Satellite IoT in 2026 for Remote Assets: Iridium SBD vs NB-IoT NTN vs Licensed Backhaul, and What India's Rules Mean for Each
The verdict first, because the details take a while. For a remote asset outside India, design new products around standards-based NB-IoT over satellite (NB-NTN) where a certified chipset and a commercial service exist for your region, and keep Iridium Short Burst Data as the proven option for anything that ships before that ecosystem settles. For an asset inside India, the technology comparison is the second question. The first is which satellite services you are legally permitted to use, and the answer rules out the default choice that most global satellite-IoT articles recommend.
That second sentence is the reason this post exists. Satellite IoT writing is overwhelmingly written for the United States and Europe, where you pick a bearer on technical merit and sign a contract. In India, satellite mobile services are a licensed activity, foreign mobile-satellite terminals need government permission, and the regulatory picture for new low-earth-orbit services moved a long way in 2026 without yet reaching commercial IoT. A firmware team that chooses a bearer on datasheet merit and discovers the licensing position after tooling is built has made the most expensive kind of mistake.
Choose the satellite bearer in this order: where the asset sits, which services are lawful and commercially live there, then message size, latency and power. Starting with the datasheet inverts the order and wastes months.
The Candidates, Placed on One Table
Four families of satellite connectivity matter for remote monitoring in 2026. Statuses below reflect what was publicly reported through October 2026 and will change; verify with providers before committing hardware.
Option | How it works | Message model | Status (Oct 2026) | Typical use |
Iridium Short Burst Data (SBD) | Proprietary L-band service over the Iridium LEO constellation, via a dedicated modem | Small binary messages: mobile-originated up to 340 bytes, mobile-terminated up to 270 bytes | Mature, global, in service for over a decade | Tracking, remote telemetry, alarms where messages are small and infrequent |
Iridium NTN Direct | 3GPP NB-IoT over the same constellation, standardised in Release 19 | IP-style NB-IoT data, with standard cellular modules | On-air trials from January 2026; live over-the-air testing of an additional chipset reported in June 2026; certification and commercial availability expected before end of 2026 | New designs wanting standard modules and carrier roaming |
GEO NB-NTN (for example Skylo-type services) | 3GPP NB-IoT over geostationary satellites, often sold through mobile operators | NB-IoT messaging through standard modules | Operator partnerships announced in early 2026; one reported offering covers 36 countries on a single SIM | Cellular-style IoT that falls back to satellite when terrestrial coverage ends |
Licensed backhaul (VSAT, authorised GEO or LEO broadband) | A gateway at the site uplinks over a licensed satellite broadband service | Full IP, many devices behind one terminal | Available through licensed operators; new LEO broadband entrants in India still awaiting commercial launch | Sites with many sensors, power available, and bandwidth needs |
Two points on that table deserve emphasis. First, NB-NTN is the direction of travel. Using the same NB-IoT stack your cellular modules already speak means one module family, carrier roaming, and eventually a path to a single product SKU for both terrestrial and satellite coverage. Second, it is not yet an obvious default. Certified silicon, finished commercial terms, and field experience are still arriving, and any decision made on a launch announcement is a decision made on a roadmap.
Message Size, Latency, and the Firmware Consequence
The shape of the message model drives firmware design more than any other property.
SBD is a binary messaging service. Your device hands the modem a payload of up to 340 bytes, the modem waits for a satellite pass with usable geometry, and the message is delivered to a gateway that forwards it to you by email, HTTP, or a direct IP link. The operational model is a mailbox, not a socket. There is no connection to keep alive, which is why SBD devices can be asleep nearly all of the time. Latency ranges from several seconds under good sky view to minutes when obstructed, and the delivery is acknowledged at the modem, not end to end at your application.
NB-NTN uses the same radio concepts as terrestrial NB-IoT: attach to a network, hold a context, exchange packets over IP or non-IP data delivery. That means familiar sockets, familiar AT commands on a standard module, and a protocol stack your team has probably already shipped. It also means session setup, timers, and power-saving modes that behave differently when the "tower" is moving at orbital speed. Satellite NB-IoT in 3GPP Release 17 and later relies on the device knowing its own position and the satellite's ephemeris so it can pre-compensate timing and frequency offsets. In practice, a NTN device needs a recent GNSS fix before it transmits, which adds acquisition time and energy to every wake-up that a terrestrial NB-IoT node does not pay.
