It’s 2045. Earth is green again. Free from cellular towers and the terrestrial radiation of yet another G, no longer needed to justify endless telecom upgrades. Humanity has finally transcended its communication needs to the sky, fully served by swarms of Low Earth Orbit (LEO) satellites.

Millions of mobile towers have vanished. No more steel skeletons cluttering skylines and nature in general. In their place: millions of beams from tireless LEO satellites, now whispering directly into our pockets from orbit.

More than 1,200 MHz of once terrestrially-bound cellular spectrum below the C-band had been uplifted to LEO satellites. Nearly 1,500 MHz between 3 and 6 GHz had likewise been liberated from its earthly confines, now aggressively pursued by the buzzing broadband constellations above.

It all works without a single modification to people’s beloved mobile devices. Everyone enjoyed the same, or better, cellular service than in those wretched days of clinging to terrestrial-based infrastructure.

So, how did this remarkable transformation come about?

THE COVERAGE.

First, let’s talk about coverage. The chart below tells the story of orbital ambition through three very grounded curves. On the x-axis, we have the inclination angle, which is the degree to which your satellites are encouraged to tilt away from the equator to perform their job. On the y-axis: how much of the planet (and its people) they’re actually covering. The orange line gives us land area coverage. It starts low, as expected, tropical satellites don’t care much for Greenland. But as the inclination rises, so does their sense of duty to the extremes (the poles that is). The yellow line represents population coverage, which grows faster than land, maybe because humans prefer to live near each other (or they like the scenery). By the time you reach ~53° inclination, you’re covering about 94% of humanity and 84% of land areas. The dashed white line represents mobile cell coverage, the real estate of telecom towers. A constellation at a 53° inclination would cover nearly 98% of all mobile site infrastructure. It serves as a proxy for economic interest. It closely follows the population curve, but adds just a bit of spice, reflecting urban density and tower sprawl.

The satellite constellation’s beams have replaced traditional terrestrial cells, providing a one-to-one coverage substitution. They not only replicate coverage in former legacy cellular areas but also extend service to regions that previously lacked connectivity due to low commercial priority from telecom operators. Today, over 3 million beams substitute obsolete mobile cells, delivering comparable service across densely populated areas. An additional 1 million beams have been deployed to cover previously unserved land areas, primarily rural and remote regions, using broader, lower-capacity beams with radii up to 10 kilometers. While these rural beams do not match the density or indoor penetration of urban cellular coverage, they represent a cost-effective means of achieving global service continuity, especially for basic connectivity and outdoor access in sparsely populated zones.

Conclusion? If you want to build a global satellite mobile network, you don’t need to orbit the whole planet. Just tilt your constellation enough to touch the crowded parts, and leave the tundra to the poets. However, this was the “original sin” of LEO Direct-2-Cellular satellites.

THE DEMAND.

Although global mobile traffic growth slowed notably after the early 2020s, and the terrestrial telecom industry drifted toward its “end of history” moment, the orbital network above inherited a double burden. Not only did satellite constellations need to deliver continuous, planet-wide coverage, a milestone legacy telecoms had never reached, despite millions of ground sites, but they also had to absorb globally converging traffic demands as billions of users crept steadily toward the throughput mean.

The radio access uplink architecture relies on low round-trip times for proper scheduling, timing alignment, and HARQ (Hybrid Automatic Repeat Request) feedback cycles. The propagation delay at 350 km yields a round-trip time of about 2.5 to 3 milliseconds, which falls within the bounds of what current specifications can accommodate. This is particularly important for latency-sensitive applications such as voice, video, and interactive services that require low jitter and reliable feedback mechanisms. In contrast, orbits at 550 km or above push latency closer to the edge of what NR protocols can tolerate, which could hinder performance or require non-standard adaptations. The beam geometry also plays a central role. At lower altitudes, satellite beams projected to the ground are inherently smaller. This smaller footprint translates into tighter beam patterns with narrower 3 dB cut-offs, which significantly improves frequency reuse and spatial isolation. These attributes are important for deploying high-capacity networks in densely populated urban environments, where interference and spectrum efficiency are paramount. Narrower beams allow D2C operators to steer coverage toward demand centers while minimizing adjacent-beam interference dynamically. Operating at 350 km is not without drawbacks. The satellite’s ground footprint at this altitude is smaller, meaning that more satellites are required to achieve full Earth coverage. Additionally, satellites at this altitude are exposed to greater atmospheric drag, resulting in shorter orbital lifespans unless they are equipped with more powerful or efficient propulsion systems to maintain altitude. The current design aims for a 5-year orbital lifespan. Despite this, the shorter lifespan has an upside, as it reduces the long-term risks of space debris. Deorbiting occurs naturally and quickly at lower altitudes, making the constellation more sustainable in the long term.

