In Bethel, Alaska, a town of 6,300 people reachable only by air or river, downloading a medical imaging file used to take the better part of an afternoon. Today, a Starlink dish bolted to the roof of the regional health clinic pulls it down in seconds. Scenes like this are playing out across rural America, on cargo ships in the South Pacific, and at Antarctic research stations, all connected by a constellation of more than 10,000 low-Earth orbit satellites that SpaceX has quietly turned into the world’s largest broadband network.
As of June 2026, independent tracking by astronomer Jonathan McDowell counts upward of 10,020 Starlink satellites launched, with the majority actively serving customers. SpaceX’s residential service tier advertises typical download speeds of 50 to 250 Mbps, depending on plan level and location, according to the company’s specifications page. Those numbers are impressive on paper. Two recent independent studies now offer the first rigorous look at whether the network delivers on them in practice.
Two studies, two angles on the same constellation
The strongest independent evidence comes from a pair of research papers posted as preprints on arXiv, each attacking the question from a different direction.
The first is a large-scale IPv6-based measurement study that maps how data actually travels through the Starlink constellation. Rather than running a speed test from a single dish, the researchers used Internet-wide measurement techniques to trace packet routes across the network’s global footprint. By analyzing IPv6 addressing, they could see how signals hop between satellites, ground stations, and user terminals. The result is a detailed technical portrait of Starlink’s routing architecture: how it manages traffic, where bottlenecks form, and how latency behaves across different paths.
The second is a separate measurement study focused on Starlink’s direct-to-cell service, the newer capability that connects standard smartphones to satellites without a dish. Using crowdsourced mobile network data collected across the United States from October 2024 through April 2025, the researchers captured real-world performance as ordinary phone users experienced it. This service, authorized by the FCC as “supplemental coverage from space” and operated in partnership with T-Mobile, represents a fundamentally different product from the dish-based broadband. It uses different radio technology, different frequency bands, and different hardware, so its performance profile should not be confused with what a rooftop terminal delivers.
What makes both studies valuable is their methodology. Neither relies on SpaceX marketing materials or user-posted speed test screenshots. The IPv6 study probes the network’s internal behavior using documented, reproducible techniques. The direct-to-cell study draws on thousands of real measurements spanning months and geographies. Together, they form the most rigorous publicly available picture of how Starlink performs at scale.
What the data shows, and what it does not
The IPv6 study confirms that Starlink operates a functioning, globally distributed routing infrastructure capable of moving traffic efficiently through a dense satellite constellation. It documents how the network’s addressing architecture handles the complexity of thousands of moving nodes, and it characterizes latency behavior across different routing paths. For users in remote areas, the practical takeaway is that the underlying plumbing works: data can travel from a ground station to a satellite to a user terminal and back with the kind of consistency needed for video calls, telehealth, and cloud-based work.
The direct-to-cell study captures something the IPv6 work cannot: the messy reality of everyday use. Because its data spans seven months and covers locations across the continental U.S., it reflects seasonal variation, geographic diversity, and the congestion patterns that emerge when thousands of users share the same orbital resources. Its connection to FCC-authorized supplemental coverage gives the findings regulatory weight, potentially informing how carriers and agencies evaluate satellite links as part of nationwide coverage obligations.
Neither study, however, publishes a single verified global average download speed. The IPv6 research characterizes routing efficiency and latency but does not reduce its findings to a simple “users get X Mbps” figure. The direct-to-cell study examines a service tier with inherently lower throughput than dish-based broadband, so its speed findings do not apply to the residential product most subscribers use. The gap between SpaceX’s advertised ranges and independently measured, globally consistent numbers remains open.
Constellation size adds another layer of uncertainty. SpaceX launches satellites in frequent batches, and tracking services maintain near-real-time counts, but the number of fully operational satellites at any given moment differs from the total launched. Some are still raising their orbits, some have failed, and some have been intentionally deorbited. The operational count fluctuates week to week, making it difficult for planners to treat the constellation as a fixed piece of infrastructure.
Context the studies do not cover
Both preprints focus tightly on network performance. They do not address several questions that prospective users and regulators will inevitably raise.
Cost. Starlink’s residential service starts at $120 per month in the U.S., with hardware costs for the dish and router on top of that. The Priority tier, which offers higher speeds and greater data allowances, costs significantly more. For rural households already stretching budgets, price remains a barrier that no amount of routing efficiency can solve.
Competition. Amazon’s Project Kuiper has begun deploying its own LEO constellation and plans to offer commercial service. Eutelsat OneWeb operates a smaller constellation focused on enterprise and government customers. How Starlink’s performance holds up as orbital space gets more crowded is an open question neither study addresses.
Sustainability. Orbital debris risks, interference with astronomical observations, and the environmental footprint of frequent rocket launches are subjects of active regulatory debate. The FCC, the International Telecommunication Union, and space agencies worldwide are developing frameworks to manage these concerns, but the available academic work here stays in its lane: network measurement, not constellation management.
Readers should treat broad claims about Starlink’s future capacity or global expansion timeline with caution until primary data from SpaceX or regulatory bodies addresses these areas with the same rigor these studies bring to performance measurement.
How to weigh the evidence
For anyone evaluating Starlink as a primary connection in a remote area, the research supports a cautious but grounded optimism. The network’s routing infrastructure functions at global scale. Direct-to-cell links work as a practical extension of terrestrial mobile coverage, at least in the U.S. and within the study window. These are not trivial achievements for a system that did not exist a decade ago.
But real-world performance will depend on local factors: how much sky the dish can see, how close the nearest ground station is, how many other subscribers share the same beam, and which service tier is in use. The studies confirm that Starlink works at scale. They do not promise it will work identically for every household, fishing vessel, or polar research station.
Both papers are preprints, meaning they have not completed formal peer review. Their methods are documented and open to scrutiny, and subsequent research may refine or challenge their conclusions. The responsible read is to treat them as the best available independent evidence while recognizing that the full picture will require additional studies, particularly outside the United States, where ground station density, regulatory environments, and user loads differ substantially.
For regulators, the message is more pointed. As satellite constellations seek subsidies, spectrum rights, and coverage credits, empirical measurement of the kind these researchers have produced offers a benchmark that operator-supplied filings alone cannot. The FCC’s recent decisions to revoke Starlink’s Rural Digital Opportunity Fund award underscore how much rides on the gap between promised and delivered performance.
Satellite broadband is infrastructure now
The larger shift these studies document is not really about SpaceX. It is about the moment when satellite broadband stopped being an exotic backup and became primary infrastructure for millions of people who have no other viable option. In Bethel, on container ships, at McMurdo Station, the Starlink dish is not a novelty. It is the connection.
As that transition accelerates, the quality of evidence will matter as much as the quality of the signal. Independent measurement, transparent methodology, and honest accounting of what remains unknown are the tools that will separate genuine progress from promotional noise. These two studies are a strong start. They will not be the last word.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
