What a 5G Non-Terrestrial Network Actually Is
A 5G Non-Terrestrial Network, or NTN, is a 5G radio access network where at least part of the path between the user equipment (UE) and the core network runs through a spaceborne or airborne platform instead of a ground-based tower. The platform can be a Low Earth Orbit (LEO) satellite at 300 to 2000 km, a Medium Earth Orbit (MEO) satellite around 8000 km, a Geostationary (GEO) satellite at roughly 35,786 km, or a High Altitude Platform Station (HAPS) drifting in the stratosphere at 20 km. What makes it a 5G NTN specifically, and not just satellite broadband as we have known it for decades, is that the radio interface follows the 3GPP 5G New Radio (NR) specification with a set of modifications defined by 3GPP for the non-terrestrial case.
The practical consequence is that the same silicon inside a 5G phone or a 5G IoT module can, in principle, connect to a satellite that behaves like a very large, very fast-moving cell tower. That is a fundamental shift from Iridium, Inmarsat BGAN, or Thuraya, which used bespoke air interfaces and specialized handsets. In NTN, the modem stack is (mostly) reused, the SIM/USIM is reused, and the core network is reused. The satellite is essentially a new type of gNB or a very long RF cable to a gNB on the ground.
Two architectural variants matter. In the transparent payload model, the satellite is a bent-pipe repeater: it receives NR signals, translates them to a feeder link frequency, and forwards them to a ground gateway where the actual gNB lives. In the regenerative payload model, the satellite hosts the gNB itself, or at least the lower layers of it, and demodulates and re-modulates on board. Release 17 NTN specifies both, but early commercial deployments in 2024 and 2025 have overwhelmingly been transparent because the space-hardened processing needed for regenerative payloads is heavier, more power-hungry, and more expensive.
The motivation is coverage economics. About 80 percent of the Earth's surface, including most oceans, deserts, and polar regions, has no terrestrial cellular signal. Even in developed markets, roughly 15 to 20 percent of the land area is a coverage hole where building a macro tower will never pay back. NTN turns those holes into revenue by piggybacking on the same 5G subscription. For operators, that is compelling. For rural users, hikers, ships, aircraft, oil platforms, and remote IoT sensors, it changes what is possible from a standard-issue handset. And for engineers, it opens an entirely new discipline that sits between classical satcom and cellular RAN, which is what the satellite communications engineer program at Refonte Learning is designed to prepare people for.
The 3GPP Release 17, 18, and 19 Timeline
3GPP Release 17, frozen in mid-2022, is where 5G NTN became a real standard rather than a study item. Release 17 introduced two parallel tracks: NR NTN, which targets smartphone-grade broadband over satellites, and IoT NTN, which targets narrowband LTE-M and NB-IoT devices over satellites. The two tracks share underlying design principles but differ in numerology, coverage assumptions, and battery expectations.
Release 17 solved a set of hard physics problems. The first is propagation delay: a LEO satellite at 600 km has a one-way path around 2 to 8 ms depending on elevation angle, and a GEO satellite has about 120 ms one way. Standard 5G NR timing loops assume terrestrial round-trip times measured in tens of microseconds, so Release 17 extended the timing advance range, redefined how the UE reports its GNSS position to help the network pre-compensate, and enlarged the HARQ (Hybrid ARQ) process count to keep the pipeline full despite the delay. The second is Doppler: a LEO satellite moves at roughly 7.5 km/s, producing Doppler shifts of tens of kHz on S-band and hundreds of kHz on Ka-band. Release 17 requires the UE to know its own position via GNSS and to pre-compensate frequency, offloading the burden from the satellite payload.
Release 18, frozen in 2024, is where NTN grew up. It added support for above-10-GHz bands, better mobility between NTN cells (which is essentially constant, because the beam moves under the user), and, critically, initial specification work for direct-to-handset service in bands originally allocated to mobile network operators (MNOs). This last point is the technical bedrock under the SpaceX-plus-T-Mobile and AST SpaceMobile services.
