The economy is moving into the Sky

Happy Sunday.

You've probably seen the Starlink story this week. For more than a week, SpaceX cut off Ukraine's military satellite links whenever the dish was moving faster than about 100 miles an hour. That's fine for a pickup truck and too slow for some of Ukraine's attack drones and interceptors. Trump called Musk on Wednesday and asked him to lift the limit. On Thursday, Trump said, "The bottom line: He put it back." The same day, Zelensky said it still wasn't fully back.

Most of the coverage asked who gave the order. Musk points at the government. Defense Secretary Pete Hegseth points at Starlink's terms of service. Trump says he didn't order it and doesn't know who did.

I'd ask a different question. How long does a Starlink satellite last?

About five years. SpaceX said so in a letter to the SEC in May. Then it comes down, burns up, and a new one has to go up in its place.

So basically, a satellite network never gets finished. Every operator on earth has to rebuild its fleet on a clock, and the biggest operators in Europe and Canada ride SpaceX rockets to do it. When a rival replaces a satellite, Musk gets paid. When Musk replaces one, he pays himself, on rockets he reuses.

Musk's real monopoly isn't the network. It's the refill. He's using the cash Starlink throws off to wire that same network into AI, cars and robots on the ground. And it all sits inside a stock worth roughly $2 trillion, priced as if Washington will never do what it did the last time one company got this close to owning the sky.

The last time Washington saw this coming

On July 10, 1962, a NASA rocket put up Telstar, the first satellite a private company ever paid to launch. AT&T owned it, its Bell Labs built it, and that same month it carried the first live TV across the Atlantic. AT&T already ran the American phone network. Now it had a piece of the sky too.

Congress was already worried about exactly that. A handful of liberal Democrats led by Estes Kefauver and Wayne Morse filibustered the satellite bill over antitrust, and on August 14 the Senate voted 63 to 27 to shut them down. It hadn't broken a filibuster since 1927. Kennedy signed the Communications Satellite Act on August 31.

The law created COMSAT, a separate company with the U.S. monopoly on overseas satellite traffic. Half its stock went to the public, half to the phone carriers, and federal regulators sat on top. Congress's answer was a monopoly with a lot of owners and a referee.

In 1962, Washington passed a law. In 2026, it made a phone call.

Musk runs most of the sky

SpaceX runs more satellites than everyone else combined, and some of its biggest rivals ride its rockets.

Starlink has more than 11,000 satellites up. On PBS's NOVA, Johns Hopkins space engineering instructor Patrick Binning put SpaceX at about 70 percent of the world's 12,000-plus working satellites last year, and the Naval War College's David Burbach put it at about two-thirds of America's national security launches. Starlink added 1.6 million subscribers last quarter, to 13.6 million.

Now the challengers. Amazon $AMZN ( ▼ 0.95% ) has fewer than 400 satellites up for its Leo network and missed its own mid-2026 service target. Eutelsat's OneWeb has about 650. Europe's IRIS² plans 348 satellites for about $17.5 billion, with the first launch in 2029. Canada's Telesat wants a working network in early 2028.

OneWeb and Telesat rely on SpaceX's Falcon rocket to get up there. Amazon, whose founder owns his own rocket company, put three batches of its satellites on Musk's Falcon 9s last year. (I'd have paid to sit in on that meeting.) And on Monday, Deutsche Telekom, Orange, Telefónica and Vodafone agreed to explore a joint bid for satellite spectrum. Their plan is to pick an outside satellite partner and keep the ground equipment in Europe.

Even Europe's escape plan is a rental.

Today’s Sponsor

B2B support leaders who made AI work

AI agents are reshaping B2B support, and a few leaders are already ahead. Join Victoria Lowney of Inside Real Estate and Alex Rivas of HoneyBook as they share how they launched, scaled, and what's next. Hear how they built the business case, connected AI to data and systems, and reshaped team roles. Live Q&A. Register now to save your spot and get the recording.

Reserve your seat

The five-year clock

In orbit, the depreciation schedule is the moat.