THE CONSTELLATION.

The satellite-to-cellular infrastructure has now fully matured into a global-scale system capable of delivering mobile broadband services that are not only on par with, but in many regions surpass, the performance of terrestrial cellular networks. At its core lies a constellation of low Earth orbit satellites operating at an altitude of 350 kilometers, engineered to provide seamless, high-quality indoor coverage for both uplink and downlink, even in densely urban environments.

To meet the evolving expectations of mobile users, each satellite beam delivers a minimum of 50 Mbps of uplink capacity and 500 Mbps of downlink capacity per user, ensuring full indoor quality even in highly cluttered environments. Uplink transmissions utilize the 600 MHz to 1800 MHz band, providing 1200 MHz of aggregated bandwidth. Downlink channels span 1500 MHz of spectrum, ranging from 2100 MHz to the upper edge of the C-band. At the network’s busiest hour (e.g., around 20:00 local time) across the most densely populated regions south of 53° latitude, the system supports a peak throughput of 60,000 Tbps for downlink and 6,000 Tbps for uplink. To guarantee reliability under real-world utilization, the system is engineered with a 25% capacity overhead, raising the design thresholds to 75,000 Tbps for DL and 7,500 Tbps for UL during peak demand.

Each satellite beam is optimized for high spectral efficiency, leveraging advanced beamforming, adaptive coding, and cutting-edge modulation. Under these conditions, downlink beams deliver 4.5 Gbps, while uplink beams, facing more challenging reception constraints, achieve 1.8 Gbps. Meeting the adjusted peak-hour demand requires approximately 16.7 million active DL beams and 4.2 million UL beams, amounting to over 20.8 million simultaneous beams concentrated over the peak demand region.

Thanks to significant advances in onboard processing and power systems, each satellite now supports up to 5,000 independent beams simultaneously. This capability reduces the number of satellites required to meet regional peak demand to approximately 4,200. These satellites are positioned over a region spanning an estimated 45 million square kilometers, covering the evening-side urban and suburban areas of the Americas, Europe, Africa, and Asia. This configuration yields a beam density of nearly 0.46 beams per square kilometer, equivalent to one active beam for every 2 square kilometers, densely overlaid to provide continuous, per-user, indoor-grade connectivity. In urban cores, beam radii are typically below 1 km, whereas in lower-density suburban and rural areas, the system adjusts by using larger beams without compromising throughput.

Because peak demand rotates longitudinally with the Earth’s rotation, only a portion of the entire constellation is positioned over this high-demand region at any given time. To ensure 4,200 satellites are always present over the region during peak usage, the total constellation comprises approximately 20,800 satellites, distributed across several hundred orbital planes. These planes are inclined and phased to optimize temporal availability, revisit frequency, and coverage uniformity while minimizing latency and handover complexity.

The resulting Direct-to-Cellular satellite constellation and system of today is among the most ambitious communications infrastructures ever created. With more than 20 million simultaneous beams dynamically allocated across the globe, it has effectively supplanted traditional mobile towers in many regions, delivering reliable, high-speed, indoor-capable broadband connectivity precisely where and when people need it.

THE SATELLITE.

The Cellular Device to Satellite Path.