Release 19, with completion targeted for 2025 and early 2026, is refining the regenerative payload architecture, adding store-and-forward for deep IoT, enabling multi-connectivity between terrestrial and non-terrestrial cells, and improving positioning. It also expands the frequency scope: FR1 (below 7.125 GHz), FR2 (24 to 52.6 GHz), and new NTN-specific bands. By the time you read a device datasheet in 2026, expect to see a Release 17 or Release 18 NTN capability line item next to the usual band list.
One subtle point that trips people up: the difference between 5G and 5G NR is not a technology gap, it is a scope question. 5G is the marketing and system-level term for the entire fifth-generation mobile system, including the 5G Core (5GC), the New Radio air interface, and the service and operational architecture. 5G NR is the radio access layer specifically, the OFDM-based air interface that replaces LTE's air interface. 5G NTN is a set of features on top of 5G NR that let it work over satellite links. All 5G NTN is 5G NR, but not all 5G NR is NTN.
Direct-to-Cell: The Race That Actually Matters
Direct-to-Cell, sometimes written Direct-to-Device or D2D, means an unmodified consumer smartphone connecting directly to a satellite without any external antenna, terminal, or dish. This was science fiction in 2020. It is a commercial reality in 2025 and will be widespread in 2026.
Three players dominate the current landscape. SpaceX Starlink, in partnership with T-Mobile in the United States and roughly a dozen international carriers, is deploying V2-mini and V3 satellites with a large phased-array antenna panel that transmits in T-Mobile's mid-band PCS spectrum around 1900 MHz. The service started with SMS in 2024, added limited data and voice through 2025, and is expected to reach broader broadband-adjacent throughputs during 2026. Because SpaceX has a fully vertically integrated launch cadence, the constellation is being built out faster than any competitor can match.
AST SpaceMobile is pursuing a different technical bet: much larger satellites (BlueBird generation, with roughly 64 square meter phased-array antennas) that can deliver higher effective isotropic radiated power and therefore support proper broadband to phones. AST has partnerships with AT&T, Verizon, Vodafone, Rakuten, and others. The tradeoff is deployment speed: each BlueBird is expensive and physically large, so the constellation grows more slowly than Starlink's D2C fleet.
Lynk Global is a smaller, focused player targeting SMS and emergency messaging in markets where MNOs want an easy roaming-style deal. Lynk has demonstrated commercial service in the Solomon Islands, Palau, and several African markets, and it is likely to remain the low-cost, narrowband option throughout 2026.
So is Starlink a non-terrestrial network? The consumer Starlink service using a Dishy terminal is a proprietary Ku-band satellite broadband system, not a 3GPP NTN. But Starlink Direct-to-Cell, which uses the satellite as a base station serving standard LTE and 5G NR handsets in MNO spectrum, is functionally an NTN, and SpaceX has been aligning it with 3GPP specifications. In 2026 the answer will be: consumer Starlink broadband is a proprietary satcom system that happens to use similar physics; Starlink Direct-to-Cell is a 5G NTN in practical terms. For a deeper look at how these constellations differ commercially, see the Starlink, OneWeb, and Kuiper competition breakdown.
IoT NTN Versus Handset NTN: Two Very Different Design Points
It is tempting to lump all NTN together, but IoT NTN and handset NTN are engineered for opposite corners of the design space, and understanding the split is essential for anyone building products for this market.
IoT NTN is based on NB-IoT and LTE-M radio profiles adapted for satellite. The design goals are low power (a battery lasting 5 to 10 years), low data rate (kilobits per second, not megabits), low device cost (a single-chip modem under 5 USD in volume), and enormous coverage tolerance. IoT NTN devices typically know their own position via a GNSS receiver, wake up on a schedule or when triggered, transmit a short burst, and go back to sleep. Latency is not a constraint: a soil moisture sensor in the Sahel does not care whether its reading arrives in 200 ms or 20 minutes, so store-and-forward architectures over LEO or MEO satellites are acceptable. The commercial players here include Skylo (running over Inmarsat and Viasat GEO satellites in L-band), Sateliot (LEO), and OQ Technology. Most of these run through operator partnerships with MNOs so that the same SIM roams onto NTN when out of terrestrial coverage.
