest the way you read a contract: skipping the poetry, hunting for the clauses.

The company was asking for $1.75 trillion.

Not a range. Not a negotiation. A fixed price: $135 per share, 555.6 million shares, $75 billion raised. The largest IPO in history, priced before the market had spoken a single word about what it thought the company was worth.

Outside my window, the Liffey was dark and quiet. Inside the prospectus, three businesses were asking to be valued as one. Starlink, a satellite network with 10 million subscribers whose revenue per user was falling. A launch division that sent rockets to orbit every two days but had not yet proven its next vehicle could come home. And xAI, an artificial intelligence unit that had burned $14 billion in a single year on margins that were getting worse.

Together, they had generated $18.67 billion in revenue in 2025. They had also lost $4.94 billion.

The public would be allowed to own 4.4% of the company. The rest was locked inside a structure that gave one man 82% of the votes.

Aswath Damodaran, the NYU professor whose valuations have become the industry's benchmark, published his analysis four days later. His number: $1.25 to $1.3 trillion.

The gap between his figure and the IPO price was $500 billion. That is not a rounding error. That is a bet — and the market may be forced to take it whether it wants to or not.

Four Point Four Percent

The number that would not leave me alone was not the valuation. It was the float.

Four point four percent. Of the roughly 13 billion shares that will exist after the IPO, only 555 million will be available to the public. The rest are Class B shares with ten votes each, all held by Elon Musk, plus restricted stock and options that will not trade for months or years.

I kept returning to this number because it reframed everything else. When you hear "$1.75 trillion," you imagine a market where buyers and sellers meet and negotiate. But SpaceX is not offering a market. It is offering a closet.

The entire public — every pension fund, every retail investor, every index tracker, every sovereign wealth fund — gets a room the size of a closet. Musk keeps the house.

And here is the part that changes the arithmetic: index funds do not care how big the closet is. They are required to buy whatever weight the index assigns them. If SpaceX joins the Russell 1000 at an estimated 2% weight, the roughly $10 trillion in passive funds tracking that index would need to purchase approximately $200 billion in SpaceX stock.

The public float at IPO is worth about $77 billion.

Forced buying would exceed available supply by roughly 2.6 times. That is not speculation. That is arithmetic. And it is why the $500 billion gap between Damodaran's valuation and the IPO price may not close — at least not quickly.

Three Businesses, One Price Tag

The prospectus organises SpaceX into three segments. Reading them in sequence, I realised the company was asking investors to believe in three different futures at the same time.

Starlink came first. It is the only business that makes money, and it is the reason SpaceX can file for an IPO at all. In 2025, the satellite network generated $11.4 billion in revenue — 61% of the company's total — with a gross margin of 48%, up from 37% a year earlier. It has 10.3 million subscribers, more than any other satellite communications network on earth.

But every new subscriber SpaceX signs is worth one-third less than the one it signed two years ago. Average revenue per user has fallen from $99 a month in 2024 to $66 in the first quarter of 2026. I pulled up my own Starlink bill on a second monitor — $110, Dublin pricing, and climbing even as the global average falls. The reason is expansion: SpaceX is growing by signing customers in markets where it cannot charge American prices. The satellite constellation also ages. The company must replace about one-fifth of its satellites every year just to maintain service. Since the beginning of 2023, SpaceX has invested more in its satellite business — $11.4 billion — than it has building Starship and its launch infrastructure — $8.4 billion. The treadmill is real.

Then there is the launch business. This is where SpaceX's reputation lives. The Falcon 9 completed 165 launches in 2025, nearly one every two days. It is the workhorse of the global space industry, with a gross margin of 67% — the best in the company. It has no real competitor at scale: Blue Origin is years behind, Rocket Lab is too small, and the Chinese space programme does not serve Western customers.

But the valuation depends on a rocket that does not yet work the way it is supposed to. Starship — the massive vehicle designed for Mars, for orbital data centres, for replacing Starlink's satellites cheaply — had its third test flight in May 2026. It deployed dummy satellites and two test vehicles. It also failed to relight its engines for a controlled return to Earth, which is the entire point of building a reusable rocket. Satellite analyst Tim Farrar wrote to clients that without full reusability, "the cost of launch on Starship may not be much lower than Falcon 9" — potentially $100 million per launch, or $1,000 per kilogram. If that turns out to be true, the Starlink replacement costs in the financial model will not fall the way the prospectus suggests they will.

I set the prospectus down and looked at the radiator. It was still ticking, steady as a clock, indifferent to the $100 million per launch problem sitting on my desk.

Then there is xAI. This is the business that burns the most and promises the most.

In 2025, xAI generated between $1 billion and $2 billion in revenue, growing at 22% annually. It also spent $9.1 billion in capital expenditure and $5.1 billion in research and development. The gross margins are the lowest of the three segments, and they are deteriorating. The prospectus mentions Colossus, xAI's compute centre in Memphis, which has been leased to Anthropic for $1.25 billion per month — roughly $15 billion in annual revenue. On June 5, Google reportedly agreed to pay SpaceX $920 million per month for compute as well. These are real numbers from real customers.