The uplink antennas aboard the Direct-to-Cellular satellites have been specifically engineered to reliably receive indoor-quality transmissions from standard (unmodified) mobile devices operating within the 600 MHz to 1800 MHz band. Each device is expected to deliver a minimum of 50 Mbps uplink throughput, even when used indoors in heavily cluttered urban environments. This performance is made possible through a combination of wideband spectrum utilization, precise beamforming, and extremely sensitive receiving systems in orbit. The satellite uplink system operates across 1200 MHz of aggregated bandwidth (e.g., 60 channels of 20 MHz), spanning the entire upper UHF and lower S-band. Because uplink signals originate from indoor environments, where wall and structural penetration losses can exceed 20 dB, the satellite link budget must compensate for the combined effects of indoor attenuation and free-space propagation at a 350 km orbital altitude. At 600 MHz, which represents the lowest frequency in the UL band, the free space path loss alone is approximately 133 dB. When this is compounded with indoor clutter and penetration losses, the total attenuation the satellite must overcome reaches approximately 153 dB or more.

Rather than specifying the antenna system at a mid-band average frequency, such as 900 MHz (i.e., the mid-band of the 600 MHz to 1800 MHz range), the system has been conservatively engineered for worst-case performance at 600 MHz. This design philosophy ensures that the antenna will meet or exceed performance requirements across the entire uplink band, with higher frequencies benefiting from naturally improved gain and narrower beamwidths. This choice guarantees that even the least favorable channels, those near 600 MHz, support reliable indoor-grade uplink service at 50 Mbps, with a minimum required SNR of 10 dB to sustain up to 16-QAM modulation. Achieving this level of performance at 600 MHz necessitated a large physical aperture. The uplink receive arrays on these satellites have grown to approximately 700 to 750 m² in area, and are constructed using modular, lightweight phased-array tiles that unfold in orbit. This aperture size enables the satellite to achieve a receive gain of approximately 45 dBi at 600 MHz, which is essential for detecting low-power uplink transmissions with high spectral efficiency, even from users deep indoors and under cluttered conditions.

Unlike earlier systems, such as AST SpaceMobile’s BlueBird 1, launched in the mid-2020s with an aperture of around 900 m² and challenged by the need to acquire indoor uplink signals, today’s Direct-to-Cellular (D2C) satellites optimize the uplink and downlink arrays separately. This separation allows each aperture to be custom-designed for its frequency and link budget requirements. The uplink arrays incorporate wideband, dual-polarized elements, such as log-periodic or Vivaldi structures, backed by high-dynamic-range low-noise amplifiers and a distributed digital beamforming backend. Assisted by real-time AI beam management, each satellite can simultaneously support and track up to 2,500 uplink beams, dynamically allocating them across the active coverage region.

Despite their size, these receive arrays are designed for compact launch configurations and efficient in-orbit deployment. Technologies such as inflatable booms, rigidizable mesh structures, and ultralight composite materials allow the arrays to unfold into large apertures while maintaining structural stability and minimizing mass. Because these arrays are passive receivers, thermal loads are significantly lower than those of transmit systems. Heat generation is primarily limited to the digital backend and front-end amplification chains, which are distributed across the array surface to facilitate efficient thermal dissipation.

The Satellite to Cellular Device Path.

The downlink communication path aboard Direct-to-Cellular satellites is engineered as a fully independent system, physically and functionally separated from the uplink antenna. This separation reflects a mature architectural philosophy that has been developed over decades of iteration. The downlink and uplink systems serve fundamentally different roles and operate across vastly different frequency bands, with their power, thermal, and antenna constraints. The downlink system operates in the frequency range from 2100 MHz up to the upper end of the C-band, typically around 4200 MHz. This is significantly higher than the uplink range, which extends from 600 to 1800 MHz. Due to this disparity in wavelength, a factor of nearly six between the lowest uplink and highest downlink frequencies, a shared aperture is neither practical nor efficient. It is widely accepted today that integrating transmit and receive functions into a single broadband aperture would compromise performance on both ends. Instead, today’s satellites utilize a dual-aperture approach, with the downlink antenna system optimized exclusively for high-frequency transmission and the uplink array designed independently for low-frequency reception.

