Starlink Direct-to-Cell Coverage Limitations: What's Approved vs Untested

A satellite passing overhead has no opinion about a stairwell, a parking garage, or a windowless office. A phone's antenna does. That gap, between where a satellite can technically reach and where a phone can actually use it, sits at the center of Starlink direct-to-cell coverage limitations. Regulatory approvals for the service have piled up steadily over the past two years, but the public record says almost nothing about whether the signal holds up where people actually spend most of their time: indoors.

The FCC's Space Bureau authorized SpaceX and T-Mobile to deploy supplemental coverage from space on November 26, 2024, permitting direct-to-cell operations using SpaceX's previously authorized Gen2 Starlink satellites, according to the Commission's later waiver order (FCC). Through T-Mobile, the resulting service supports texting, limited app data, and low-bandwidth voice and video (Wireless Infrastructure Association). And when the FCC granted SpaceX a conditional waiver to raise its interference limits in March 2025, T-Mobile's own throughput model in that record was built specifically around an outdoor use case; AT&T's competing model, based on a market-area scenario, didn't specify indoor or outdoor conditions at all (FCC).

"Coverage," "service availability," and "reliable indoor connectivity" get treated as interchangeable in most conversations about satellite-to-phone service. They aren't. This piece separates them and traces exactly what the regulatory record establishes about each, showing what has been tested, what hasn't, and why that gap matters to anyone deciding how much to rely on Starlink mobile service indoors.

What Starlink's direct-to-cell service actually does today

The supplemental coverage from space model is a partnership, not a replacement network. Think of it as a satellite operator borrowing a carrier's phone number, so to speak: SpaceX extends T-Mobile's existing coverage rather than building an independent competing service, per the framework the FCC laid out in its waiver order (FCC).

Current functionality is narrow by design. Texting works. App data is limited. Voice and video calls run at low bandwidth. That's the current shape of Starlink's satellite-to-cell service, established through T-Mobile on modern smartphones (WIA).

A jump to something resembling 4G LTE performance is farther off, and the timeline for it comes with real hedging built in. In its October 2025 paper, the Wireless Infrastructure Association said Starlink is acquiring additional spectrum that could eventually support LTE-level speeds, but only once a second-generation constellation is built, compatible handsets exist, and the spectrum sale itself clears regulatory approval. At the time WIA wrote that assessment, it placed that capability at least two years out (WIA).

That baseline already separates two ideas that get treated as one. A satellite being technically overhead is not the same as a specific feature working under specific conditions, and better Starlink cellular connectivity down the road doesn't retroactively tell readers anything about how today's service behaves indoors right now.

What the FCC's interference waiver actually measured

The March 2025 waiver let SpaceX exceed the standard -120 dBW/m²/MHz out-of-band emissions limit, raising it to as much as -110.6 dBW/m²/MHz, but only within two narrow 5 MHz segments adjacent to its proposed satellite-to-cell downlinks (FCC). The grant was conditional: SpaceX has to resolve any harmful interference to terrestrial carriers or stop operating under the relaxed limits.

Three carriers filed competing technical analyses that differ in method as much as in conclusion. T-Mobile modeled network-throughput degradation for an outdoor use case and found effects in the range of 0.2% to 1.8%. Verizon argued from handset noise-floor math, noting that flagship phones have a noise floor around -108 dBm/MHz and that SpaceX's proposal would push the interference-to-noise ratio to -3 dB, a level it called insufficient to protect terrestrial networks. AT&T built a market-area scenario using a PCS C Block region and projected an 18% reduction in average network throughput (FCC).

All three analyses answer the same underlying question from different angles: how much does SpaceX's satellite emission degrade someone else's terrestrial network? None of them asks how well a phone inside a structure receives the satellite signal in the first place, and that's a meaningful distinction because interference modeling and receive-link testing measure entirely different things. Interference modeling looks at aggregate signal levels across a network footprint, which is what determines whether a downlink is quietly degrading a neighboring carrier's spectrum. A receive-link test would need to isolate one handset, one location, and one satellite pass, then ask whether that specific link holds together long enough to send a text or complete a low-bandwidth call. The waiver record is full of the first kind of measurement and empty of the second.

