GEO vs. LEO vs. Direct-to-Device: Satellite IoT Options Compared

“Satellite IoT” isn’t one thing. The term gets slapped on everything from a battery-powered tracker sending a location update every few hours to a smartphone messaging home from a satellite, and the industry’s marketing hasn’t done buyers any favors by treating those as the same technology.

A shipping container tracker sending a 20-byte location update and a smartphone connecting to a satellite for broadband service don’t solve the same connectivity problem. Treating them as interchangeable can lead to the wrong hardware choice, an unrealistic power budget, or a connectivity strategy that doesn’t scale beyond the pilot.

So instead of asking which satellite technology is “best,” start with a more useful question: What does your device actually need to send, how often does it need to send it, and where does it need to work?

In this blog, we’ll break down GEO, LEO, and Direct-to-Device satellite IoT options: the comparison table, where NB-IoT over NTN actually fits, the D2D IoT landscape in 2026, and a decision guide for matching the option to the deployment. (If you haven’t yet, start with our primer on what NTN actually is.)

In short: GEO, LEO, and D2D are not interchangeable, and they’re not competitors either. In most real deployments, they end up as different layers of the same connectivity architecture.

GEO vs. LEO vs. Direct-to-Device: The Quick Comparison

What GEO Satellites Are Best For

GEO (geostationary orbit) sits at roughly 35,786 kilometers, spinning at the same rate as the Earth so it appears fixed over one spot in the sky. That means a small number of satellites (as few as three) can cover most of the planet. The tradeoff is distance: signals travel much farther, latency runs ~480–600 ms round-trip, and the link budget is less forgiving.

For many IoT applications, none of that matters. A tracker sending a location update doesn’t need a millisecond response. Neither does a pipeline sensor sending an alert or an industrial monitor reporting a periodic measurement. What actually matters for those use cases:

• Coverage
• Power efficiency
• Device compatibility
• Payload size
• Battery life
• The ability to communicate outside terrestrial cellular coverage

That’s why GEO is the practical choice for asset tracking, telemetry, monitoring, and alerting, particularly when devices need to operate across large or unpredictable geographic areas. It’s also, not coincidentally, where the commercially live NB-IoT NTN market actually is right now (more on that below).

What LEO Satellites Change for IoT

LEO (low Earth orbit) operates much closer to Earth (roughly 500–1,200 kilometers up), which cuts latency dramatically, down to ~20–40 ms round-trip. Close enough to terrestrial cellular that it barely registers as “satellite.”

The tradeoff flips to coverage. A GEO satellite holds its position relative to Earth; a LEO satellite moves across the sky and hands off to the next one within minutes. Global LEO coverage takes a constellation of hundreds to thousands of satellites, not a handful.

LEO has real potential for IoT use cases where lower latency matters more. But orbit alone doesn’t answer the deployment question; you still need to ask:

• Is the service available where the device operates?
• Does the device module support the required NTN technology?
• What spectrum and radio technology are being used?
• What is the expected power profile?
• Does the application need continuous connectivity or intermittent communication?
• How does the network behave when terrestrial coverage becomes available again?

LEO is not automatically “better” than GEO for IoT; it solves a different set of tradeoffs. A tracker sending a few updates a day gets little practical benefit from lower latency. A more time-sensitive application (oilfield monitoring, for instance) is a different story.

Direct-to-Device Isn’t the Same as Satellite IoT

This is where much of the current satellite conversation gets confusing. Direct-to-Device (D2D), sometimes called Direct-to-Cell, is not a separate orbital category. It’s a connectivity model, an unmodified device talking straight to a satellite. Most of today’s D2D players (Starlink Direct to Cell, AST SpaceMobile) fly in LEO, so they inherit LEO’s low latency, but they are built around a completely different device class: smartphones, not sensors.

D2D IoT headlines get written, but the substance underneath is almost always about extending mobile connectivity to phones: messaging, emergency communications, voice, and eventually broader mobile data. Satellite IoT has a different job: an asset tracker, environmental sensor, or utility monitor needs to run for months or years on a constrained battery, sending a few bytes at a time. It needs:

• Efficient communication
• Compatible hardware
• Predictable power consumption
• Coverage beyond terrestrial networks
• A service model designed for small and infrequent payloads

That’s not just a device-class distinction; the throughput math backs it up. AST’s own materials cite peak speeds of “over 150 Mbps per coverage cell,” but that’s shared cell capacity split across every device in range, not a per-device number, and it assumes spectrum most markets won’t clear at that scale (AST’s own operator low-band allocations run in 10 MHz-class channels). Starlink’s real-world T-Satellite speeds today are closer to 2–4 Mbps. Divide either figure across a realistic concurrent user population, and what actually reaches one device lands in Cat-1/Cat-1bis territory, not broadband.