In order to deliver 500 Mbps per user with full indoor coverage, each downlink beam must sustain approximately 4.5 Gbps, accounting for spectral reuse and beam overlap. At an orbital altitude of 350 kilometers, downlink beams must remain narrow, typically covering no more than a 1-kilometer radius in urban zones, to match uplink geometry and maintain beam-level concurrency. The antenna gain required to meet these demands is in the range of 50 to 55 dBi, which the satellites achieve using high-frequency phased arrays with a physical aperture of approximately 100 to 200 m². Because the downlink system is responsible for high-power transmission, the antenna tiles incorporate GaN-based solid-state power amplifiers (SSPAs), which deliver hundreds of watts per panel. This results in an overall effective isotropic radiated power (EIRP) of 50 to 60 dBW per beam, sufficient to reach deep indoor devices even at the upper end of the C-band. The power-intensive nature of the downlink system introduces thermal management challenges (describe below in the next section), which are addressed by physically isolating the transmit arrays from the receiver surfaces. The downlink and uplink arrays are positioned on opposite sides of the spacecraft bus or thermally decoupled through deployable booms and shielding layers.

The downlink beamforming is fully digital, allowing real-time adaptation of beam patterns, power levels, and modulation schemes. Each satellite can form and manage up to 2,500 independent downlink beams, which are coordinated with their uplink counterparts to ensure tight spatial and temporal alignment. Advanced AI algorithms help shape beams based on environmental context, usage density, and user motion, thereby further improving indoor delivery performance. The modulation schemes used on the downlink frequently reach 256-QAM and beyond, with spectral efficiencies of six to eight bits per second per Hz in favorable conditions.

The physical deployment of the downlink antenna varies by platform, but most commonly consists of front-facing phased array panels or cylindrical surfaces fitted with azimuthally distributed tiles. These panels can be either fixed or mounted on articulated platforms that allow active directional steering during orbit, depending on the beam coverage strategy, an arrangement also called gumballed.

Satellite System Architecture.

The Direct-to-Cellular satellites have evolved into high-performance, orbital base stations that far surpass the capabilities of early systems, such as AST SpaceMobile’s Bluebird 1 or SpaceX’s Starlink V2 Mini. These satellites are engineered not merely to relay signals, but to deliver full-featured indoor mobile broadband connectivity directly to standard handheld devices, anywhere on Earth, including deep urban cores and rural regions that have been historically underserved by terrestrial infrastructure.

As described earlier, today’s D2C satellite supports up to 5,000 simultaneous beams, enabling real-time uplink and downlink with mobile users across a broad frequency range. The uplink phased array, designed to capture low-power, deep-indoor signals at 600 MHz, occupies approximately 750 m². The DL array, optimized for high-frequency, high-power transmission, spans 150 to 200 m². Unlike early designs, such as Bluebird 1, which used a single, large combined antenna, today’s satellites separate the uplink and downlink arrays to optimize each for performance, thermal behavior, and mechanical deployment. These two systems are typically mounted on opposite sides of the satellite and thermally isolated from one another.

Thermal management is one of the defining challenges of this architecture. While AST’s Bluebird 1 (i.e., from mid-2020s) boasted a large antenna aperture approaching 900 m², its internal systems generated significantly less heat. Bluebird 1 operated with a total power budget of approximately 10 to 12 kilowatts, primarily dedicated to a handful of downlink beams and limited onboard processing. In contrast, today’s D2C satellite requires a continuous power supply of 25 to 35 kilowatts, much of which must be dissipated as heat in orbit. This includes over 10 kilowatts of sustained RF power dissipation from the DL system alone, in addition to thermal loads from the digital beamforming hardware, AI-assisted compute stack, and onboard routing logic. The key difference lies in beam concurrency and onboard intelligence. The satellite manages thousands of simultaneous, high-throughput beams, each dynamically scheduled and modulated using advanced schemes such as 256-QAM and beyond. It must also process real-time uplink signals from cluttered environments, allocate spectral and spatial resources, and make AI-driven decisions about beam shape, handovers, and interference mitigation. All of this requires a compute infrastructure capable of delivering 100 to 500 TOPS (tera-operations per second), distributed across radiation-hardened processors, neural accelerators, and programmable FPGAs. Unlike AST’s Bluebird 1, which offloaded most of its protocol stack to the ground, today’s satellites run much of the 5G core network onboard. This includes RAN scheduling, UE mobility management, and segment-level routing for backhaul and gateway links.