Property | SBD | NB-NTN (LEO) | NB-NTN (GEO) | Licensed IP backhaul |
Max app payload per message | 340 B up, 270 B down | Limited by NB-IoT packet size; hundreds of bytes typical | Same family of limits | Effectively unlimited |
Latency (typical) | Seconds to minutes | Seconds | About half a second round trip minimum, plus access delay | Roughly 600 ms round trip (GEO) down to tens of milliseconds (LEO) |
Needs GNSS fix before transmit | No; the modem handles satellite acquisition itself | Yes, for timing and frequency pre-compensation | Yes | No |
Antenna | Small patch with clear sky view | Small patch, clear sky view | Fixed-pointing patch or small dish, depends on service | Dish, pointing and power required |
Software model | AT commands, mailbox | Sockets or NIDD over a standard module | Same | Ordinary IP |
The practical consequence: design the application message format independent of the bearer. A compact, versioned binary envelope with its own authentication tag (the same discipline described in our LoRaWAN application-layer security post) travels over SBD, NB-NTN, or IP unchanged. The bearer becomes a driver, not an architecture.
Aggregate at a gateway: one satellite endpoint is cheaper, simpler to license and easier to power than forty.
What a Satellite Uplink Does to Your Power Budget
Power is the second great misunderstanding. A satellite modem transmitting to a satellite hundreds or thousands of kilometres away draws a different load from a terrestrial radio.
SBD modems are well known for their burst behaviour: transmit current peaks on the order of an ampere or more for short intervals, which is beyond what most small primary cells can source directly. The standard engineering response is a supercapacitor or capacitor bank charged slowly from the battery or solar supply, then discharged during the transmit burst. That adds board area and a charge-management design problem, and it sets a minimum time between transmissions while the bank recharges.
NB-NTN modules draw cellular-class currents (hundreds of milliamps at peak) but spend longer awake per message because of the GNSS fix and the satellite access procedure. A node that uplinks once an hour pays that cost 24 times a day.
A back-of-envelope model keeps the comparison honest. Fill in your own measured figures; the numbers below are placeholders for the structure, not vendor data.
# energy_model.py -- annual energy and message-count model for a remote node
# Replace every placeholder with measured values from your own bench tests.
def annual_energy_mah(msgs_per_day, e_msg_mah, sleep_ua, other_mah_per_day=0.0):
tx = msgs_per_day * e_msg_mah * 365
sleep = (sleep_ua / 1000.0) * 24 * 365
return tx + sleep + other_mah_per_day * 365
scenarios = {
# name: (msgs/day, energy per message in mAh INCLUDING wake, fix, tx, rx; sleep uA)
"SBD, hourly": (24, 0.35, 8), # placeholder
"NB-NTN LEO, hourly": (24, 0.60, 6), # placeholder: GNSS fix + attach cost
"NB-NTN GEO, hourly": (24, 0.80, 6), # placeholder: longer link setup
"SBD, 6 per day": (6, 0.35, 8),
}
battery_mah = 19000 # e.g. one D-cell primary lithium
derate = 0.7 # temperature, self-discharge, margin
for name, (m, e, s) in scenarios.items():
use = annual_energy_mah(m, e, s)
life_years = battery_mah * derate / use
print(f"{name:24s} {use:7.0f} mAh/yr -> {life_years:5.1f} years")
The model's value is not the output. It forces three questions that datasheets avoid: what does one complete message cost including the fix and the retries, how many messages per day does the application truly need, and which dominates, the radio or the sleep current. For hourly reporting on a large primary cell, sleep current and the cell's own self-discharge often dominate, which means the choice between SBD and NB-NTN matters less than people expect. For frequent reporting, per-message energy dominates and a measured number from your own bench is the only honest input.
Standards Versus Proprietary: What You Are Actually Buying
A proprietary service such as SBD offers maturity: a decade of field history, predictable behaviour, a modem with a stable interface. It also ties you to one operator's network, one modem family, and one set of commercial terms.
A standards-based NB-NTN service offers a different set of assets. Modules come from multiple semiconductor vendors; roaming agreements can in principle extend a single SIM across operators and between terrestrial and satellite coverage; and the engineering skills transfer from your existing cellular IoT work. The costs are the immaturity noted above and one subtlety that rarely makes the brochure: until certification completes and a service is commercially live in your region, you are developing against a trial or a roadmap.
