SpaceX’s S-1 Flags a Cascade That Could Kill Low Earth Orbit

With two-thirds of active satellites already Starlink, a crash clock measured in days, and filings for up to a million more craft, the company’s own IPO risk language says a debris cascade could wipe out the shell that runs the internet.

One dead satellite is enough. At orbital closing speeds, a thumbnail-sized shard hits like a hand grenade, and the fragments become the next round of grenades. SpaceX put that cascade in its public filing as a material business risk. The margin left to absorb a single failure is already thin.

Key Takeaways

• Starlink now accounts for about 10,000 of the planet’s roughly 16,000 active satellites - call it two-thirds of everything still working overhead.
• SpaceX’s June 2026 S-1 warns that a debris-generating collision could trigger a cascading series that renders certain altitudes unusable and hit the business hard.
• Crash Clock math: if every craft stopped dodging, first collision between tracked objects fell from 164 days of margin in 2018 to about 2.5 days by May 2026, with roughly a 30% chance inside 24 hours.
• Starlink alone ran about 300,000 collision-avoidance maneuvers in 2025, up roughly 50% in one year - thrusters firing on the order of every couple of minutes, fully automated.
• Models put ~1.2 million objects between 1 and 10 cm and around 140 million in the millimeter class; almost everything under 10 cm stays invisible to tracking, so nothing dodges it.
• Blue Origin’s Project Sunrise targets 52,000 orbital data-center satellites; SpaceX has tabled proposals that scale toward 1 million for space-based compute - same low floors, 500–1,000 km.
• China’s 2007 Fengyun-1C hit and Russia’s 2021 Cosmos 1408 test still leave thousands of trackable fragments decades later; the U.S. 2022 kinetic ASAT ban binds only the country that announced it.
• At Starlink’s ~550 km parking altitude, atmospheric drag still works - a dead bird can reenter in about five to six years without a grabber mission.

---

Closing Speed Turns Thumbnails Into Grenades

When two objects cross paths in low Earth orbit, closing speed runs about 7–8 km per second - New York to Miami in roughly four minutes. At that velocity, a 1 cm fragment packs the energy of an exploding hand grenade. The first impact sprays hundreds of new pieces. Each of those pieces is now another high-energy projectile on a crossing path through the same crowded shell.

I care less about the poetry of “space is big” and more about density. Most of the valuable hardware sits in a few hundred kilometers of altitude where latency is low and launch cost is low. Everyone is fighting for the same floors. Once fragments multiply faster than the thin upper atmosphere can drag them down, you are in a race the sky can lose without launching another rocket.

What SpaceX Told Investors in the S-1

When SpaceX went public in June 2026 in what became the largest IPO on record, the S-1 had to list risks that could seriously hurt the business. Deep in that document, the company itself warned that a collision or debris-generating event could set off a cascading series of collisions, leave certain orbital altitudes unusable, and materially and adversely affect the business.

Lawyers write that language so no investor can claim surprise. I read it as the company acknowledging the physics in public. Starlink is already the internet for well north of 10 million customers - ranches, islands, ships, war zones. The same company is a roughly $2 trillion public name with plans that include up to a million satellites. The filing is not a blog post. It is the risk stack the market is supposed to price.

Two-Thirds of Working Orbit Belongs to One Fleet

Back in 2019, the whole planet ran on the order of 2,000 working satellites. Seven years later, SpaceX alone flies more than five times that old global total. Today’s rough census is about 16,000 active craft. About 10,000 of them are Starlink. One operator owns roughly 66% of the machines still answering commands.

That concentration is a product story and a systemic-risk story at the same time. Starlink is the cash engine bankrolling Starship. If the shell between about 500 and 1,000 km becomes a debris furnace, the company with the most hardware and the most revenue tied to that shell takes the largest hit. Blue Origin is chasing the same prize with Project Sunrise and a proposed 52,000-satellite orbital data-center fleet. Rocket Lab is racing into the same economy. The sky of 2026 is still a rounding error next to the constellations already on paper.

Kessler’s 1978 Threshold Still Sets the Frame

Don Kessler and Burton Cour-Palais laid out the mechanism in a 1978 NASA paper. Objects collide. They shatter into many fragments. More objects mean more collisions. The only natural cleanup is residual atmosphere - a faint drag that, over years, pulls debris down until it burns. Below a density threshold, orbit forgives mistakes. Above it, the debris population can keep growing for decades even if launches stop cold.

That is Kessler syndrome. The field still wears his name. The movie version is softer than the paper. We can track on the order of 40,000 objects today - softball-sized and up. Models put the real count above 10 cm nearer 54,000. Between 1 and 10 cm, about 1.2 million. Millimeter-class paint flecks and panel slivers: around 140 million. At 8 km per second, every one of those is a bullet. Everything under about 10 cm is effectively invisible. You cannot dodge what never appears on the screen.

