The pitch deck for direct-to-device connectivity always leads with a download number. AST SpaceMobile demonstrated a 98.9 Mbps peak to an unmodified smartphone over international waters; its next-generation satellites are designed to approach 200 Mbps. SpaceX quotes up to 150 Mbps per user on its forthcoming hardware. Those numbers are real, and they are also the wrong way to understand what this technology does — because every one of them describes the easy direction.

The constraint

Why is the uplink the bottleneck, not the download?

A satellite the size of a tennis court can pour an enormous amount of power and antenna gain into the downlink. It cannot do anything about the other half of the conversation. The uplink originates at a phone in someone's hand — roughly 200 milliwatts of transmit power through a tiny, non-directional antenna, trying to close a link to a spacecraft several hundred kilometers overhead. No amount of satellite engineering adds power to the handset's transmit side. That asymmetry is structural, it applies to every direct-to-device system regardless of constellation size, and it is the single most important fact about the category.

The best independent evidence we have puts numbers on it. A peer-reviewed study from the Universidad Politécnica de Madrid and WePlan Analytics, published by the IEEE and built on more than a million crowdsourced measurements of Starlink's deployed direct-to-cell network, estimates real-world performance at roughly 3 Mbps per beam in outdoor conditions — and "per beam" means shared across everyone in that beam's footprint. The authors are explicit that this sits far below terrestrial averages, which is precisely why the service is positioned as supplemental coverage rather than a replacement. The ceiling they model, around 18 Mbps, isn't an engineering inevitability either; it's contingent on regulators granting more power, more spectrum, and more satellites.

You can see the constraint in the architecture, not just the spreadsheets. Every commercial service that's actually live today runs on store-and-forward SMS — messages queued and delivered with seconds of latency — rather than the sustained, real-time, two-way stream that voice and data require. Store-and-forward is the engineering signature of a link that cannot hold a fat symmetric pipe open. The technology is telling you what it's for.

What people actually use

What does direct-to-cell do today?

Look at what has shipped, and the answer stops being a matter of opinion. Not one mainstream direct-to-device product on the market sells broadband. They all sell messaging and safety.

Apple's service — the one that started the category in 2022 — is, by its own design, a personal locator beacon built into the phone: emergency SOS, Find My, roadside assistance, and off-grid texting, explicitly not voice or internet. It has already saved lives, with multiple documented rescues of people stranded in remote areas whose iPhone delivered an SOS message and a GPS fix when nothing else could. T-Mobile's T-Satellite, running on Starlink, offers messaging on satellite-optimized phones — ten dollars a month for non-subscribers, free on premium plans — and cannot make calls or browse. Google's Pixel and Samsung's Galaxy flagships connect through Skylo for the same purpose. The honest line buried in the consumer coverage says it best: satellite on a phone is not cellular, and it does not behave like it.

This is the part the throughput debate misses. The entire shipping product category — built by companies that compete fiercely and share nothing — has independently converged on low-uplink messaging. That convergence isn't a coincidence or a phase. It's what the link budget allows.

Where the revenue is

Where is the money actually going?

Follow the revenue and you arrive at the same place. The paying volume in satellite-direct connectivity is not in consumer broadband; it's in low-rate telemetry and messaging. Counterpoint Research projects direct-to-device-supported IoT connections reaching roughly 11 million in 2026, and analysts name enterprise IoT as one of the strongest monetization pathways in the whole sector. The broader satellite-IoT market sits near $3.6 billion in 2026, growing in the low-20s percent annually — and within it, the single largest application is asset tracking and logistics at about 35 percent: a position ping measured in bytes, the lowest-uplink workload imaginable.

Even the optimists concede the near-term shape of it. Industry trackers note that standardized satellite IoT remains a small fraction of the market for now, that immediate commercial returns come from consumer mobile messaging, and that mass-scale momentum waits on pricing and revenue-share models that aren't settled yet. The growth is real. It is also growth in exactly the low-uplink workloads the physics permits — not in the broadband the marketing implies. (A note on the market figures: estimates range from roughly $2.4 billion to $4.2 billion depending on whether a given firm counts "satellite IoT" or "direct-to-satellite," so treat the absolute number as a band, not a point.)

The convergence

When the science, the engineers, and the money all agree

Here is the argument in one move. The physics says the uplink is the binding constraint. The engineers — at Apple, SpaceX, Google, Samsung — have all shipped messaging-only products. And the revenue is concentrated in asset tracking and emergency texting. Three independent witnesses, who coordinated on nothing, give the same testimony. When the measured science, the shipped products, and the booked revenue all line up against the marketing deck, the marketing deck is the outlier.

So the verdict isn't that the technology is overhyped. It's that the use case was mis-sold. Direct-to-device is a low-uplink messaging-and-telemetry layer wearing a broadband costume — and as a messaging-and-telemetry layer, it's genuinely valuable. A text where there was nothing. An SOS from a canyon. A sensor reading from a pipeline no tower will ever reach. That's the Jeep you keep not because you drive it off-road every day, but because of the one day the pavement runs out.