That reframes the argument. D2D isn’t the wrong fit for satellite IoT because it’s “too advanced” or built for some abstractly different device class. It’s that once real cell capacity gets divided across real users, the throughput a single device actually sees is roughly what Cat-1bis already delivers today, terrestrially, everywhere, on cheaper hardware and a far better power budget. Paying for a satellite radio to land you back at a service tier you can already buy over LTE isn’t a good trade for most sensor deployments.

A D2D announcement is not automatically a new connectivity option for your fleet of sensors. Keep that distinction straight and half the confusion in this market disappears.

Where NB-IoT Over NTN Actually Fits

Here’s the distinction the marketing tends to erase: the NB-IoT NTN service that’s commercially live today runs mostly over GEO, in L-band, through Skylo, with LEO IoT-NTN services from Iridium, OQ Technology, and Satelliot now coming up alongside it. It is not the same technology, spectrum, or hardware as Starlink Direct-to-Cell or AST SpaceMobile.

That’s not a knock on GEO. For asset tracking, telemetry, and alerts, the ~500 ms round-trip is irrelevant; nobody’s location ping needs to feel real-time. What matters for that use case is coverage (near-global reach with a handful of satellites) and power efficiency (the same modem and battery budget as terrestrial NB-IoT). LEO-based NB-IoT/LTE-M NTN is following behind as more operators light up constellations built for the standard, and Release 19’s regenerative payloads and native store-and-forward are what will push LEO IoT-NTN toward mainstream.

The practical takeaway: if a module or provider says “NTN-ready,” ask which orbit, which band, and which operator certification. “NTN” is not a single technology. Neither is “satellite IoT.”

The Band Question: What Actually Determines Which Module You Can Use

Orbit gets the attention, but spectrum is what decides whether a given module can actually talk to a given network, and it’s the detail most “NTN-ready” claims quietly skip.

Standardised IoT-NTN lives in mobile-satellite spectrum. 3GPP Release 17 defined two bands for it: n255 in L-band (1525–1559 MHz down, 1626.5–1660.5 MHz up) and n256 in S-band (2170–2200 MHz down, 1980–2010 MHz up). L-band propagates slightly better; S-band has the cleaner path to global harmonisation, because L-band has to be coordinated with incumbent MSS operators. Later releases keep widening this: Release 18 extended the band set, and Release 19 work has pushed further still, Keysight and Samsung demonstrated live NR-NTN in S-band n252 in January 2026.

Direct-to-Device works completely differently. Starlink and AST don’t use satellite spectrum for D2D at all — they borrow terrestrial spectrum from their mobile operator partners, which is exactly why an unmodified phone can connect without new hardware. Starlink’s service runs in T-Mobile’s PCS holdings; AST uses partner low-band and has separately secured access to up to 45 MHz of lower mid-band, alongside its own S-band and L-band rights.

That distinction has a very practical consequence: the band determines the module, and the module determines where your device can work. A tracker certified for L-band n255 over GEO is not going to attach to a LEO constellation running S-band, and neither will it attach to a D2D service operating in an operator’s terrestrial band. “NTN-ready” on a datasheet means nothing on its own. The questions that matter are which bands the module supports, which operators have certified it, and whether your connectivity provider can give you commercial access to those networks, because radio compatibility and a commercial agreement are two different problems, and you need both.

The Direct-to-Device Landscape in 2026

Direct-to-Device connectivity is no longer theoretical, but it’s earlier and messier than the headlines suggest, and it’s still primarily a phone story, not an IoT one.

There’s a real difference between a service being demonstrated, available for emergency SOS, available for messaging, commercially launched in select markets, available for voice, and capable of broadband data at scale. Those are very different stages, and both major players are currently spread across several of them at once.

Starlink Direct to Cell, partnered with T-Mobile in the US and a growing list of international carriers, has the satellite-count lead and is live today for basic texting and data. Voice is in testing through 2026. Full broadband data comparable to normal cellular service depends on the Starlink V3 satellite generation, targeted for the second half of 2027.

AST SpaceMobile, backed by AT&T, Verizon, and Vodafone, is betting on fewer, much larger satellites with bigger antennas for stronger per-satellite capacity. 2026 hasn’t gone to plan. On April 19, New Glenn’s upper stage underperformed and left BlueBird 7 in an orbit too low to operate. On May 28, a separate New Glenn exploded during a static fire at LC-36, destroying the vehicle and damaging Blue Origin’s only operational New Glenn pad. AST moved its BlueBird launches to SpaceX Falcon 9 and now guides to commercial service in the first half of 2027, with a partial-coverage beta planned ahead of that. In mid-August, the FCC granted a 30-day authorization for non-commercial testing on up to 100 handsets, which expired September 12. That was a real step, but a test, not a launch.