This computational load compounds the satellite’s already intense thermal environment. Passive cooling alone is insufficient. To manage thermal flows, the spacecraft employs large radiator panels located on its outer shell, advanced phase-change materials embedded behind the DL tiles, and liquid loop systems that transfer heat from the RF and compute zones to the radiative surfaces. These thermal systems are intricately zoned and actively managed, preventing the heat from interfering with the sensitive UL receive chains, which require low-noise operation under tightly controlled thermal conditions. The DL and UL arrays are thermally decoupled not just to prevent crosstalk, but to maintain stable performance in opposite thermal regimes: one dominated by high-power transmission, the other by low-noise reception.

To meet its power demands, the satellite utilizes a deployable solar sail array that spans 60 to 80 m². These sails are fitted with ultra-high-efficiency solar cells capable of exceeding 30–35% efficiency. They are mounted on articulated booms that track the sun independently from the satellite’s Earth-facing orientation. They provide enough current to sustain continuous operation during daylight periods, while high-capacity batteries, likely based on lithium-sulfur or solid-state chemistry, handle nighttime and eclipse coverage. Compared to the Starlink V2 Mini, which generates around 2.5 to 3.0 kilowatts, and the Bluebird 1, which operates at roughly 10–12 kilowatts. Today’s system requires nearly three times the generation and five times the thermal rejection capability compared to the initial satellites of the mid-2020s.

Structurally, the satellite is designed to support this massive infrastructure. It uses a rigid truss core (i.e., lattice structure) with deployable wings for the DL system and a segmented, mesh-based backing for the UL aperture. Propulsion is provided by Hall-effect or ion thrusters, with 50 to 100 kilograms of inert propellant onboard to support three to five years of orbital station-keeping at an altitude of 350 kilometers. This height is chosen for its latency and spatial reuse advantages, but it also imposes continuous drag, requiring persistent thrust.

The AST Bluebird 1 may have appeared physically imposing in its time due to its large antenna, thermal, computational, and architectural complexity. Today’s D2C satellite, 20 years later, far exceeds anything imagined two decades earlier. The heat generated by its massive beam concurrency, onboard processing, and integrated network core makes its thermal management system not only more severe than Bluebird 1’s but also one of the primary limiting factors in the satellite’s physical and functional design. This thermal constraint, in turn, shapes the layout of its antennas, compute stack, power system, and propulsion.

Mass and Volume Scaling.

The AST’s Bluebird 1, launched in the mid-2020s, had a launch mass of approximately 1,500 kilograms. Its headline feature was a 900 m² unfoldable antenna surface, designed to support direct cellular connectivity from space. However, despite its impressive aperture, the system was constrained by limited beam concurrency, modest onboard computing power, and a reliance on terrestrial cores for most network functions. The bulk of its mass was dominated by structural elements supporting its large antenna surface and the power and thermal subsystems required to drive a relatively small number of simultaneous links. Bluebird’s propulsion was chemical, optimized for initial orbit raising and limited station-keeping, and its stowed volume fit comfortably within standard medium-lift payload fairings. Starlink’s V2 Mini, although smaller in physical aperture, featured a more balanced and compact architecture. Weighing roughly 800 kilograms at launch, it was designed around high-throughput broadband rather than direct-to-cellular use. Its phased array antenna surface was closer to 20–25 m², and it was optimized for efficient manufacturing and high-density orbital deployment. The V2 Mini’s volume was tightly packed, with solar panels, phased arrays, and propulsion modules folded into a relatively low-profile bus optimized for rapid deployment and low-cost launch stacking. Its onboard compute and thermal systems were scaled to match its more modest power budget, which typically hovered around 2.5 to 3.0 kilowatts.