Handset NTN is the exact opposite. The device is a smartphone with a fixed antenna, a fixed maximum transmit power around 23 dBm, and a user who expects to make a voice call or send a photo in near real time. Latency budgets are tight, throughput expectations start at hundreds of kilobits per second and scale to a few megabits, and the link budget is dominated by the phone's tiny antenna and body loss. This is why AST SpaceMobile's massive aperture matters: to close the link to a phone at reasonable rates, you need either a huge satellite antenna or a very low orbit or both.
A subtle intermediate case is fixed and vehicular NTN, where the terminal is not a phone but a purpose-built antenna on a truck, ship, aircraft, or fixed rooftop. These terminals can afford directional antennas and higher transmit power, and they are the natural evolution of the VSAT industry into a 3GPP framework. For an introduction to the classical side of that world, the VSAT and teleport ground station basics article covers the ground-segment fundamentals that still apply.
One consequence of these very different design points: an engineer working on IoT NTN spends a lot of time on power-optimized MAC layer behavior, GNSS-assisted synchronization, and store-and-forward routing. An engineer working on handset NTN spends time on link budgets, beamforming, spectrum coordination with terrestrial networks, and Doppler compensation. The core skills overlap heavily, which is why RF and 5G-literate engineers with a satcom sensibility are the ones being aggressively recruited in 2026.
Link Budgets, Doppler, and the RF Physics That Actually Bind
The reason NTN is hard, and the reason it took 3GPP two full releases to get workable, is the RF physics. Consider a LEO satellite at 600 km altitude serving a 5G phone in S-band around 2 GHz. The free space path loss at that distance is roughly 154 dB. Add 2 to 5 dB of atmospheric and rain loss, another 3 to 6 dB of body loss and antenna mismatch on the phone, and you are looking for the satellite to close a link that is 165 dB down from the transmitter output.
On the downlink, the satellite can compensate with a large phased-array antenna: 30 to 40 dBi of gain, plus enough transmit power to deliver a per-beam EIRP in the 60 to 70 dBW range. That is achievable with current bus platforms. On the uplink, however, the phone is limited to 23 dBm from a near-omnidirectional antenna, which means the satellite receiver needs a very low noise temperature and a very high aperture gain to hear it. This is why AST SpaceMobile's antenna is 64 square meters, and it is why Starlink's D2C service currently offers narrowband messaging and voice more comfortably than high-throughput data. The uplink is the binding constraint.
Doppler is the second physics problem. A LEO satellite crossing overhead produces a Doppler frequency shift that swings from about +50 kHz to -50 kHz at S-band over the course of a pass, plus a Doppler rate of change that must be tracked. Release 17 solves this by requiring the UE to know its own position and the ephemeris of the serving satellite (broadcast in system information), then pre-compensating both frequency and timing before transmission. Without that trick, the base station would need extraordinary Doppler search ranges during random access, which is not practical.
Modulation and coding choices flow from these constraints. Handset NTN typically uses QPSK and low-rate turbo or LDPC codes on the uplink, reserving 16QAM or 64QAM for the downlink where the link budget is comfortable. Higher-order schemes such as 8PSK and 16APSK, common in DVB-S2X satcom, appear more often in fixed-terminal NTN and feeder links than on the direct-to-handset service. The tradeoff between spectral efficiency and required Es/N0 is the core design lever, and we cover it in depth in the article on modulation schemes for satcom.
An often-overlooked issue is interference. A LEO D2C constellation reusing MNO spectrum broadcasts into a country where that spectrum is licensed to a specific carrier, but the satellite footprint spills across national borders. Coordinating those spillovers with regulators, and with adjacent-country MNOs that hold the same or nearby spectrum, is a nontrivial regulatory exercise. The ITU coordination process, national regulator agreements, and inter-operator commercial agreements together form a matrix that is currently being worked out in real time.
Spectrum, Regulation, and the ITU
Spectrum is where NTN gets political. The bands used for NTN fall into two categories: MSS (Mobile Satellite Service) bands, which have been allocated for satellite use for decades, and IMT bands, which are the terrestrial mobile bands that MNOs paid billions for.