A launch announcement is not a service level agreement. Before committing a hardware platform, ask for the commercial availability date in writing, the list of certified modules, and which countries and regulators have approved the service.
The India Question Comes Before the Technology Question
In India, providing satellite mobile services requires a licence from the Department of Telecommunications under the Global Mobile Personal Communication by Satellite (GMPCS) framework. Only licensed operators may provide such services, and foreign mobile-satellite handsets and two-way communicators generally need DoT permission to be brought in and used. Iridium equipment is a recurring subject in these rules. The Directorate General of Shipping's order dated 11 March 2026 restates that Iridium-based equipment in Indian waters is permitted only for distress and safety communications under the maritime GMDSS framework, and that unauthorised use of this class of equipment can be prosecuted under the Indian Telegraph Act and the Indian Wireless Telegraphy Act. Reporting from several legal and industry sources adds that portable Iridium sets may not be taken ashore and used on Indian land without authorisation.
What that means for an IoT product designer needs careful reading, not a quick conclusion. These rules are written mostly around handsets, personal communicators, and maritime use. Whether and how an Iridium SBD modem embedded in a fixed industrial asset can be lawfully operated on Indian land is a licensing question for DoT and for a licensed service provider, not something to settle by reading a modem datasheet. The responsible position for a product team is the following:
Treat any design that relies on an end customer importing and activating a foreign mobile-satellite terminal inside India as unproven until you have written permission or a licensed provider's service agreement in hand.
Ask any satellite connectivity provider you consider for their Indian authorisation: which licence they operate under, which Indian gateway handles the traffic, and which terminals are approved.
For assets outside Indian jurisdiction (overseas deployments, exports to customers in Africa, the Middle East, or elsewhere), the technical comparison above applies directly, subject to the licensing rules of the country where the asset sits.
The wider Indian satcom picture is also moving. As publicly reported through 2026, three operators have secured GMPCS licences (a Bharti-backed operator, a Jio-linked venture, and Starlink), the government notified rules for administrative assignment of satellite spectrum in 2026 with additional security clearances required before consumer services start, and one report described the Digital Communications Commission approving most of the regulator's spectrum recommendations in early September. None of that has yet produced a commercial satellite NB-IoT service for Indian industrial customers, and reporting also describes mandatory security conditions such as in-country gateways, lawful interception capability, terminal registration, and geo-fencing for licensed operators. For a product team, the implication is a planning horizon, not a bearer: licensed LEO broadband and satellite IoT services are likely to become options in India over the coming years, on the operators' timelines.
What Indian Deployments Do in the Meantime
Remote Indian sites that cannot reach terrestrial networks usually solve the problem with a gateway at the site and a licensed backhaul. A local LoRaWAN or wired sensor network aggregates readings at a gateway, which uplinks over a VSAT or other licensed satellite broadband service procured through an authorised operator. The satellite terminal belongs to the licensed service, and the product's own radio design stays on licence-exempt terrestrial bands (the 865 to 867 MHz band for LoRa and similar). That arrangement keeps the product's compliance story simple and puts the satellite licensing burden with a party that already holds the licence.
Where the site has partial cellular coverage, a second approach is store-and-forward: nodes buffer data and a gateway or a vehicle-mounted collector transfers it when it reaches coverage. That works for non-urgent data and costs nothing per message in airtime fees.
In India, unresolved licensing is a stop sign, not a risk to be managed later.
A Remote Tank Farm and Three Candidate Designs
A representative composite, genericised, shows how the pieces fit. A monitoring system covers an unattended storage site: level, temperature, and a pressure alarm across 40 sensors within a few hundred metres, plus a requirement that an alarm reaches a control room within two minutes. The site has power from a small solar array and no reliable cellular signal.
Design A: one satellite modem per sensor. Each of 40 nodes carries a satellite modem. Forty airtime subscriptions, forty antennas, forty supercapacitor banks, and forty independent failure points. Airtime cost scales linearly with node count. This design is only reasonable for sensors scattered over a wide area with nothing nearby to aggregate them, which is not this site.