Here is the contradiction the prospectus does not resolve: xAI plans to compete with Anthropic in the enterprise AI market. SpaceX is both Anthropic's landlord and Anthropic's rival. It is building the infrastructure that trains Anthropic's models while simultaneously building its own models to compete against them. This is not a stable arrangement. It is a temporary alignment of incentives that will end the moment one side decides the other is more valuable as a target than as a tenant.

Damodaran doubled his revenue target for xAI from $80 billion to $160 billion after reading the filing, but he also cut his target operating margin from 45% to 25%. The prospectus claims a total addressable market of $28 trillion, with $26 trillion from AI alone. Damodaran called this "bordering on fantasy."

The company also won $6.45 billion in Space Force contracts in May, and Starlink secured an American Airlines in-flight connectivity contract the same month. These are not speculative. They are signed, contracted revenue. The businesses are real. The question is whether they are worth $1.75 trillion today.

The Profitability Question

One number I expected to find in the prospectus and did not, at first, was the one that changes the framing of the entire debate.

On a GAAP basis, SpaceX lost $4.94 billion in 2025. That is the number the S&P 500 uses when it asks whether a company is profitable, and it is why S&P Global rejected SpaceX's request for fast-track index inclusion on June 4.

On an adjusted EBITDA basis, SpaceX generated $8 billion in profit. It is cash-flow positive. It is not a money-losing company in the way a pre-revenue startup is money-losing. It is a company investing aggressively — $14 billion in capex and $9 billion in R&D — and reporting the cost of that investment as losses.

This distinction matters. It means the S&P exclusion is partly a technicality: GAAP profitability penalises heavy R&D spenders, and SpaceX is one of the heaviest in history. Some investors will look at the EBITDA figure and conclude the profitability requirement is a timing problem, not a fundamental judgment. They may be right.

But S&P was clear: "Exceptions to the financial viability, seasoning, and IWF requirements should not be granted solely based on market capitalization." The company must be profitable under GAAP in its most recent quarter and over the trailing four quarters. SpaceX fails both tests. It also fails the 12-month seasoning requirement and the 10% minimum float requirement.

S&P had consulted investors about relaxing all three rules. It rejected all three proposals.

FTSE Russell made the opposite choice. On May 26, it adopted a rule allowing newly listed megacaps to join its indexes after just two months of trading. SpaceX could enter FTSE Russell indexes by August 2026. Nasdaq's own fast-entry rules could bring SpaceX into the Nasdaq 100 by September.

This split is the quiet story of the IPO. S&P says no. FTSE and Nasdaq say yes. And the difference between those answers may determine whether SpaceX trades at Damodaran's number or at the IPO price.

But accounting rules only tell you who gets to play. They do not tell you what the game is worth. For that, you need to read the assumptions.

What Perfection Requires

To understand the $500 billion gap, I stopped reading the prospectus and started reading the assumptions.

Damodaran's DCF model projects ten years of cash flows for each of SpaceX's three businesses, discounts them at 8.37%, and sums them up. His inputs: Starlink reaches $200 billion in annual revenue at 60% operating margins. Launch reaches $50 billion at 45% margins. xAI reaches $160 billion at 25% margins.

These are optimistic but not impossible. The question is what the IPO price requires beyond them.

At $1.75 trillion, SpaceX trades at roughly 94 times revenue. For context: Meta trades at 7 times. Tesla trades at 7 times. The combined market capitalisation of every publicly traded space company on earth — Rocket Lab, Planet Labs, Virgin Galactic, and the rest — is roughly $15 billion. SpaceX is asking for more than 100 times that.

This is also the largest IPO in history by a wide margin. Arm Holdings went public in 2023 at a $54 billion valuation. Saudi Aramco, the previous record holder, raised $25.6 billion in 2019 at a $1.7 trillion valuation — but Aramco had decades of proven profitability and the world's largest oil reserves. SpaceX is attempting to raise three times Aramco's capital at the same headline valuation, with a fraction of the operating history and none of the profitability.

To justify the IPO price, you would need Starlink to reach roughly $50 billion in revenue at 50% margins, generating $25 billion in operating income. You would need Starship to cut launch costs tenfold by 2028 and become the only vehicle capable of deploying next-generation satellites at scale. You would need xAI to capture $30 billion or more in AI infrastructure revenue while competing against OpenAI (last valued at $300 billion), Google, and Anthropic (which filed its own IPO on June 1).

Each of those outcomes is possible. None is probable. If you assign each business a 60% chance of hitting its targets — generous for moonshots — the probability that all three succeed simultaneously is 22%. The IPO price implies the market is pricing in something closer to 60-70%. That is the premium perfection demands.

That is what perfection looks like in a spreadsheet. The question is whether the index rules will force the market to believe in it anyway.

The Governance Trap

I flipped to the risk factors section, page 187, where the lawyers list everything that could go wrong. The governance structure is buried three pages in, but once you see it, you cannot unsee it.

SpaceX will have two classes of shares. Class A shares, sold to the public, carry one vote each. Class B shares, all held by Musk, carry ten.