MSS bands include L-band (around 1.5 to 1.6 GHz), S-band (around 2 GHz), and portions of Ka-band (17 to 30 GHz). These are the natural home for satellite services. 3GPP Release 17 initially targeted these bands. The problem is that MSS spectrum is fragmented, held by a small number of legacy operators (Inmarsat, EchoStar, Ligado, Iridium), and expensive to acquire.
The more disruptive move, which Release 18 formalized, is using IMT bands directly. When Starlink uses T-Mobile's PCS spectrum from orbit, the satellite is broadcasting in a band that the ITU has allocated for terrestrial mobile use. This works legally only because T-Mobile owns the spectrum in the United States and has authorized Starlink to use it there, under a Supplemental Coverage from Space (SCS) framework that the FCC formalized in 2024. Other regulators, including Ofcom in the UK, the European Commission, and various Asian regulators, are working out their own equivalent frameworks in 2025 and 2026.
The consequences are large. If SCS is broadly authorized, an MNO with a good satellite partner effectively extends its network coverage across the planet without new tower CapEx. If it is restricted, NTN stays in the MSS ghetto and grows more slowly. The ITU World Radiocommunication Conference decisions of the mid-2020s will shape which path dominates by decade's end.
A second regulatory frontier is emergency service obligations. Regulators increasingly want NTN services to support emergency calling, location reporting, and lawful intercept. That is not trivial when the serving cell is moving at 7.5 km/s and covering multiple jurisdictions. Expect this to be a compliance workstream at every serious NTN operator through 2026 and 2027.
Ground Segment: Gateways, Feeder Links, and Core Integration
Even the most futuristic NTN needs a very grounded ground segment. Every satellite eventually talks to a gateway on Earth, and that gateway hosts (in the transparent payload case) the gNB or (in the regenerative case) the interface between the satellite-hosted gNB and the 5G Core.
Gateways for LEO NTN constellations are geographically distributed to ensure that every operational satellite has line of sight to at least one gateway most of the time. Feeder links typically operate in Ka-band or higher (17 to 30 GHz, or Q/V-band at 40 to 50 GHz for newer designs) because those bands offer the bandwidth needed to backhaul dozens or hundreds of user beams. Rain fade at Ka-band and above is significant, so gateway sites are chosen for climate as well as connectivity, and site diversity (two gateways separated by enough distance that they are unlikely to be in rain at the same time) is a standard mitigation.
The 5G Core (5GC) integration is where NTN starts to look like normal 5G. The gNB, whether on the satellite or in the gateway, connects to an AMF, SMF, and UPF via standard N2 and N3 interfaces. UEs authenticate with a normal USIM. Session management, QoS, and handover mostly work the same. The interesting engineering happens in the edges: how does mobility work when the UE is stationary but the cell is moving? How does location-based routing work when the UE is in one country but the gateway is in another? How does lawful intercept work when the traffic never touches a tower in the country where the user is standing?
The distinction between what happens in space and what happens on the ground is fundamental to how NTN is engineered and operated, and the tradeoffs between the two are covered more fully in the ground segment versus space segment breakdown.
Inter-satellite links (ISLs), typically optical, are the other piece that changes ground segment economics. A constellation with fast optical ISLs can route traffic from a satellite over Kenya to a gateway in Ireland without ever touching a ground station in between. This reduces the number of gateways needed, but it also means the constellation is now a routing fabric in its own right, running something like a space-hardened BGP or a custom routing protocol. That is a new discipline for network engineers, and it is where an increasing share of NTN engineering hiring is concentrated.
Use Cases That Actually Pay in 2026
Hype cycles produce lists of exciting use cases that never generate revenue. Here are the NTN use cases where money is already changing hands in 2025 and where the volumes will scale meaningfully in 2026.
First, emergency and off-grid messaging. Apple's satellite Emergency SOS on iPhone (a Globalstar-based service that is not strictly 3GPP NTN but demonstrates the market) proved that consumers value satellite fallback. T-Mobile plus Starlink is generalizing that to any 5G phone, initially for SMS. This is a feature that MNOs use to differentiate their subscription plans, and the willingness to pay is real: 5 to 15 USD per month uplifts in early market research.