Design B: LoRaWAN sensors, one gateway, one satellite uplink. Sensors report over low-power radio to a gateway on the solar array. The gateway batches readings hourly, applies the authentication tag described earlier, and sends one aggregated message plus any alarm immediately. One modem or terminal, one subscription, and the gateway can hold a few days of data if the satellite path is down. Where the satellite link must be Indian-licensed, the uplink is a VSAT or other licensed broadband service rather than a mobile-satellite modem.
Design C: sensors with embedded NB-NTN. Wait for certified modules and commercial service. Worth a pilot in 2027, not a basis for a product shipped this year.
Design B won, for a reason that holds well beyond this example: aggregating at a gateway turns 40 expensive satellite endpoints into one, moves the licensing question to a single well-understood terminal, and leaves the low-power sensor design on a cheap, licence-exempt radio. The alarm path adds one more rule: alarms bypass batching and go out within seconds of detection, while routine readings wait for the hourly batch. That single rule is what lets airtime stay cheap without breaking the two-minute requirement.
Reading a Provider Quote Without Getting Surprised
Satellite airtime pricing is where lifetime cost hides. Per-message prices look small on a quote and compound across a fleet and a product life. Ask every provider the same set of questions and put the answers in one table so that quotes can be compared on the same basis.
Question | Why it matters |
How is a message billed: per message, per byte block, per kilobyte, or flat bundle? | A 340-byte limit billed in small blocks rewards compact payloads; a flat bundle rewards chatty devices. |
Is there a monthly minimum per device, and does it apply when the device is idle? | A fleet of seasonal devices can pay for months of silence. |
What is charged for mobile-terminated messages (commands to the device)? | Downlink is often priced higher and is the path you need for updates and key rotation. |
What happens to a failed or unacknowledged message: is it billed? | Retries from bad sky view can inflate cost. |
Is there an activation or suspension fee, and can devices be paused? | Matters for stock held in inventory and seasonal use. |
Which country gateways handle traffic, and under which licence? | This answers the India question and any data-residency requirement. |
What are the contract term and price-change terms? | A product life of seven years outlasts most price lists. |
What is the service-level commitment for message delivery latency, if any? | Many services give coverage, not a latency guarantee. |
A quote that cannot answer the gateway and licence question for your deployment region has not finished answering the first question in this post.
Sky View and Antenna Placement
A surprising share of underperforming satellite installs fail for ordinary reasons. Satellite modems and NTN modules need a clear view of a large part of the sky, and the antenna is the part of the design that gets the least attention until installation day.
A few rules that save field visits:
Survey the obstructions at the actual mounting point, not at the base of the pole. A tree canopy, a gantry, or a steel tank can block a large wedge of the sky. A phone photo taken from the antenna location, pointed straight up, shows the opening immediately.
Keep the antenna cable short. Cable loss between modem and patch is a direct deduction from a link margin that is already thin. Mount the modem near the antenna and run the longer cable on the interface side, where loss is harmless.
Give the patch antenna its ground plane. Many patch antennas assume a specific ground plane size. Mounting one on a small plastic bracket or a thin sheet of metal changes its pattern and can cost several dB.
Mount it level and away from metal clutter. A tilted antenna is steering part of its coverage toward the ground.
Allow for GNSS on NTN designs. The GNSS antenna has the same sky-view needs, and a design that places it under the same metal lid as the satellite antenna often works on the bench and fails on site.
Test sky view systematically. Run a long-duration soak at the installed location, log every send attempt with its outcome, and judge the install on the success rate over days, not on a single successful message in the first five minutes.
A Session Log That Tells the Truth
Whatever bearer you choose, keep a compact diagnostic record per attempt. When a customer reports missing data, this log separates antenna problems from firmware bugs from service outages without a site visit.
// One record per send attempt, stored in a ring buffer and uploaded opportunistically
typedef struct __attribute__((packed)) {
uint32_t ts; // seconds since epoch (or boot, if no time yet)
uint8_t bearer; // 0=SBD, 1=NB-NTN LEO, 2=NB-NTN GEO, 3=IP backhaul
uint8_t result; // 0=ok, 1=no_signal, 2=timeout, 3=modem_error, 4=low_power
uint8_t attempts; // retries used for this message
uint8_t rssi_or_snr; // modem-reported link quality, scaled to 0..255
uint16_t t_active_s; // seconds the radio was powered for this attempt
uint16_t v_batt_mv_min; // minimum supply voltage seen during the attempt
} send_log_t; // 12 bytes
Two fields in that record pay for themselves. t_active_s gives measured energy per message once multiplied by your bench current profile, which keeps the power model honest against real field behaviour. v_batt_mv_min catches the supply-droop failures that look like radio problems, where a burst current collapses a weak cell before the transmission completes.