For IoT buyers, the bigger point is simpler: D2D’s growth doesn’t eliminate the need for purpose-built satellite IoT connectivity. A network designed to message a smartphone has a fundamentally different requirements list than one designed to support thousands of battery-constrained sensors sending small, infrequent payloads. As D2D matures and expands the overall connectivity landscape, it should be evaluated as another option, not treated as a replacement for NB-IoT, LTE-M, or other NTN technologies.

Decision Guide: Which Satellite Connectivity Option Should You Choose

• Tracking, telemetry, and alerts from remote or moving assets (shipping containers, heavy equipment, livestock, pipelines): NB-IoT NTN over GEO, today. Latency doesn’t matter for a location ping; coverage and battery life do.
• Deployments where a handful of satellites need to cover a huge, unpredictable footprint (global supply chain, maritime): GEO NB-IoT NTN’s near-global reach from ~3 satellites is the practical fit right now. The exception is far-northern routes: GEO coverage drops out above roughly 70° latitude, so Arctic shipping lanes and polar routes need LEO coverage for those legs.
• Latency-sensitive IoT applications, once LEO NB-IoT/LTE-M NTN coverage matures: worth planning for, not yet a universal option, check constellation coverage for the specific region before committing.
• Consumer-facing SOS, texting, or basic connectivity for phones or phone-adjacent hardware: D2D/Direct-to-Cell, understanding that broadband voice/video/data is a 2027 story, not a 2026 one.
• Anything that needs live video, continuous streaming, or firmware pushes over satellite: none of the above, yet. Narrowband NTN wasn’t built for it, and broadband D2D isn’t there yet either.

The Bigger Decision Isn’t GEO vs. LEO

For most deployments, this won’t be a one-time, permanent choice. A device may use terrestrial cellular where coverage is available, GEO satellite where it isn’t, and LEO NTN as those networks mature, and the requirement can change while the device is already in the field:

• A vehicle moves outside cellular coverage.
• A shipping container crosses into a new region.
• A device shifts from routine monitoring into an exception state.
• A deployment expands into areas where the original connectivity strategy never accounted for.

That’s not a “pick the network with the biggest coverage map” problem. It’s an orchestration problem, building a connectivity architecture that uses the right network for the situation without forcing a redesign every time coverage changes.

The future of satellite IoT isn’t GEO replacing LEO, or D2D replacing satellite IoT. It’s more options, and the real advantage goes to whoever can use them together without re-architecting every time the map changes. We dig into exactly how that orchestration works in the next piece in this series, on hybrid cellular + satellite connectivity.

The takeaway? Don’t ask “GEO or LEO or D2D?” as if it’s one choice. Start with the deployment. Ask what the device needs to send, how often, how latency-sensitive it is, what the power budget allows, and where it actually needs to work — then match the technology to the workload, not the other way around.

Frequently Asked Questions

Is D2D the same as satellite IoT?
No. D2D (Direct-to-Device/Direct-to-Cell) is built for unmodified smartphones, borrowing mobile operators’ terrestrial spectrum. It’s marketed as broadband, but once shared cell capacity is divided across users, a single device sees roughly Cat-1-class throughput. Satellite IoT, specifically NB-IoT/LTE-M over NTN, is a narrowband service built for small, infrequent sensor and tracker payloads, in dedicated satellite spectrum, on different hardware.

Is NB-IoT over NTN GEO or LEO?
Today, commercially live NB-IoT NTN service runs mostly over GEO, in L-band, through Skylo. Iridium, OQ Technology and Satelliot are building out LEO IoT-NTN services alongside it, so “NTN” on its own tells you nothing about orbit. LEO-based NB-IoT/LTE-M NTN is emerging as more constellations are built out for the standard.

Which is faster, GEO or LEO?
LEO has dramatically lower latency (~20–40 ms round-trip vs. GEO’s ~480–600 ms), because the satellite is thousands of kilometers closer. For most IoT telemetry use cases, the difference doesn’t matter; for anything closer to real-time, it does.

Are Starlink Direct to Cell and AST SpaceMobile available now?
Partially. Starlink Direct to Cell is live for basic texting and data, with voice in testing, but full broadband data depends on the V3 satellite generation, targeted for the second half of 2027. AST SpaceMobile ran FCC-authorized non-commercial handset testing from mid-August until that authorization expired on September 12, 2026, and guides to commercial service in the first half of 2027 after launch-vehicle setbacks.

Which option should I choose for a global asset-tracking deployment?
For most tracking, telemetry, and alert use cases today, GEO-based NB-IoT NTN is the practical choice: near-global coverage from a handful of satellites, and the same power and battery profile as terrestrial NB-IoT. D2D broadband isn’t built for this use case at all. The exception is assets that operate above roughly 70° latitude, such as Arctic shipping routes, where GEO coverage drops out and LEO is needed for those segments.