Second, maritime and aviation connectivity. Ships, aircraft, and offshore platforms have been paying premium prices for satcom for decades. NTN offers these operators the promise of using standard 5G modems, drastically reducing hardware and integration cost. Airline in-flight connectivity vendors are actively building NTN-capable aero terminals. Ships, especially in the fishing and small-vessel segment where legacy VSAT is too expensive, are a large addressable market.
Third, IoT for agriculture, mining, energy, and logistics. A pipeline monitoring sensor in Kazakhstan, a soil sensor in the Australian outback, a container tracker on an intermodal route: all of these want a globally roaming, low-power, low-cost cellular connection. IoT NTN, priced at a few dollars per device per year for a few messages per day, is a plausible commercial fit.
Fourth, first responder and government communications. Public safety agencies want redundant coverage that does not depend on terrestrial infrastructure surviving a wildfire, earthquake, or hurricane. Purpose-built NTN services for first responders are being contracted in the United States, Europe, and parts of Asia.
Fifth, private 5G extensions. A mining company running a private 5G network on a remote site can use NTN as backhaul or as a coverage extension for staff who wander outside the on-site network. This is a smaller market by volume but a high-margin one where operators can charge for the specific integration work.
Markets that will probably not pay in 2026 include consumer broadband to phones as a primary connection (the throughput and cost are wrong for that), and dense urban augmentation (terrestrial 5G already does that better and cheaper). Being clear about what NTN is not is as important as being clear about what it is.
Career Paths for RF, Comms, and Systems Engineers
The NTN wave is creating a job market that did not exist five years ago. It sits at the intersection of three previously separate specializations: cellular RAN engineering (Ericsson, Nokia, Huawei, Samsung, and the open-RAN vendors), classical satcom engineering (Airbus, Thales, Lockheed, Northrop, and the operator side at SES, Intelsat, Viasat), and constellation-native new-space engineering (SpaceX, AST SpaceMobile, Lynk, OneWeb, Kuiper, and a long tail of startups). Very few engineers have deep experience in all three, and the ones who do are commanding significant premiums.
The most in-demand roles in 2026 fall into a few clusters. RF systems engineers with link budget expertise are needed to design and verify the physical layer of NTN services. NR RAN software engineers who understand the specific Release 17 and 18 modifications for NTN are needed by every base station and modem vendor. Payload engineers, especially those with digital regenerative payload experience, are increasingly rare and highly compensated. Network engineers with a mix of 5G Core, IP routing, and space-aware architecture skills are being hired to build the ground-plus-space fabric.
Compensation is aggressive. In the United States, senior NTN systems engineers at established primes were reporting total compensation in the 220,000 to 350,000 USD range in 2024, and new-space companies competing for the same talent were often paying more in equity. In Europe, base salaries are lower but the availability of interesting programs (Airbus, Thales Alenia Space, OHB, and various ESA-funded initiatives) makes the market thick. A detailed breakdown of the wider satcom career landscape is available in the satellite communications engineer career guide.
How do you break in without existing experience? Three things help disproportionately. First, learn 3GPP specifications directly. Read TS 38.300, TS 38.211, TR 38.821, and the NTN-specific work items. This is dry material, but the people who can quote it fluently are the people who get hired. Second, build hands-on RF intuition. Playing with SDRs (HackRF, USRP, LimeSDR), running GNU Radio, working through link budgets in a spreadsheet, and simulating basic OFDM modems in Python builds a physical intuition that no amount of theory replaces. Third, work on integration and testing exposure. The gap between a specification and a working system is where most of the real engineering happens.
Practical Learning Path for 2026
If you are an engineer, student, or career-switcher planning to move into NTN in 2026, here is a sequence that has worked for people who came through Refonte Learning programs and similar tracks.
Start with foundational cellular. Understand LTE first (it is simpler and the material is mature), then move to 5G NR. Ken Wesolowski's textbooks and 3GPP's own specifications are the gold standard, but do not underestimate the value of hands-on tools: srsRAN and OpenAirInterface let you actually run a base station on a laptop with an SDR frontend. You will learn more in a weekend of debugging RACH failures than in a month of reading.