An acceptance rule worth adopting: for each installed site, define the minimum acceptable delivery success rate over a seven-day soak (for example, 99 percent of scheduled messages delivered within the permitted window) and treat anything below it as an installation defect, not a service outage, until the log shows otherwise.
When Satellite Is the Wrong Answer
Satellite is the connectivity of last resort, and treating it as a default wastes money.
When terrestrial coverage exists, even poor coverage. A LoRaWAN gateway with a cellular or wired backhaul, or a cellular IoT module with a rooftop antenna, costs far less per message than any satellite service.
When data can wait. Readings that are valuable but not urgent can be buffered and moved by store-and-forward when a vehicle or technician visits. If a weekly download answers the business need, do not pay per-message airtime.
When the device cannot see the sky. All of these services need a reasonably clear view of the sky. Sensors in pits, under steel roofs, inside vessels, or in dense canopy need an external antenna run or a different architecture.
When the volume is high. Per-message pricing punishes chatty devices. A design that sends 288 messages a day over satellite is an application design problem to fix before it is a connectivity cost to negotiate.
When legal permission is unclear. In India, unresolved licensing is a stop sign, not a risk to be managed later.
The Decision Rules, Written Down
If your situation is... | Then choose... | Because... |
Asset outside India, ships in 2026, small infrequent messages | SBD today, with the application envelope bearer-independent | Proven service; certified NB-NTN supply is still arriving |
Asset outside India, product life of 7 to 10 years, ships in 2027 or later | NB-NTN, with a pilot in 2027 before volume commitment | Standard modules, roaming, and one stack for terrestrial and satellite |
Asset inside India, remote, many sensors in one place | Local sensor network plus licensed satellite backhaul at a gateway | Licence held by the service provider; product stays on licence-exempt radio |
Asset inside India, widely scattered sensors, no cellular | Do not design around a foreign mobile-satellite terminal; ask a licensed provider which services and terminals are authorised | The legal position needs written confirmation first |
Data is non-urgent | Store and forward | Zero airtime fee |
Mixed fleet across several countries | One bearer-independent envelope, bearer drivers per region | One firmware application, several bearers |
Where to Start Next Week
Write the legal question down. For every deployment region, list who the licensed satellite providers are and what terminals they authorise. For India, send the question to licensed GMPCS operators and obtain answers in writing before any hardware is ordered.
Define the envelope. Specify a compact, versioned binary message with a sequence number and an authentication tag, independent of bearer. Keep it under 200 bytes so it fits SBD comfortably with room to grow.
Abstract the bearer. In firmware, hide the radio behind a small interface (send(msg), poll_downlink(), link_state()) so a different bearer is a new driver, not a rewrite.
Measure, do not quote. Put the candidate module on a bench with a current analyser and record energy per complete message, including GNSS fix and retries, at the temperature extremes of the site.
Split alarms from telemetry. Decide which events must go out immediately and which can batch, and implement that policy in the gateway, not in each node.
Plan the downlink. Decide how you will push configuration and key-rotation commands over the satellite path, and what happens if a node misses one. Mobile-terminated capacity is small and metered.
Pilot before you commit. Run any new service, especially an NB-NTN one, on a small fleet for at least one full season before putting it in a product's bill of materials.
The Order of the Questions
A satellite link feels like a technology choice because the technology is the interesting part: orbits, waveforms, link budgets. In practice, the decisions that decide whether a remote product works are made earlier and are duller: where the asset sits, which services are lawful there, what the message really needs to carry, and how often. Get those right and the choice between SBD, NB-NTN, and a licensed backhaul usually falls out of the answers. Get them wrong and the best radio in orbit does not save the project.
EurthTech is an engineering-first product design and embedded systems company with a development center in Hyderabad. To discuss a project, write to connect@eurthtech.com or see https://www.eurthtech.com/case-studies.
Comments