NASA SpaceX Crew-10 launch, March 14, 2025. Photo: NASA/Aubrey Gemignani. Documentary image of crew transportation, not the November 2026 statement. Original source
#ElonMusk has offered the American government the chance to cancel #SpaceX’s contracts. Not one in particular: all the ones #Washington no longer wants. «Happy to cancel any that the government doesn’t want. Just say the word!», he wrote on November 6, replying to a supporter who suggested sending the #Democrats, fresh winners of the midterm #elections, a letter asking which contracts to kill. In the same message he added two forecasts: government #revenue will fall below 10 percent of SpaceX’s turnover this very quarter, and below 5 percent within a year. And he threw in the sales pitch: best product, lowest price.
Here is the intellectual con worth dismantling: a company that lives on #NASA certifications, federal launch #licenses and multi-year billion-dollar #contracts presents itself in public as if a shrinking percentage of revenue were a certificate of #independence from the state. It is not. That share measures how much the #government weighs in SpaceX’s income statement. It does not measure how much the government weighs on SpaceX’s very ability to operate, nor how badly the public client depends on its supplier, nor how many physical limits wait for the company outside the balance sheet. Those are four different quantities, and Musk’s rhetoric crushes them into one convenient number. The whole trick lives in that compression.
What a revenue share actually tells you
The number Musk is waving around, below 10 percent this quarter and below 5 within a year, is his own #forecast, not accounting: no filing has confirmed it yet, and the man pronouncing it is the interested party. That alone would justify handling it with gloves. But grant him the numbers: even if they were exact, they would still not measure independence from the state.
The third-quarter 2026 results, communicated by the company itself in the exhibit filed with the #SEC, portray a transformed business: 14.6 billion dollars in revenue, up 161 percent year over year. Of that, 7.56 billion comes from the #SuperIntelligence segment, the #AI division, which alone is worth more than half the total. #Connectivity, the #Starlink universe with its consumer, enterprise and also government customers, brings in 5.2. The #Space segment, the one with the #rockets and the launches, stops at 1.8 billion: it was the heart of the company, and today it hovers around 12 percent of revenue.
Operating income tells a different story, and that contrast appears nowhere in Musk’s message about the contracts, even though the data are just as official and sit in the same #filing. Of the quarter’s 2,791 million dollars of operating income, Connectivity generates 2,367, roughly 85 percent of the total; Super Intelligence, which out-bills everyone with its 7,560 million out of 14,575, contributes 431 million, about 15.4 percent; Space closes 7 million in the red. These are shares of the combined result, not #profitability margins of each segment, and operating income is not cash. But the picture is sharp: the segment that dominates revenue is not the one that carries the profit, and the historic rocket business produces none this quarter. The company proclaiming its emancipation from the state lives today, in operating-income terms, mostly off Starlink’s antennas, which also serve government customers.
When turnover grows 161 percent in a year, driven by AI, any line that grows more slowly shrinks in percentage terms without anyone cutting it. That is #arithmetic, not emancipation: you do not need to reduce the numerator, you just need the #denominator to explode. I do not have the data to claim that the whole announced decline depends on this effect, nor that public revenue is not falling in absolute terms. But that is precisely the point: Musk does not provide those numbers either. He throws out a forecast percentage and lets the public read it as liberation. Meanwhile, the 946 million that NASA added to the astronaut transport contract, bringing it to 5.92 billion in total, has not vanished: it has merely become optically small next to the AI monster. And beware of another trick already in circulation: that 5.92 billion is the multi-year value of a contract with performance running through 2030, not a quarterly receipt. Comparing it with three months of revenue is barroom #accounting.
One more detail the single number buries: not all of SpaceX’s public contracts sit in the Space segment. Starlink also serves government customers inside Connectivity, and the breakdown of public revenue by segment does not appear in the documents I consulted, which also do not allow anyone to compute which part of the company would bleed if the government ever said the famous word. One more reason to distrust anyone who reduces everything to a percentage: we do not even know exactly what it is made of.
What the percentage does not measure: the stamps
There is a first quantity the revenue share cannot capture by definition: #authorizations. Every SpaceX launch and reentry goes through the #FAA, with licenses for vehicle operators, separate licenses for launch and reentry sites, and financial responsibility requirements. The documents I have read record no revocation and no threat of one: the point is not an ongoing retaliation, it is structural #dependence. A rocket company can give up Washington’s revenue far more easily than it can give up Washington’s stamps. Driving public revenue down to 5 percent does not move that #regulatory dependence by a millimeter. This is why the indicator Musk flaunts is rigged at the source: it measures only the slice of the relationship with the #state that suits him, and stays silent about the slice where the state has the last word every single day.
Then there is the procedure. Musk has rescinded nothing: he has invited the other side to do it. Federal procurement law, the #FAR, assigns the decision to a precise figure, the contracting officer, who may terminate a contract for convenience only when it serves the government’s own interest. A settlement phase follows: the general rule provides fair #compensation for work done and preparations made, with a reasonable allowance for profit, and the same regulation contemplates a no-cost settlement when the contractor accepts one and specific conditions are met. Musk’s offer, taken literally, resembles exactly that consensual availability. What it would truly cost depends on contract, clauses and negotiation: I have not read the clauses of SpaceX’s individual contracts and I will not pretend otherwise. What can be said without reading them: between a tweet and a formal termination lies an ocean of #bureaucracy, deadlines, public-interest assessments and political responsibility. «Just say the word» sounds like courage; it works as an offer that costs its maker nothing and loads the entire #political bill onto whoever would have to accept it.
