Hourwiz Explains · video notes

Do Satellites Collide in Space? Who Dodges, and Who Can't

From the Hourwiz video “Look up. They're dodging.” · Checked Oct 19, 2026

Watch on YouTube (opens in a new tab)Instagram (opens in a new tab)Facebook (opens in a new tab)

The short answer: yes, but rarely. On Feb 10, 2009, a long-dead Russian satellite hit a working US one: the first time two whole satellites collided by accident (NASA). Most of the time, working satellites get out of the way. SpaceX's own report to the FCC works out to a Starlink dodge about every 76 seconds (our math). Dead satellites can't dodge.

Our map: where the satellites fly

Our test. We downloaded CelesTrak's public list of active satellites once (Oct 19, 2026, 04:55 UTC), worked out each satellite's height from its orbit, and moved every orbit to one shared moment (Oct 19, 2026, 02:52 UTC). Our script uses no outside libraries; the numbers below are its own printout.

So most satellites crowd into a few height lanes, but there's still a lot of room between them. ESA's report also finds the biggest concentrations of active, maneuverable satellites at 400–600 km.

Our take: it's not luck. It's room, plus a lot of dodging.

Who keeps them apart

• Trackers flag close passes. US Space Force trackers screen everything they track and send a Conjunction Data Message for each close pass. NASA: these messages “are not collision warnings”. Operators then work out the odds themselves.
• A common rule: move if the chance of a hit is worse than 1 in 10,000 (NASA, 2009; SpaceX calls it “the industry standard”).
• SpaceX says Starlink moves at 3 in 10 million, about 300 times more cautious than that rule (our math: 1 in 10,000 ÷ 3 in 10 million ≈ 333).

They dodge, a lot

• SpaceX told the FCC its Starlink satellites made 207,152 collision-avoidance maneuvers in six months (Dec 1, 2025 – May 31, 2026): 65,137 by Gen 1 satellites and 142,015 by Gen 2 (our sum).
• That's about 1,138 a day, or one every 76 seconds on average (our math: 207,152 ÷ 182 days).
• Per satellite, SpaceX reports about 36 (Gen 1) to 46 (Gen 2) maneuvers a year: roughly one every 8 to 10 days (our math).
• SpaceX says an onboard collision-avoidance system plans the dodges.
• It isn't perfect. In the same six months SpaceX logged 22 cases where a planned dodge didn't happen (late or bad tracking data, gaps after launch, a Space-Track outage).

Self-reported. These Starlink numbers come from SpaceX's own semiannual report to the FCC (filed Jul 1, 2026). We did the per-day and per-second math.

2009: the crash

• When and where: Feb 10, 2009, at 16:56 GMT (11:56 a.m. EST, our conversion), over northern Siberia, about 790 km (about 490 miles) up (NASA).
• Who: Iridium 33, a working US communications satellite, and Cosmos 2251, a retired Russian communications satellite (NASA). Cosmos 2251 was launched in 1993 and is thought to have stopped working about two years later (Kelso), so nobody could steer it.
• How hard: their orbits crossed at a nearly right angle, and they hit at more than 11 km/s, over 24,000 mph (NASA; our conversion).
• A first: NASA called it “the first accidental hypervelocity collision of two intact spacecraft”. Earlier known collisions involved a satellite or rocket stage and a smaller piece of debris, such as the French satellite Cerise in 1996.
• That day's screening: CelesTrak's public screening tool, SOCRATES, predicted the two would pass 584 m (about 1,900 feet) apart. It wasn't in that report's top ten; at the time of the crash it ranked 152nd. Close calls were everyday: that week Iridium 33 alone had 10 to 15 predicted passes within 5 km, and SOCRATES reported over 1,600 close approaches a day (Kelso).
• What it left: two clouds of debris. 783 larger pieces were catalogued by the end of March 2009 (NASA) and more than 1,300 by August 2009 (Kelso, our sum).
• 17 years later, pieces from that crash still show up in collision checks: ESA lists “Cosmos-2251 Fragmentation Debris” and “Iridium 33 Fragmentation Debris” among the objects in 2025 close-approach events.

The crash in the video is our own reconstruction, not footage.

Dead satellites still drift through the lanes

• ESA: “Not enough satellites leave heavily congested orbits at the end of their lives, creating a growing collision risk.”
• ESA's DISCOSweb statistics (Oct 14, 2026) count about 19,440 satellites still in space, about 15,900 of them still working, and 17,402 in low orbit. So at least ~1,500 dead satellites are in low orbit (our math, a lower bound: 17,402 − 15,900).
• Confirmed collisions are still rare. ESA counts 661 confirmed break-ups in orbit through 2025, and collisions are 0.89% of them: about 6, a handful (our math).

What our map assumes

• One snapshot. CelesTrak's list as downloaded at 04:55 UTC on Oct 19, 2026. 24 records with orbit data more than 3 days old were left out of the spacing check.
• Approximate orbits. Each satellite's average orbit, with Earth's equator bulge (J2) but no air drag. Good to a few km, which is fine for spacing measured in tens of km.
• Height is the average height above Earth's mean radius (6,371 km, NASA).
• The picture stretches heights so the lanes don't overlap. The counts in the table are not stretched.