The Crash Clock and 300,000 Dodges a Year

Researchers publish a metric called the Crash Clock: if every satellite stopped maneuvering, how long until the first collision between two tracked objects? In 2018 the answer was about 164 days. By June 2025 it was about five days. By May 2026, about two and a half. At that level, the same math puts roughly a 30% chance of a first hit inside 24 hours if the dodging stops.

It has not stopped. Starlink satellites alone logged about 300,000 collision-avoidance maneuvers in 2025 - up roughly 50% year over year. Somewhere overhead, thrusters fire on the order of every couple of minutes, automated, continuous. The safety case for those crowded floors is software and fuel. Both objects in a close pass need to move. A dead bird, a hacked thruster burn, or a cloud of untracked shrapnel breaks the assumption.

I know how that sounds - like fear theater. The counter is the actual maneuver rate and the crash-clock compression. Vast empty sky is a mood. The telemetry is a traffic jam.

Storms, Failures, and Weapons Already Proven

February 2022: SpaceX launched 49 Starlinks. A geomagnetic storm heated the upper atmosphere as they deployed. Drag spiked. Roughly 38 of them were dragged down and burned within days. May 2024: the strongest solar storm in about two decades puffed the atmosphere enough that orbit predictions used for collision screening degraded for days, while thousands of satellites maneuvered at once. Ground systems that cannot predict cannot protect.

Some fraction of any large fleet dies on its own - battery, radio, flight computer. Scale that fraction across tens of thousands of machines and you get a permanent population of uncontrolled mass. Then there is intent. A satellite is a computer with a thruster. Own the thruster in a dense shell and you can steer kinetic damage through the busiest altitudes humans have ever built.

China’s 2007 test against Fengyun-1C near 860 km produced over 3,000 trackable fragments; most still orbit because air there is too thin to clean for decades. Russia’s November 2021 hit on Cosmos 1408 added on the order of 1,500 trackable pieces and forced ISS crews into return capsules as a contingency. The U.S. announced a kinetic ASAT test ban in 2022. It binds one country. China and Russia never signed on.

Rules, Autonomy, and Cleanup That Removes One Object at a Time

The FCC tightened U.S. disposal rules so a dead satellite must come down within five years of end of mission - down from a 25-year guideline. ESA wants net-zero debris by 2030. That helps future launches. It does almost nothing for junk already flying.

Software is Plan B. Starlink dodges autonomously and moves when collision odds approach a few in 10 million - far more cautious than the old industry habit of sweating near 1 in 10,000. That discipline is a big reason the maneuver count is so high. The blind spot stays the same: dead craft, hacked craft, and shrapnel do not take commands.

Plan C is physical cleanup. ESA’s ClearSpace-1 is built to capture one piece of ESA’s own leftover hardware and drag it to burn. Astroscale’s ADRAS-J already crept up on a dead rocket stage for close inspection. Rocket Lab is winning launch work in that niche. Cost per object can run to nine figures. Industry is still lofting thousands of new satellites a year. One down, thousands up. At that pace, cleanup is a rounding error.

Why 550 Kilometers Still Buys Time - and Why Starship Tightens the Race

SpaceX’s deliberate low parking altitude near 550 km still has a whisper of atmosphere. A dead Starlink can decay and burn in about five to six years without a rescue mission. That self-cleaning shell is real physics working in our favor. NASA’s long-term LEGEND debris model, run under current mitigation practices, projects roughly linear debris growth through 2100 - not a sudden runaway. Serious modelers will tell you a lot has to fail at once for a true cascade. If they are right, the odds sit under 1%.

I put real weight on that steel-man. Operators are also improving disposal discipline. But the launch curve is about to bend. Starship aims for full reusability - same booster flown hard, many times, cheap enough that more companies will put hardware up. SpaceX’s own filings for space-based compute scale toward a million satellites; Blue Origin’s Project Sunrise is 52,000 in the same low-Earth prize belt. Density goes up. Careless operators become a statistical certainty the way plane crashes stay rare per flight but certain over enough flights.

Humanity has run dangerous systems before and come out wealthier on the other side. I am still banking on that arc. The difference this time is that one cascade does not just kill a fleet - it can strand GPS, defense links, and the satellite internet that already serves more than 10 million customers, and it can lock us out of the altitudes we need to leave Earth. SpaceX has the most to lose if that shell dies. It also has the most hardware already living there. The risk is not theoretical. It is already in orbit, and the company’s own S-1 said so out loud.