SpaceX's May letter to the SEC put a Starlink broadband satellite's useful life at about five years, and its first-generation direct-to-phone satellites at about three. (Those are accounting estimates. Some will fly longer.) A fiber line in the ground can work for decades. A satellite in low orbit is closer to a set of tires.

SpaceX had about 9,600 broadband satellites up at the end of June. On a five-year life, that's roughly 1,900 replacements a year just to stand still. Every rival is on the same treadmill and pays a launch bill to stay on it, often to SpaceX.

Musk pays his own cost, and he's pushing it down. On September 28, Starship flew its first revenue-generating mission and dropped off 26 next-generation Starlink satellites, adding about 10 times the downlink capacity of a Falcon 9 Starlink launch. On the next flight, SpaceX will try to catch both the booster and the ship with the tower's mechanical arms. If that works, his cost per satellite falls again, and nobody else's does.

A railroad gets cheaper to defend once the track is laid. A constellation has to be relaid every five years, so the winner is whoever owns the cheapest way back up.

The sky pays for the brain

Starlink makes the money, and the AI business spends it.

SpaceX reported third-quarter results on Tuesday. Revenue was $14.6 billion, up 161 percent, and the AI business, which SpaceX now calls Super Intelligence, passed Starlink as the biggest segment at $7.56 billion. The profit came from somewhere else. Starlink's segment made $2.37 billion in operating income, about 85 percent of the company's total, while AI took $16.4 billion of the $19.6 billion SpaceX spent on capital projects. This quarter, SpaceX expects to spend $35 billion to $40 billion.

Now look at the other companies renting out AI compute. On the last day of Nvidia's conference in Berlin, CoreWeave $CRWV ( ▼ 2.46% ) fell 10 percent and Nebius $NBIS ( ▼ 0.27% ) fell 7 percent. An AI infrastructure executive told CNBC the number of these companies with no credit rating was spooking the market. Jensen Huang said, "The demand of the capital is much greater than the availability of the capital."

CoreWeave and Nebius lean on borrowed money for their chips. Musk's AI arm has 13.6 million monthly Starlink bills behind it, plus the $75 billion SpaceX raised in June in the biggest IPO ever. He got the money in the door before the squeeze. When capital is the bottleneck, the guy with his own cash machine sets the pace.

The ground game

A Tesla merger would close the loop.

On the same call, Musk said combining SpaceX with Tesla $TSLA ( ▼ 0.63% ) "would be a good idea," and both boards have formed special committees of independent directors. Musk, who calls himself "a conflicted party," stays out of their talks. The two companies already do plenty of business together. Tesla sells SpaceX cars, batteries and power equipment. The $2 billion Tesla put into xAI turned into SpaceX stock when SpaceX absorbed xAI and X earlier this year. And they share Terafab, a chip venture Musk says reaches volume production in late 2028, with "big error bars." On Wednesday, Musk said Tesla has started early production of Optimus, its humanoid robot, which he still calls "the biggest product on Earth."

The Wall Street Journal asked whether a merged company would be a futuristic Berkshire Hathaway or a mess. I'd pick a third comparison, the old Bell System. Western Electric built the equipment, Bell Labs invented what came next, AT&T owned the network, and you rented the phone on your wall. Each layer was the next layer's customer, and nearly every call in America paid for the whole thing.

In Musk's version, Starlink is the network and the AI is the brain. Tesla builds the batteries, cars and robots, and Terafab, if it works, makes the chips.

AI's first wave lived in data centers. The next one moves around: robotaxis, drones, robots, phones out of tower range. Machines that move need a link everywhere, and the network built for everywhere is in orbit. SpaceX is already preparing to expand its phone service on spectrum it bought from EchoStar for about $17 billion. Ukraine just showed what the owner of that network can do to a machine in motion. He can set its speed limit.

(Valerie Grimba of RBC said on CNBC this week that you should be able to explain to a friend what's in your fund. Try it with a merged Musk company: rockets, satellites, a social network, a chatbot, a $60 billion coding startup, cars, semis, robots, solar panels and a chip fab that isn't built yet. Your friend will need a minute.)