In contrast, today’s satellites occupy an entirely new performance regime. The dry mass of the satellite ranges between 2,500 and 3,500 kilograms, depending on specific configuration, thermal shielding, and structural deployment method. This accounts for its large deployable arrays, high-density digital payload, radiator surfaces, power regulation units, and internal trusses. The wet mass, including onboard fuel reserves for at least 5 years of station-keeping at 350 km altitude, increases by up to 800 kilograms, depending on the propulsion type (e.g., Hall-effect or gridded ion thrusters) and orbital inclination. This brings the total launch mass to approximately 3,000 to 4,500 kilograms, or more than double ATS’s old Bluebird 1 and roughly five times that of SpaceX’s Starlink V2 Mini.

Volume-wise, the satellites require a significantly larger stowed configuration than either AST’s Bluebird 1 or SpaceX’s Starlink V2 Mini. While both of those earlier systems were designed to fit within traditional launch fairings, Bluebird 1 utilizes a folded hinge-based boom structure, and Starlink V2 Mini is optimized for ultra-compact stacking. Today’s satellite demands next-generation fairing geometries, such as 5-meter-class launchers or dual-stack configurations. This is driven by the dual-antenna architecture and radiator arrays, which, although cleverly folded during launch, expand dramatically once deployed in orbit. In its operational configuration, the satellite spans tens of meters across its antenna booms and solar sails. The uplink array, built as a lightweight, mesh-backed surface supported by rigidizing frames or telescoping booms, unfolds to a diameter of approximately 30 to 35 meters, substantially larger than Bluebird 1’s ~20–25 meter maximum span and far beyond the roughly 10-meter unfolded span of Starlink V2 Mini. The downlink panels, although smaller, are arranged for precise gimballed orientation (i.e., a pivoting mechanism allowing rotation or tilt along one or more axes) and integrated thermal control, which further expands the total deployed volume envelope. The volumetric footprint of today’s D2C satellite is not only larger in surface area but also more spatially complex, as its segregated UL and DL arrays, thermal zones, and solar wings must avoid interference while maintaining structural and thermal equilibrium. Compared to the simplified flat-pack layout of Starlink V2 Mini and the monolithic boom-deployed design of Bluebird 1.

The increase in dry mass, wet mass, and deployed volume is not a byproduct of inefficiency, but a direct result of very substantial performance improvements that were required to replace terrestrial mobile towers with orbital systems. Today’s D2C satellites deliver an order of magnitude more beam concurrency, spectral efficiency, and per-user performance than its 2020s predecessors. This is reflected in every subsystem, from power generation and antenna design to propulsion, thermal control, and computing. As such, it represents the emergence of a new class of satellite altogether: not merely a space-based relay or broadband node, but a full-featured, cloud-integrated orbital RAN platform capable of supporting the global cellular fabric from space.

CAN THE FICTION BECOME A REALITY?

From the perspective of 2025, the vision of a global satellite-based mobile network providing seamless, unmodified indoor connectivity at terrestrial-grade uplink and downlink rates, 50 Mbps up, 500 Mbps down, appears extraordinarily ambitious. The technical description from 2045 outlines a constellation of 20,800 LEO satellites, each capable of supporting 5,000 independent full-duplex beams across massive bandwidths, while integrating onboard processing, AI-driven beam control, and a full 5G core stack. To reach such a mature architecture within two decades demands breakthrough progress across multiple fronts.

