Almost every argument about Starlink is about the wrong thing.
The usual debate is about launch: how many satellites it takes, how big the rockets are, how fast SpaceX can fly them. By that measure the whole enterprise looks impossible — tens of thousands of satellites, replaced forever. But launch cadence is the part SpaceX actually solved. Reusable Falcon 9s made orbit cheap and routine. If you’re hunting for the thing that doesn’t work, you’re staring at the one part that does.
The problems that matter are quieter, and none of them are about rockets.
Start with the money. A Starlink satellite lasts about five years, then burns up on the way down by design. That’s not a footnote — it’s the whole economic shape. A constellation has to replace roughly a fifth of itself every single year just to stand still. Not to grow. To stay the same size. SpaceX’s own audited filings show the connectivity capital spend climbing: $2.46 billion in 2023, $3.50 billion in 2024, $4.18 billion in 2025, and $1.33 billion in the first quarter of 2026 alone. That’s not a one-time build cost you pay down. It’s a fire you feed forever.
Compare it to the thing it claims to replace. Fiber laid in the ground in the 1980s still carries traffic today. When it needs more capacity, you swap the electronics on each end and leave the glass alone. Fiber separates the part that lasts decades from the part you upgrade. Starlink welds them into a single disposable object that falls out of the sky on a five-year clock. One architecture amortizes; the other incinerates.
Then there’s the neighborhood. Low Earth orbit broadband lives in a thin shell a few hundred kilometers up — proportionally thinner, wrapped around the Earth, than the skin on an apple. It’s a shared, finite, barely-policed volume, and right now SpaceX, Amazon, China, and several others are all racing to cram it full at once. There’s no air-traffic control up there. In 2009 a dead Russian satellite slammed into a working Iridium one and shattered both — and Iridium had done everything right. That’s the warning in one event: in a commons with no referee, behaving responsibly doesn’t save you from everyone who doesn’t.
And then there’s the part nobody measured first. When these satellites die, they vaporize in the upper atmosphere, seeding the stratosphere with aluminum and other metals. Direct sampling up there now finds spacecraft metal in roughly one in ten of the particles measured — and we’re scaling that injection up sharply before anyone understands what it does to the ozone layer or the climate. We’re running the experiment and reading the results later.
Here’s the part that should bother the boosters most: none of this is unique to Starlink. Amazon’s Kuiper, the Chinese state constellations, OneWeb, Telesat — every one of them is signing up for the same five-year treadmill, the same crowded shell, the same metal in the sky. Starlink is just the leading indicator. The whole field is on course to build, launch, and deliberately burn up satellites worth tens of billions of dollars a year, indefinitely, just to keep the lights on.
And there is a great deal riding on the optimistic version of this story. In June 2026 SpaceX went public in the largest IPO in history, near a $1.77 trillion valuation — pitched substantially on Starlink as a broadband business that competes with terrestrial fiber across the whole populated world. That premise is exactly what the full paper weighs against the company’s own filings. This is an independent analysis, not investment advice — no position, no prediction; only a question about whether the math under the story holds.
So is the technology bad? No. And that’s the actual point.
There’s a version of this that makes complete sense: fewer satellites, longer-lived, responsibly disposed of, aimed at the places nothing on the ground can reach — oceans, poles, aircraft, the real backcountry. It already exists. Iridium has run that model, profitably, for decades. Satellite broadband is a brilliant answer to “how do I get a signal in the middle of nowhere.” It’s a poor answer to “how do I replace fiber for everyone.” The trouble is it’s being built, sold, and valued as the second when the physics, the economics, and the atmosphere all say it’s only ever the first.
That’s the Starlink math problem. The rockets work. The business case for draping a private, disposable internet over the entire populated planet is the part that doesn’t close.
Want the whole thing — every figure sourced, every claim end-noted, the filings, the orbital mechanics, the atmospheric papers? It is a long read. Bring coffee.