This is the most detailed technical fight in the public record, and it's entirely about protecting terrestrial networks from satellite interference. That's a different question from whether a Starlink-served phone can get a usable signal through a wall, and the two shouldn't be confused just because they share a docket.

Starlink direct-to-cell building penetration: the evidence gap in the public record

Neither the March 2025 waiver order nor the October 2025 WIA paper reports building-penetration measurements or indoor connection-success rates, and nothing else examined for this piece fills that gap either (FCC; WIA).

A satellite passing overhead is also not synonymous with a phone being able to use it. Three separate gates have to open before a connection happens: the satellite needs an active beam covering that exact patch of ground at that moment, the handset and account need to be recognized as eligible for the service, and the radio path between satellite and antenna has to be clear enough to actually establish and hold a link. Regulatory filings document the first two gates in general terms, through coverage areas and eligibility rules. Almost nothing in the cited record addresses the third gate once a building wall gets introduced into that radio path.

The physics behind that third gate aren't exotic or in dispute. Wireless engineering generally recognizes that materials like reinforced concrete, steel framing, and low-emissivity coated glass attenuate and scatter radio signals to different degrees, and that dense urban surroundings introduce multipath effects that can intermittently strengthen or kill a link. None of that is unique to satellite-to-phone service, and none of it has been measured for this particular service anywhere reviewed here. The absence is a gap in the record, not a verdict on performance. Building penetration is a legitimate open question shaped by well-documented physics; no source cited in this piece has quantified it for Starlink's direct-to-cell service in either direction, which means confident claims about indoor reliability, positive or negative, are currently unsupported by the public record.

A framework for consumers, not a verdict

Rather than assuming the service is "outdoor-oriented" or "emergency-only," claims the cited record doesn't actually support, readers evaluating this service should separate three distinct questions. First: is a satellite currently serving this location at all? Second: is the specific feature being attempted, whether texting, limited data, or low-bandwidth voice and video, one the service actually supports? Third: will the connection hold reliably at this exact indoor location, right now?

The public record answers the first two questions in general terms. Supplemental coverage from space is live, and its feature set is defined (WIA). The third question is where satellite-to-phone signal limitations actually live, and it's the one nobody in the public record has measured.

SpaceX's own filing language is aimed at gap-filling and emergency scenarios. The company told the FCC that a higher emissions threshold "will protect adjacent band networks from harmful interference while ensuring that consumers and first responders can use an increasingly strong set of features even in the most challenging circumstances" (FCC). That's a stated goal, not a measured indoor performance result, and treating it as a guarantee misreads what the filing actually claims.

There's a practical limit to how much any single experience proves, too. A successful text sent from inside a building shows that an indoor connection is possible under those specific conditions. It doesn't show that the connection is reliable across different buildings, floors, or weather. Read anecdotal reports and marketing claims with that distinction in mind, not as evidence either way about the broader question.

Conclusion

Starlink's direct-to-cell service has cleared genuine regulatory milestones. The November 2024 supplemental coverage authorization and the March 2025 conditional emissions waiver establish it as an authorized supplemental service, not a hypothetical one (FCC). Today's defined feature set, texting, limited data, low-bandwidth voice and video, is a matter of public record (WIA).

How reliably that feature set performs inside buildings isn't on record anywhere reviewed for this piece. The next meaningful milestone for this story isn't another FCC filing. It's the first independently verified test of connection success inside the buildings where people actually use their phones, and that test needs to publish specifics to mean anything: which handset models were used, which frequency bands were active, what building materials and floor levels were tested, how much sky visibility each location had, how large the sample size was, and what the resulting message-delivery rates and time-to-connect figures actually were.

Until a test like that exists and gets published, two claims need to stay separate: "the record hasn't measured indoor performance" and "the service doesn't work indoors." Only the first one is currently supported.