The most daunting challenge lies in achieving indoor-grade cellular uplink at frequencies as low as 600 MHz from devices never intended to communicate with satellites. Today, even powerful ground-based towers struggle to achieve sub-1 GHz uplink coverage inside urban buildings. For satellites at an altitude of 350 km, the free-space path loss alone at 600 MHz is approximately 133 dB. When combined with clutter, penetration, and polarization mismatches, the system must close a link budget approaching 153–160 dB, from a smartphone transmitting just 23 dBm (200 mW) or less. No satellite today, including AST SpaceMobile’s BlueBird 1, has demonstrated indoor uplink reception at this scale or consistency. To overcome this, the proposed system assumes deployable uplink arrays of 750 m² with gain levels exceeding 45 dBi, supported by hundreds of simultaneously steerable receive beams and ultra-low-noise front-end receivers. From a 2025 lens, the mechanical deployment of such arrays, their thermal stability, calibration, and mass management pose nontrivial risks. Today’s large phased arrays are still in their infancy in space, and adaptive beam tracking from fast-moving LEO platforms remains unproven at the required scale and beam density.

Thermal constraints are also vastly more complex than anything currently deployed. Supporting 5,000 simultaneous beams and radiating tens of kilowatts from compact platforms in LEO requires heat rejection systems that go beyond current radiator technology. Passive radiators must be supplemented with phase-change materials, active fluid loops, and zoned thermal isolation to prevent transmit arrays from degrading the performance of sensitive uplink receivers. This represents a significant leap from today’s satellites, such as Starlink V2 Mini (~3 kW) or BlueBird 1 (~10–12 kW), neither of which operates with a comparable beam count, throughput, or antenna scale.

The required onboard compute is another monumental leap. Running thousands of simultaneous digital beams, performing real-time adaptive beamforming, spectrum assignment, HARQ scheduling, and AI-driven interference mitigation, all on-orbit and without ground-side offloading, demands 100–500 TOPS of radiation-hardened compute. This is far beyond anything that will be flying in 2025. Even state-of-the-art military systems rely heavily on ground computing and centralized control. The 2045 vision implies on-orbit autonomy, local decision-making, and embedded 5G/6G core functionality within each spacecraft, a full software-defined network node in orbit. Realizing such a capability requires not only next-gen processors but also significant progress in space-grade AI inference, thermal packaging, and fault tolerance.

On the power front, generating 25–35 kW per satellite in LEO using 60–80 m² solar sails pushes the boundary of photovoltaic technology and array mechanics. High-efficiency solar cells must achieve conversion rates exceeding 30–35%, while battery systems must maintain high discharge capacity even in complete darkness. Space-based power architectures today are not yet built for this level of sustained output and thermal dissipation.

Even if the individual satellite challenges are solved, the constellation architecture presents another towering hurdle. Achieving seamless beam handover, full spatial reuse, and maintaining beam density over demand centers as the Earth rotates demands near-perfect coordination of tens of thousands of satellites across hundreds of planes. No current LEO operator (including SpaceX) manages a constellation of that complexity, beam concurrency, or spatial density. Furthermore, scaling the manufacturing, testing, launch, and in-orbit commissioning of over 20,000 high-performance satellites will require significant cost reductions, increased factory throughput, and new levels of autonomous deployment.

Regulatory and spectrum allocation are equally formidable barriers. The vision entails the massively complex undertaking of a global reallocation of terrestrial mobile spectrum, particularly in the sub-3 GHz bands, to LEO operators. As of 2025, such a reallocation is politically and commercially fraught, with entrenched mobile operators and national regulators unlikely to cede prime bands without extensive negotiation, incentives, and global coordination. The use of 600–1800 MHz from orbit for direct-to-device is not yet globally harmonized (and may never be), and existing terrestrial rights would need to be either vacated or managed via complex sharing schemes.

From a market perspective, widespread device compatibility without modification implies that standard mobile chipsets, RF chains, and antennas evolve to handle Doppler compensation, extended RTT timing budgets, and tighter synchronization tolerances. While this is not insurmountable, it requires updates to 3GPP standards, baseband silicon, and potentially network registration logic, all of which must be implemented without degrading terrestrial service. Although NTN (non-terrestrial networks) support has begun to emerge in 5G standards, the level of transparency and ubiquity envisioned in 2045 is not yet backed by practical deployments.