The bet inside the price

SpaceX's valuation is a bet that 1962 doesn't happen again.

In August, Deutsche Bank valued SpaceX's businesses one by one. It said the market was giving the AI business close to zero credit, and it valued connectivity, mostly Starlink, at up to 57.5 times the cash profit (EBITDA) it expects that business to earn in 2027. Its price target is $235, about 47.5 percent above where the stock traded Wednesday.

Nobody pays 57.5 times cash profit for a network they expect a regulator to cap. That multiple assumes no rate cap, no COMSAT and no Bell breakup anywhere on the calendar. It's a political bet sitting inside a stock price.

This month that bet got shakier, from both parties and from Europe. Sen. Richard Blumenthal, a Connecticut Democrat, opened an inquiry on Thursday into how Ukraine's service got cut without warning and how much the Pentagon depends on SpaceX. Sen. Kevin Cramer, a North Dakota Republican, said "we're putting too many eggs in one company's basket when it comes to our space assets." Poland's foreign minister, Radek Sikorski, whose government helps pay Ukraine's Starlink bill, went after Musk on his own platform: "As your European customers, we are watching you." Germany has committed about $40 billion to space defense by 2030.

Washington's 1962 fix had two halves. Sell shares to the public, and put a regulator on top. We're getting the first half. SpaceX listed in June at $135 a share, and Musk still controls it. A new Nasdaq fast-entry rule put it in the Nasdaq-100, and every fund that tracks it, after 15 trading days. S&P didn't bend its rules, so the S&P 500 is about a year away. Tesla has been in the S&P 500 since 2020. (A merger is one way around that wait. I don't know yet how S&P would treat it.) And on Monday, Musk promised a SpaceX share to "every kid in America" next year, with no details.

The shares are getting handed out. Nobody has written the second half.

Nobody's directing traffic up there

More of the economy runs through orbit every year, and orbit has no traffic cop.

GPS atomic clocks timestamp stock trades and ATM withdrawals. About 500,000 Canadians get their internet from Starlink, and Ukraine's army fights on it. SpaceX moved its satellites about 300,000 times last year to dodge collisions. Astrophysicist Jonathan McDowell told NOVA the number of working satellites could go from about 12,000 to more than 100,000 in a decade, and at that level he worries the dodging stops working. The main space treaty dates to 1967, and nothing since governs how close one satellite can get to another.

Concentration risk used to mean too much of your index in a few names. Now it also means too much of the physical economy running through one network, controlled by one man, in an orbit nobody polices.

The best argument against me is that this lead expires too. Amazon has built 1,000 satellites. Europe is writing checks. And Washington is SpaceX's customer as well as its regulator. SpaceX won more than $6 billion in U.S. contracts in the second quarter alone, so Washington has ways to push back if it wants to. But Telesat shows up in 2028, IRIS² in 2029, and Telesat is riding Falcons to get there. I'd give the moat at least three more years. What I don't know is what Washington does with them.

Two tells I'm watching. If SpaceX sells off on a regulation headline in a quarter when Starlink still grows faster than 60 percent, the market has started pricing a 1962 moment. If Amazon gets Leo running at scale without buying more Falcon launches, the launch moat is thinner than I think.

The first one is still up there

Telstar didn't make it to its first birthday. The day before it launched, the U.S. set off a 1.4 megaton hydrogen bomb 250 miles over the Pacific in a test called Starfish Prime. The radiation it left behind fried Telstar's electronics. It went quiet in late 1962, came back briefly, and died for good on February 21, 1963.

It's still up there, one of about 49,000 objects the government tracks. Washington spent the summer it launched making sure whatever came next would answer to somebody.

Trump's line on Thursday was "He put it back." The word I keep coming back to is "he."

Stay curious 😎

John

Today’s Sponsor

Leave Granola and get up to 12 months free of Wispr Flow Notetaker + Dictation

If you have paid time left on an individual Granola plan, we'll match it with a Wispr Flow subscription that includes Notetaker and dictation, and add bonus time, up to 12 months total. Sign in or create a Wispr account and submit proof of your plan to check eligibility.

Try Notetaker Free