Generated by Codex with GPT-5

When satellites stop inspiring wonder

Phil Plait begins with a change in perspective. As a teenager, he felt awe when he first spotted a satellite sliding across the night sky. Today the same sight fills him with dread. The difference is scale. Satellites are no longer occasional visitors overhead. Companies and governments are building constellations of thousands of spacecraft, and proposals for much larger swarms could turn low-Earth orbit into industrial infrastructure.

Starlink shows how quickly that transformation is happening. Plait notes that SpaceX had about 4,000 Starlink satellites in orbit when he wrote about the problem in 2023. The total has since passed 10,000, exceeding the number of all other operational satellites combined. In January 2026 SpaceX applied for permission to launch as many as one million more satellites, with the goal of creating an orbital data center: a distributed network in which each satellite would handle a small share of computational work and transmit the results back to Earth.

The pitch has an appealing surface logic. Data centers consume large amounts of electricity and water, and the rapid expansion of artificial intelligence is increasing that demand. Satellites could draw power from sunlight and radiate heat into space without using water for cooling. But Plait argues that this framing hides the much larger system required to put a million machines into orbit and keep replacing them.

The launch schedule is the warning

The proposal becomes harder to defend when translated into launches. If each orbital data-center satellite weighed about two metric tons and SpaceX’s Starship could carry roughly 100 metric tons into low-Earth orbit in regular operation, each flight would transport around 50 satellites. Deploying one million of them would therefore require about 20,000 launches. Because satellites fail and need replacement after several years, maintenance alone could eventually demand roughly 10 Starship launches every day.

That cadence would carry environmental costs of its own. Rocket launches release greenhouse gases, contribute to ozone damage, create intense noise, and can harm nearby habitats. Reentry creates another pollution stream. Retired satellites burn up in the atmosphere, releasing vaporized metal and plastic into a region of the planet that scientists are still trying to understand. At least one Starlink satellite is already reentering each day, Plait writes. A vastly larger constellation would turn an occasional event into a continuous industrial process.

Orbit itself would also become more dangerous. Satellites move fast enough that a collision can scatter destructive fragments across space. Those fragments can hit other spacecraft, generating still more debris in a cascade known as Kessler syndrome. Low-Earth orbit gradually clears some debris as gravity pulls objects into the atmosphere, but multiplying the number of satellites by thousands would make collision avoidance much harder and the consequences of failure much worse.

The sky is part of the cost

Plait’s deepest concern is astronomy. Satellite trails already interfere with ground-based observations. A study he cites found that with roughly half a million satellites in orbit, essentially every observation made by the Hubble Space Telescope would contain at least one contaminating satellite. Reentry debris could also add to sky glow, making faint cosmic objects harder to see. The damage would not be limited to professional observatories. Ordinary stargazing would increasingly take place beneath a sky crowded with moving lights.

SpaceX is not the only source of pressure. China has filed plans for a 200,000-satellite network, while Amazon and Blue Origin have proposed thousands more spacecraft. Plait also points to Reflect Orbital, a company that wants to place large mirrors in orbit to provide sunlight on demand. Even when aimed carefully, those mirrors could scatter light through the atmosphere, disrupt wildlife and further erase the natural darkness of the night sky.

The article does not argue that satellites are inherently bad. Communications networks, scientific instruments and other spacecraft can provide real value. Its target is the assumption that orbit is an unlimited dumping ground for private ambition. A proposal that appears efficient from the perspective of a data-center operator can impose costs on the atmosphere, the orbital environment, astronomy and everyone who values an unspoiled night sky.

Plait closes by treating the heavens as a shared natural resource. Humanity would not accept an industrial project that destroyed a national park without serious scrutiny merely because it promised cheaper computing. The night sky deserves the same care. Before megaconstellations become irreversible facts overhead, their environmental and cultural costs need to be counted as seriously as their commercial benefits.