After the IPO, Musk will control more than 82% of the voting rights.

This is not unusual for founder-led companies. Meta has a similar structure. So does Alphabet. But those companies have decades of operating history and proven business models. SpaceX has an unproven rocket, a satellite business with declining per-user revenue, and an AI unit competing for market share against better-funded rivals.

The governance concentration means shareholders cannot restrain capital allocation decisions. If Musk decides to pour another $50 billion into xAI data centres after it becomes clear the AI market is smaller than anticipated, shareholders cannot stop him. The prospectus warns of "significant" future share issuance, which hints at a possible merger with Tesla. Shareholders cannot block that either.

Damodaran put it plainly: "If you add to this mix the antipathy that exists between Musk and Sam Altman, you have the potential for a UFC match between two monstrous egos, funded by tens of billions of dollars of shareholder money."

The amended S-1, filed on June 1, added a new risk factor that caught my attention: water. SpaceX now warns investors that "water scarcity, drought conditions, competition for local water resources, or regulatory restrictions on water use could limit our ability to obtain sufficient water for cooling" its data centres. This is not an abstract risk. It is the kind of line that appears in a prospectus because the lawyers have already seen the problem.

At $1.75 trillion, investors are paying a premium for a company they cannot control, built on infrastructure that may not have enough water to cool it.

What I Might Be Wrong About

The strongest argument against my thesis is not that Damodaran's numbers are wrong. It is that DCF models systematically undervalue optionality.

SpaceX controls the only launch system capable of deploying Starlink at scale. If Starlink reaches 100 million subscribers at even $40 per month — conservative for a global broadband near-monopoly — that single business generates $48 billion in annual revenue. At 50% margins, that one segment is worth more than $500 billion on a DCF. Add the launch monopoly and xAI's compute optionality, and the bull case writes itself.

I might also be wrong about the index effect working the way I expect. Index inclusion creates demand, but it also creates supply: insiders unlock shares on a schedule, and early investors take profits. The forced buying may be absorbed by forced selling. The closet may be bigger than I think.

I might be underestimating the value of the Space Force contracts, the Google compute deal, the Anthropic lease, and the American Airlines contract. These are not speculative revenue projections. They are signed agreements with named counterparties. They suggest the businesses are further along than the financial statements imply.

And I might be wrong that the float matters as much as I think it does. SpaceX is not the first company to go public with a low float. Snap did it. Pinterest did it. In both cases, the stock was volatile but eventually found its level. The market is patient in ways that individual investors are not.

I want the clean version of this story to be true — that the valuation gap is irrational and the index effect is mechanical — because it makes my own discomfort with mega-IPOs feel like evidence rather than instinct. But the truth is more complicated. SpaceX is building critical infrastructure for multiple trillion-dollar markets simultaneously. The optionality alone may be worth more than a ten-year DCF captures.

All of that may be true. But the S-1 will still exist. The 4.4% float will still be real. The governance structure will still concentrate power in one person. And the three businesses will still have to deliver on their promises at the same time, without a single one of them failing.

The Closet and the House

There is a difference between a great company and a great investment. I learned this the hard way, watching infrastructure I believed in trade at prices I could not justify.

SpaceX is a great company. It has transformed the launch industry, built the world's largest satellite network, and positioned itself at the centre of the AI infrastructure boom. It has executed on difficult technical problems that its competitors could not solve. It has real customers writing real cheques: Google, Anthropic, the US Space Force, American Airlines.

But at $1.75 trillion, it is asking investors to pay for perfection across three businesses that have not yet proven their most important capabilities, with governance that prevents course correction when things go wrong.

Damodaran values it at $1.25 to $1.3 trillion. The IPO asks for $1.75 trillion.

The gap is not about rockets or satellites or artificial intelligence. It is about what happens when the index funds are forced to buy a company that only lets the public own four percent of it.

When I closed the prospectus that night, the rocket photos were still loading in my browser — 377 pages of orbital trajectories and satellite renderings, the most expensive picture book ever published. The numbers were still in the footnotes, where they had always been.

Outside, the Liffey was beginning to catch the first grey light of morning. The radiator had stopped ticking. Somewhere in a data centre, xAI's Colossus cluster was drawing power and water and $1.25 billion a month from Anthropic, generating the compute that would train the models that would justify the valuation that would determine whether the closet was worth the house.

The door to that closet opens on June 12. The question is not whether there will be a line. The question is whether anyone in that line has actually read page 246.

And what's on page 246 is four point four percent.

A data engineer June 2026

Source note: This article draws on the SpaceX S-1 filing with the SEC (May 20, 2026), the S-1/A amendment (June 1, 2026), and Aswath Damodaran's post-prospectus valuation (June 4, 2026). Index rule reporting from Reuters, Bloomberg, and CNBC. Financial data and segment breakdowns from the prospectus. Subscriber projections from Quilty Space. Starship reusability analysis from satellite market analyst Tim Farrar. Index flow estimates from SpotGamma. All valuation estimates attributed to their sources. This article does not constitute investment advice.