Next, build classical satcom fundamentals. Learn the geometry of orbits (LEO, MEO, GEO, HEO), the mechanics of link budgets, the tradeoffs between frequency bands (L, S, C, X, Ku, Ka, Q, V, W), and the structure of a satcom system from ground modem to satellite payload to gateway. Pratt and Bostian is a classic reference. Then layer on the NTN-specific modifications: extended timing advance, GNSS-assisted UE pre-compensation, extended HARQ, feeder link handover.
After that, pick a specialization. If your background is more software and networking, lean into 5G Core integration and constellation routing. If your background is more physics and electronics, lean into RF payload design and antenna arrays. If you are more of a systems generalist, lean into mission architecture and program engineering. The field is broad enough that you do not have to be a jack of all trades to be valuable; you have to be a genuine expert in one thing while being conversant in the adjacent things.
One pragmatic recommendation: engage with the industry's open forums early. Attend a 3GPP RAN plenary as an observer if your employer permits, follow the NTN work-item threads on 3GPP's site, read the technical papers from IEEE Globecom and Milcom, and watch the ITU-R working party proceedings. Because the standards and regulations are being written now, the people who read the primary sources have a real edge over those who wait for tutorials to be published.
For a structured pathway that combines RF fundamentals, satellite systems, and hands-on 5G NTN exposure, Refonte Learning's satellite communications engineer program is designed to compress this journey while providing mentorship and portfolio-grade projects.
What to Watch in 2026 and Beyond
A few developments will define the NTN landscape by the end of 2026 and into 2027.
The first is whether Direct-to-Cell scales beyond messaging into real broadband. This depends on how quickly SpaceX and AST SpaceMobile can put larger, more capable satellites in orbit, and on whether the FCC and international regulators approve the higher-power operations needed to close broadband links. If broadband to unmodified phones becomes routine even at a few Mbps, the terrestrial coverage gap effectively closes for messaging, voice, and light data use in most of the world.
The second is the maturation of Release 19 and the start of Release 20 study items. Release 20 is expected to formally begin the work that eventually becomes 6G, and NTN is going to be a first-class citizen from the start rather than an afterthought added late as it was for 5G. Expect 6G to assume from day one that terrestrial and non-terrestrial coverage are a single integrated fabric.
The third is the fate of the smaller and mid-tier constellations. OneWeb, now merged with Eutelsat, is repositioning around enterprise and government. Amazon's Project Kuiper is beginning commercial deployment. Chinese constellations (Guowang and Qianfan) are launching at increasing pace. The competitive dynamics between these players will determine whether NTN is a duopoly, an oligopoly, or something more open.
The fourth is the interaction with terrestrial 5G evolution. Terrestrial 5G-Advanced (Release 18 features) is deploying now, and it introduces massive MIMO improvements, network energy savings, and better positioning. NTN inherits some of these features but has to adapt them for the satellite context. The pace at which NTN keeps up with terrestrial 5G-Advanced will determine whether it feels like a natural extension of the same network or a permanently second-class add-on.
The fifth is spectrum. The next World Radiocommunication Conference (WRC-27) will make decisions on additional NTN spectrum allocations, on the terms of IMT-band SCS operation, and on adjacent-service protection. These decisions will shape the economics of NTN for the following decade.
About Refonte Learning
Refonte Learning is an EdTech platform training the next generation of engineers in the domains where the industry is moving fastest, including satellite communications, 5G, AI, cloud, and data. The satellite communications track is taught by working engineers with experience across primes, operators, and new-space constellations, and it covers RF link budgets, modulation, ground stations, payload architecture, and the 5G NTN standards discussed in this article.
If you are planning to move into satellite communications or NTN engineering in 2026, the satellite communications engineer program at Refonte Learning is built to compress the learning curve, connect you with mentors already in the field, and give you portfolio-grade project experience that hiring managers actually look at. The market is real, the standards are settled enough to teach, and the demand for capable engineers is only going to grow through the second half of the decade.