Download our map

• Our script (.mjs, 8 KB): downloads CelesTrak's active list once, counts the height lanes, moves every orbit to one moment, measures nearest-neighbor spacing and does the FCC math. Run it with Node (node sat-sky.mjs); it needs no outside libraries.
• The script's run log (.txt, 1 KB): its own printout from Oct 19, 2026.

Please be kind to CelesTrak. CelesTrak updates its lists every 2 hours and asks for one download per update. The script saves the file and re-uses it instead of downloading again, and it stops on any error instead of retrying.

Edited for the web. In the script we changed the header comments and the folder it saves to (now ./raw next to the script). The code that counts and measures is unchanged.

N011 sat-sky run
Run time (UTC): 2026-10-19T04:55:21.993Z
Data: https://celestrak.org/NORAD/elements/gp.php?GROUP=active&FORMAT=json
File: raw/celestrak-active-2026-10-19T04-55-21-962Z.json (downloaded 2026-10-19T04:55:21.988Z)
Records in file: 16754
Low orbit (perigee >= 150 km, apogee <= 2,000 km): 15939
Top height bands (mean height, 25 km bins):
  450-475 km: 5340 satellites
  475-500 km: 4782 satellites
  350-375 km: 833 satellites
Share between 400 and 600 km: 12230 of 15939 (76.7 %)
Snapshot instant T (latest EPOCH): 2026-10-19T02:52:36.178Z · used 15915 · skipped stale (>3 d): 24
Nearest neighbor, all: median 123.4 km · 10th pct 51.8 km · min 0.0 km
Nearest neighbor, excluding < 1 km (docked/formation): median 123.5 km · objects with a partner < 1 km: 28
Volume 200-2,000 km: 1.269 trillion km^3
Even-spread box side: N=15939 -> 430.2 km · N=11,300 (ESA) -> 482.4 km
FCC: 207152 burns / 182 days = 1138.2 per day = 47.4 per hour = one every 75.9 s
Odds ratio 1e-4 / 3e-7 = 333.3

Images in the video

No outside photos or footage. Every dot on our map is a real satellite from CelesTrak's public data. Land outlines come from Natural Earth (public domain). Everything else, including the 2009 crash reconstruction, is drawn in code by Hourwiz.

Sources

Every fact on this page comes from these pages, opened on Oct 19, 2026. Things change: check the live page before you rely on it.

1. NASA Orbital Debris Quarterly News, Vol. 13, Issue 2 (April 2009): “Satellite Collision Leaves Significant Debris Clouds” Date, time, place, height, speed and angle of the 2009 crash; Iridium 33 operational, Cosmos 2251 decommissioned; 783 pieces catalogued by the end of March; the 1 in 10,000 rule.
2. T.S. Kelso, AAS 09-368 (2009), CelesTrak Cosmos 2251 “long-defunct”; SOCRATES predicted 584 m, ranked 152nd; 10–15 close passes a week for Iridium 33; debris counts as of Aug 5, 2009.
3. SpaceX letter and Gen 1 + Gen 2 semiannual reports to the FCC (Jul 1, 2026) 207,152 maneuvers (65,137 + 142,015), Dec 1, 2025 – May 31, 2026; 36–46 per satellite per year; 3 in 10 million threshold; onboard system; 22 outage events.
4. NASA CARA, Conjunction event prediction Space Force screening; Conjunction Data Messages “are not collision warnings”.
5. ESA Space Environment Report 2026 (report page) Not enough satellites leave congested orbits at the end of their lives.
6. ESA Space Environment Report 2026, Issue 10.1 (Sep 8, 2026), PDF 400–600 km concentrations; 661 break-ups, 0.89% collisions; Iridium 33 and Cosmos-2251 debris in 2025 conjunctions.
7. ESA DISCOSweb statistics (updated Oct 14, 2026) About 19,440 satellites still in space, about 15,900 working; 17,402 payloads in low orbit.
8. CelesTrak, active satellites (GP data, JSON) The list behind our map, downloaded Oct 19, 2026.
9. NASA NSSDC, Earth fact sheet Earth's radius and GM, inputs to our script.
10. Natural Earth, terms of use Public-domain land outlines used on our map.

Our map is one snapshot with approximate orbits, rounded as shown. The Starlink numbers are SpaceX's own. Hourwiz is not affiliated with CelesTrak, NASA, ESA, SpaceX, the FCC, the US Space Force or Iridium, and none of them made, reviewed or endorsed this page or the video.

No sponsor and no affiliate links on this page. Vendors are named so you can check them yourself, not as recommendations.

Follow Hourwiz

• YouTube@hourwiz (opens in a new tab)
• Instagram@hourwiz.ai (opens in a new tab)
• Facebookhourwizhq (opens in a new tab)

Get the week’s videos in one email

All video pages · Back to Hourwiz home