Starlink's Direct-to-Cell

Authors:

Jorge García-Cabeza, WePlan Analytics & Universidad Politécnica de Madrid
Javier Albert-Smet, Universidad Politécnica de Madrid
Zoraida Frias, Universidad Politécnica de Madrid (corresponding author)
Luis Mendo, Universidad Politécnica de Madrid
Santiago Andrés Azcoitia, Universidad Politécnica de Madrid
Eduardo Yraola, WePlan Analytics

Based on the article “Direct-to-Cell: A First Look into Starlink’s Direct Satellite-to-Device Radio Access Network through Crowdsourced Measurements”.

Introduction

After years of anticipation and foundational technical development, mobile connectivity is finally reaching beyond-terrestrial infrastructure. With the launch of Direct Satellite-to-Device (DS2D) technologies, regular smartphones can now connect directly to satellites, with no specific hardware required. At the forefront of this transformation is Starlink, the satellite communications branch of SpaceX, which in late 2024 began beta testing the world’s first large-scale DS2D service in partnership with T-Mobile under the Supplemental Coverage from Space (SCS) framework, authorized by the Federal Communications Commission (FCC) in the U.S.

In a recent study conducted by the Universidad Politécnica de Madrid, Spain (UPM) and WePlan Analytics, now available on the preprint repository arXiv, we present the first empirical analysis of a real-world DS2D network, using large-scale crowdsourced data to evaluate what this space-based Radio Access Network (RAN) looks like and how it may perform in the future.

What Is DS2D, and Why Now?

DS2D is a new communications paradigm in which unmodified smartphones connect directly to Low Earth orbit (LEO) satellites acting as mobile base stations. Enabled by recent advances in beamforming, signal processing, and embedded electronics, DS2D makes it possible to fill in the gaps of terrestrial networks—particularly in remote or underserved regions—without requiring specialized user equipment.

Until now, much of the research in this space has focused on architectural models or simulation tools. What has been missing is concrete, measurement-based insight into how DS2D behaves in the real world. Our study aims to fill this gap.

Measuring the Starlink/T-Mobile beta testing

Starlink has become the global leader in DS2D communications, supported by a constellation of over 500 satellites. In partnership with T-Mobile, it has been the first to launch large-scale beta testing in early 2025 of a SCS service using T-Mobile’s spectrum. Initially limited to short messages (SMS), it is planned to support voice and data services by mid-2025. Using mobile network measurements passively collected by Android phones from October 2024 to April 2025, we tracked the evolution of the Supplemental Coverage from Space service. The data reveal several key findings:

• Crowdsourced data reflects DS2D network expansion. A steady increase in the number of cell identifiers observed aligns with the timeline of Direct-to-Cell satellite launches. This suggests that crowdsourced data can effectively track DS2D network growth and coverage, even in sparsely populated regions, challenging the notion that such data is biased toward urban areas.
• Geographically Targeted Activation. Although satellite coverage is inherently global, Starlink’s SCS service was selectively activated in specific regions ahead of the official beta launch, for example, in response to hurricanes Helene and Milton in fall 2024, and the wildfires in Southern California in January 2025.
• Direct-to-Cell is most actively used in accessible but underserved areas, especially near national parks. While designed for low-density regions, DS2D usage clusters around reachable yet poorly covered zones such as Mendocino, Redwood, and Rocky Mountain parks. A county-level analysis shows a moderately strong negative correlation (–0.59) between DS2D usage and population density. Counties like Knott (KY), Florence (WI), Hancock (TN), and Martin (KY) show usage intensities above 85%.
• Stable RAN performance across deployment and regulatory milestones. Despite the increasing number of satellites and key milestones —such as the FCC’s approval of a +10 dB increase in out-of-band emissions— key RAN indicators (RSRP, RSRQ, SINR) remained steady throughout the beta phase. Continued monitoring will help assess the impact of these changes over time.
• Lower Signal Strength, Higher Signal Quality than terrestrial networks. Compared to terrestrial networks, DS2D technologies offer significantly lower RSRP (signal strength) but higher RSRQ (signal quality). This reflects the long-distance, low-interference nature of satellite links and the unloaded state of the network (as it is SMS-only). However, these metrics are likely to shift once the service begins supporting higher-traffic applications like voice and data.
• Complementary Service. With an average throughput per beam currently estimated at around 4 Mbps, DS2D is not positioned to replace terrestrial networks in the U.S. Instead, it plays a relevant role in filling coverage gaps.

A Glimpse into the Future

The FCC has already approved an increase in allowable radiated power for DS2D transmissions, and Starlink is actively exploring additional spectrum options. Following the FCC’s recent authorization to raise out-of-band emissions by +10 dB, Starlink has indicated that this could lead to SINR improvements of around 3 dB. Based on our calculations, this gain could increase expected throughput to approximately 6 Mbps per beam under current network conditions. Furthermore, if Starlink succeeds in doubling the available spectrum for its DS2D operations, data rates could rise to around 12 Mbps per beam. While still below the average speeds of terrestrial networks, these improvements represent a significant step toward delivering functional mobile broadband to remote and underserved regions.

In parallel, expanding the satellite constellation would further enhance DS2D network performance. A greater number of satellites allows for increased spatial reuse of spectrum and more simultaneous connections, thereby boosting both overall network throughput and the per-user capacity—although the exact gains are difficult to quantify at this stage. While the FCC denied Starlink’s request to deploy up to 30,000 DS2D-capable satellites, the company remains authorized to operate up to 7,500—well above the approximately 600 currently in service for DS2D. This highlights the substantial remaining potential for densification, which could be crucial for supporting higher traffic volumes and advanced service offerings in the near future.

Why This Matters

Our study demonstrates that DS2D is not merely a theoretical concept—it is already being deployed and used in the real world. While technical, regulatory, and operational challenges remain, we offer the first empirical look into an active DS2D network, providing valuable insights into what can be expected as these services transition into broader commercial deployment. The findings underscore the potential of crowdsourced mobile measurements as a scalable and effective tool for evaluating DS2D performance, even in remote and underserved regions.

This research arrives at a crucial moment. With Starlink planning to expand its DS2D offerings beyond SMS to include voice and data services after July 2025, timely measurement-based analysis is essential. In light of the rapid evolution and intensifying competition within the satellite connectivity ecosystem, empirical insights like these are vital for guiding both industry strategies and policy decisions as the DS2D landscape takes shape.

A preprint version of the full study, “Direct-to-Cell: A First Look into Starlink’s Direct Satellite-to-Device Radio Access Network through Crowdsourced Measurements,” is available here.