The client’s dependence: who needs whom
Then comes #substitutability, that is, the question of who needs whom. NASA’s #CommercialCrew program has rested since 2014 on two providers, SpaceX and #Boeing, precisely so as to depend on neither. One leg has been limping for years. The July 2026 report by the congressional watchdog photographed an awkward situation: Boeing’s #Starliner capsule still not certified for human flight, a cargo mission planned before the #certification review, launch dates under review. All of this after a 2024 test flight that returned without crew due to #thruster failures and helium leaks, a NASA report describing hardware failures, systems engineering shortfalls and cultural breakdowns, and two Boeing missions cancelled from the contract. It is July’s photograph, not today’s status, and it does not rule out later progress. But the economic asymmetry survives every caveat.
The NASA page updated on September 18 documents the sole-source purchase from SpaceX of three additional missions to the International #SpaceStation, from a provider certified since November 2020, with twelve crew rotations already flown on #Dragon and #Falcon9. This is not charity, and calling every public purchase a subsidy is mental laziness: the agency buys, at a fixed price, a service it needs. The problem lies elsewhere. For the client, replacing a crew transport service is not switching couriers: certifications, availability and continuity impose timelines and costs that the documents I consulted do not quantify. For the supplier, by contrast, that client’s percentage weight shrinks on its own if other businesses grow faster. The gap between those two clocks is pure negotiating #power. When a contractor can publicly tell its own state «cancel whatever you don’t want», in a relationship where the client has every reason to think ten times, we are not watching a triumph of the free #market: we are watching a position of strength built also on twelve years of public orders, certifications and awards, now resold as independence. The client financed the lever it is being invited to step away from.
The physical limits: where no ratio protects you
There remains the quantity that the autonomy rhetoric does not even graze: material limits. In a single quarter SpaceX buried 19.6 billion dollars of #capex, more than its entire revenue for the period, with about 16.4 billion, roughly 84 percent, going to Super Intelligence: #datacenters, #chips, computing #infrastructure. Capex is not a loss and does not equal burning cash: it is spending to build future capacity. But it points unambiguously at the direction of the bet: SpaceX’s expansion runs through concrete, #energy and #water in industrial quantities, that is, through everything that depends on #permits, power grids and local politics.
And local politics has just spoken, twice and in two languages. First an ally spoke: #GregAbbott, Republican governor of #Texas and a political friend of Musk, ordered the state’s power #regulators on August 3 to run a full #audit of every data center project queued for grid connection, ruling that no project may advance before verification. The reason sits in his own numbers: over 474 gigawatts of connection requests, more than five times the state’s record peak electricity demand, roughly 90 percent of them data centers. These are requests, not real consumption, and filing a request does not guarantee a project gets built. But when the plug-in applications quintuple the consumption record of an entire state, even the industry’s best friend understands the #grid cannot carry the fantasy.
Then the ballot box spoke: Texas elected Democrat #GinaHinojosa as governor. During the campaign she promised a #moratorium on all new data center approvals until the legislature passes protections on water, #transparency and #taxation, with the twenty-year tax breaks granted by a law signed by Abbott himself in her sights, according to her own campaign platform. The Texas Tribune’s post-election report, the source of these contextual elements, records her inauguration set for January 19, 2027, a Democratic-majority state house, anger at data centers documented in red and blue counties alike, and Musk’s comment after the result: data centers in space are looking pretty appealing right now.
A campaign promise is not a law, a moratorium on new approvals does not switch off operating facilities, and the documents I have read do not say where SpaceX’s AI infrastructure physically sits or which of its projects wait in the Texas queues: the risk concerns the projects actually touched by audits and future moratoriums, not the entire capex by decree. The logic of the contrast, however, survives every footnote. On the federal contracts front, Musk brags he can do without the government; on the physical infrastructure front, the game is played with plugs, #gigawatts, environmental permits and local taxes, things no revenue growth dilutes, because they sit in the denominator of nothing. A data center without available #electricity is an expensive shed. And public power, Republican or Democratic, has just demonstrated it can freeze the project queue with one letter to two regulators.
The word that does not exist
Seen whole, the move is this: the offer to cancel the contracts is free for the one making it. It starts no #procedure, dumps the political and administrative costs on the other side, and leans on a forecast percentage that can fall on its own if commercial business grows faster than public business. The real dependence runs on other rails: the federal stamps without which nothing launches, a hard-to-replace position in astronaut transport that makes every threatened cut hurt the client first, and a hunger for energy, water and permits that hands the company’s expansion to local regulators.
The deeper trouble is that the state plays several roles here at once: it is the client buying the missions, the #regulator signing every license, and the architect of its own supplier’s dominant position. The American state outsourced access to space with such enthusiasm that it finds itself, twelve years later, with one dominant certified supplier and a second that in July still was not. Texas rolled out the red carpet for data centers so generously that it had to invent the audit after the fact, the handbrake pulled with the truck already on the slope. In both cases the public built the private player’s strength with its own hands, and now discovers that renegotiating costs more than conceding.
Musk says: just say the word. The truth is that the word does not exist. What exist are procedures, #clauses, #settlements, competent authorities, alternative suppliers to certify, power grids to size and state laws to write. Until all of that happens, every revenue percentage flaunted as a certificate of freedom should be read for what it is: an interested forecast, used to sell an independence that number does not measure.
Raffaele Di Marzio
All my “insane” books on cybersecurity and governance are here 👇 https://cyberium.limited/bookshelf.html
Photo: NASA/Aubrey Gemignani. Source: NASA. SpaceX images carry their own credits and links in the respective captions.