The phone is the terminal now

A Falcon 9 lifted off from Cape Canaveral at 3:50 in the morning Eastern time on 5 August carrying three satellites called BlueBird 11, 12 and 13. They are the first of the Block 2 design, folded into a standard five metre fairing on a stackable carbon composite frame, and they talk to ordinary phones. Not to a dish. Not to a terminal bolted to a roof. To an unmodified handset with nothing installed on it (AST SpaceMobile, 2026).

That last detail is the whole story, and it keeps getting buried under the throughput numbers. The earlier Block 1 satellites demonstrated a peak download of 98.9 Mbps straight to a standard smartphone, and the new ones are expected to roughly double that. Fine. The number that matters more is zero, as in the amount of hardware the person on the ground has to carry, power, mount, or be trained on.

Coverage stopped being a hardware problem

Satellite connectivity has been available for decades to anyone willing to buy a terminal and point it at the sky. What it was not was ambient. Somebody had to specify the kit, budget for it, ship it to site, and then find the one person on the crew who knew how to make it work in the rain.

The direct-to-device model deletes that entire chain. It is also not one company having a go. Starlink has more than 650 direct-to-cell satellites in orbit with service live across 22 countries, having gone from text only at commercial launch in July 2025 to data by that October and voice rolling out through this year (SatelliteInternet.com, 2026). Closer to home, Vodafone and AST SpaceMobile have a joint venture, SatCo, based in Luxembourg, set up to sell direct-to-device capacity to European mobile operators from this year (Vodafone, 2025).

Three separate commercial routes to the same capability usually means the capability is real and the argument is now about price.

It is a thin pipe, and that is the design brief

Here is where we would push back on how this gets sold. A peak figure measured in a trial is not what a person standing in a valley gets on a Tuesday. Early European trial results were reported above 20 Mbps over a narrow 5 MHz channel, which was described as enough for voice, texts and limited data (Via Satellite, 2025). Read that as the honest shape of it.

So the thing arriving is not broadband everywhere. It is a narrow, intermittent, latency-tolerant channel almost everywhere, with a hard ceiling and windows that open and close as satellites pass. Those are two completely different engineering briefs and only one of them is being designed for.

"

Ambient connectivity does not remove the offline mode. It changes what the offline mode is allowed to assume.

What it changes for a system in a field

We build things that get used away from a desk. A site inspection recorded on a phone. A check-in queue at an event in a car park with one bar. A property record updated by somebody standing in the room the record is about. The failure mode has always been the same: the connection goes, the app has to keep working, and the sync has to survive whatever happened in between.

Direct-to-device does not remove that work, and anyone selling it as an offline-mode killer has not deployed anything into a field. What it does is shift the numbers you design against:

• 01The gap gets shorter. Planning for eight hours of no signal turns into planning for eight minutes. That changes how much you queue locally and how aggressively you can retry.
• 02The pipe gets narrower, not wider. If your sync payload is a fat JSON blob with every field the server might want, it will not fit in the window. Send deltas, and send the important ones first.
• 03Priority becomes a real feature. A safety flag and a photo upload should not queue equally. Most sync layers we inherit treat them identically.
• 04Latency stops being a rounding error. Anything that assumes a quick round trip inside a user interaction, a validation call, a lock check, a token refresh, needs a local answer and a later reconciliation.
• 05Coverage stops being a procurement question. You no longer have to price a terminal per site to put a system somewhere remote, which quietly widens where a project is viable at all.

What we would do about it now

Nothing dramatic. This is not a rebuild, it is a set of assumptions worth writing down before the next brief lands. If you are commissioning something that will be used away from reliable signal, ask the build team three questions: what is the largest payload a single sync attempt has to move, what happens to the queue if the window closes halfway through, and which records are important enough to send first. If the answers are vague, the system was designed for an office.

The part we find genuinely interesting is the install. We have spent years specifying connectivity for installations and operational systems, and it is consistently one of the least glamorous and most expensive lines in a project. A capability that arrives in a device somebody already owns, with no fitting and no training, is the kind of change that does not look like much on the day and then quietly rewrites what you can propose a year later.