Starlink: useful connectivity, shared costs

Starlink can be valuable where a usable terrestrial connection is missing. That benefit does not erase measured interference with astronomy or the work of keeping crowded orbits safe. The useful question is which services justify which impacts, under conditions that outsiders can verify.

This is an analysis of evidence, not a hardware review or a live coverage guide. The dated publications below span 2022–2026; their dates matter. An old measurement cannot describe every satellite now flying, and a regulatory authorization is not a count of operational spacecraft.

Where does Starlink add the most value?

A household without dependable broadband has a different choice from one comparing satellite service with working fiber. In the first case, an additional connection may enable ordinary tasks that were difficult before: schoolwork, a video appointment or sending business files. In the second, the relevant test is whether it improves reliability, cost or service enough to justify switching. Those are decision criteria, not claims that every installation delivers those outcomes.

In its service specifications, Starlink describes two-way internet using a terminal, router and power supply, and says the service supports video calls and other household uses. These are company statements. The same document says speeds and uninterrupted service are not guaranteed; performance depends on location, time of day and the priority assigned to a customer's plan.

A satellite connection may also provide a path separate from a damaged local cable. But an emergency plan still needs power, a usable installation and a service that works at that location. Buying a second connection from the same operator is not equivalent to having two independent networks. A resilience claim should name the failure it is meant to survive.

What should a customer check before relying on it?

Start with the site, not the advertised peak download rate. Starlink itself lists dish field of view, weather, power reliability and installation conditions among performance factors. A plan suitable for occasional browsing is not automatically suitable for a business that cannot tolerate interruptions. Upload performance and delays during busy periods can matter more than a striking speed-test result.

Ask for the terms that apply to the actual address and intended use. Check the complete equipment and subscription cost, mounting work, electricity needs, cancellation conditions and treatment of traffic when the network is busy. This article does not quote a universal price or infer permission to operate from a satellite passing overhead. Local authorization and commercial availability are separate questions.

For a backup connection, decide how long power must last, which devices need access and how switching will work. Test the arrangement before an outage, if you acquire it. Get2Mars has not conducted such a test; this is a suggested acceptance check, not a report of firsthand experience.

What has optical astronomy actually measured?

The clearest distinction is between a streak crossing an image and the whole image becoming unusable. In their 2022 Zwicky Transient Facility study, Przemek Mróz and colleagues identified 5,301 Starlink streaks in observations collected from November 2019 to September 2021. The share of twilight images with streaks rose from less than 0.5% in late 2019 to 18% in August 2021.

The authors also said ZTF's science operations were not yet strongly affected during the period studied. Both findings belong in the discussion. Eighteen percent streaked does not mean eighteen percent of all astronomy was lost. Nor does a manageable impact on one survey demonstrate that every instrument or observing programme is protected.

The study estimated that essentially all ZTF twilight images could contain streaks if the constellation reached 10,000 satellites. That was a projection under the study's assumptions, not an observation of a 10,000-satellite fleet. Visors reduced measured brightness by a factor of 4.6 ± 0.1 relative to the original design in the study's optical bands. It is evidence that mitigation can help, not a measurement of today's entire fleet.

Is radio interference the same problem?

No. Reflected sunlight and radio emissions require different remedies. ASTRON's report of the Bassa team’s 2024 research describes unintended radio emissions detected with LOFAR in two one-hour sessions on 19 July 2024. In those observations, V2-mini satellites emitted up to 32 times brighter unintended radio waves than the earlier generation.

“Up to” is not an average across every satellite and frequency. These detections are measured effects; the resulting threat to future research depends on the instruments, bands, satellite designs and mitigation. Unintended emissions from equipment also differ from the deliberate radio links carrying subscribers' data. Turning a communications beam away is not proof that all unwanted emissions disappear.

A 2025 study by Grigg and colleagues adds evidence from EDA2, a prototype station at the SKA-Low site in Western Australia, using observations from June–October 2024. It detected unintended broadband emissions, including in bands allocated to radio astronomy. That strengthens the case for monitoring beyond one telescope. It does not measure lost science across all observatories, and its detections should not be added to the ZTF image percentage: the instruments, frequencies and denominators differ.

There is evidence for useful cooperation. Nhan and colleagues' 2024 experimental preprint, co-authored by NRAO, Green Bank Observatory and SpaceX researchers, describes tests in October 2023 and February 2024. Temporarily disabling downlinks near the Green Bank Telescope's pointing direction substantially reduced interference in the tested configuration. The authors called for follow-up observations. This was a specific demonstration, not independent certification that every observatory and frequency was safe.

Does returning satellites to Earth solve orbital risk?

Removing a satellite from a busy orbit addresses one part of the problem. It does not remove the need to avoid collisions while it is operating, or manage a spacecraft that can no longer manoeuvre. Operators share orbital space with debris and other missions; good performance by one fleet cannot guarantee the behaviour of everything it may encounter.

ESA's 2025 Space Environment Report, based on data through the end of 2024, describes increasing congestion and collision-avoidance demands. Its estimate of more than 1.2 million debris objects larger than one centimetre is a modeled global population, not a catalogue of individually tracked objects and not Starlink debris alone.

ESA also describes improved disposal compliance alongside continued debris growth. That combination matters: better practice can reduce risk without making the system sustainable. Forecasts of future fragmentation are not evidence that a particular collision has already happened. Useful scrutiny asks about failures, time spent in hazardous orbits, coordination with other operators and independently checkable disposal outcomes, rather than treating a large manoeuvre count as either proof of disaster or proof of safety.

What should regulation require in exchange for expansion?

Permission and deployment must be kept separate. The FCC order released on 9 January 2026 authorized another 7,500 Gen2 satellites, bringing the Gen2 total authorized in that decision to 15,000, with conditions and orbital changes. It did not say 15,000 were already providing service. This is a dated licensing milestone, not a claim about the fleet size or the complete licensing position today.

Our judgment is that expansion should be assessed against observable outcomes: connectivity gained where alternatives are poor, interference remaining after mitigation, reliable disposal and transparent failure reporting. Scientific cooperation is useful, but affected observatories should not have to accept a promise as a substitute for measurements. Equally, evidence of interference is not evidence that every satellite connection lacks public value.

The unresolved choice is how much residual impact society accepts, who pays to reduce it and who can require a change when safeguards fail. That is a policy decision informed by science, not a vote on whether the measurements are real.

Which questions would change your assessment?

• For a customer: what usable alternative exists at your location, and what happens during a power or operator outage?

• For an operator: which independent observations show that a mitigation works across relevant hardware generations and frequencies?

• For a regulator: what measured failure or interference threshold triggers corrective action rather than another promise?

• For an observatory: what fraction of data or observing time is lost, and who bears the mitigation cost?

• For readers: what evidence would make you revise your view? Comments can compare priorities; they cannot establish scientific truth by majority.

For related context, our space technology overview separates engineering requirements from aspirations, our Starship guide examines the launch system, and our Mars mission record keeps actual mission outcomes distinct from plans. These are background links, not evidence that an Earth-orbit broadband network proves readiness for Mars.

Get2Mars analysis. Drafting and translation were AI-assisted; source comparison and language consistency checks were performed by an AI assistant. No human expert or native-speaker review is claimed. We did not test the service, calculate a current fleet count or audit national coverage. The linked records support the dated claims above; they do not settle every later design change or regulatory decision.