While the 2045 architecture described so far assumes a single unified constellation delivering seamless global cellular service from orbit, the political and commercial realities of space infrastructure in 2025 strongly suggest a fragmented outcome. It is unlikely that a single actor, public or private, will be permitted, let alone able, to monopolize the global D2C landscape. Instead, the most plausible trajectory is a competitive and geopolitically segmented orbital environment, with at least one major constellation originating from China (note: I think it is quit likely we may see two major ones), another from the United States, a possible second US-based entrant, and potentially a European-led system aimed at securing sovereign connectivity across the continent. This fracturing of the orbital mobile landscape imposes a profound constraint on the economic and technical scalability of the system. The assumption that a single constellation could achieve massive economies of scale, producing, launching, and managing tens of thousands of high-performance satellites with uniform coverage obligations, begins to collapse under the weight of geopolitical segmentation. Each competitor must now shoulder its own development, manufacturing, and deployment costs, with limited ability to amortize those investments over a unified global user base. Moreover, such duplication of infrastructure risks saturating orbital slots and spectrum allocations, while reducing the density advantage that a unified system would otherwise enjoy. Instead of concentrating thousands of active beams over a demand zone with a single coordinated fleet, separate constellations must compete for orbital visibility and spectral access over the same urban centers. The result is likely to be a decline in per-satellite utilization efficiency, particularly in regions of geopolitical overlap or contested regulatory coordination.

Finally, the commercial viability of any one constellation diminishes when the global scale is eroded. While a monopoly or globally dominant operator could achieve lower per-unit satellite costs, higher average utilization, and broader roaming revenues, a fractured environment reduces ARPU (average revenue per user). It increases the breakeven threshold for each deployment. Satellite throughput that could have been centrally optimized now risks duplication and redundancy, increasing operational overhead and potentially slowing innovation as vendors attempt to differentiate on proprietary terms. In this light, the architecture described earlier must be seen as an idealized vision. This convergence point may never be achieved in pure form unless global policy, spectrum governance, and commercial alliances move toward more integrated outcomes. While the technological challenges of the 2045 D2C system are significant, the fragmentation of market structure and geopolitical alignment may prove an equally formidable barrier to realizing the full systemic potential. While a monopoly or globally dominant operator could achieve lower per-unit satellite costs, higher average utilization, and broader roaming revenues, a fractured environment reduces ARPU (average revenue per user). It increases the breakeven threshold for each deployment. Satellite throughput that could have been centrally optimized now risks duplication and redundancy, increasing operational overhead and potentially slowing innovation as vendors attempt to differentiate on proprietary terms. In this light, the architecture described earlier must be seen as an idealized vision. This convergence point may never be achieved in pure form unless global policy, spectrum governance, and commercial alliances move toward more integrated outcomes. While the technological challenges of the 2045 D2C system are significant, the fragmentation of market structure and geopolitical alignment may prove an equally formidable barrier to realizing the full systemic potential.

Despite these barriers, incremental paths forward exist. Demonstration satellites in the late 2020s, followed by regional commercial deployments in the early 2030s, could provide real-world validation. The phased evolution of spectrum use, dual-use handsets, and AI-assisted beam management may mitigate some of the scaling concerns. Regulatory alignment may emerge as rural and unserved regions increasingly depend on space-based access. Ultimately, the achievement of the 2045 architecture relies not only on engineering but also on sustained cross-industry coordination, geopolitical alignment, and commercial viability on a planetary scale. As of 2025, the probability of realizing the complete vision by 2045, in terms of indoor-grade, direct-to-device service via a fully orbital mobile core, is perhaps 40–50%, with a higher probability (~70%) for achieving outdoor-grade or partially integrated hybrid services. The coming decade will reveal whether the industry can fully solve the unique combination of thermal, RF, computational, regulatory, and manufacturing challenges required to replace the terrestrial mobile network with orbital infrastructure.

POSTSCRIPT – THE ECONOMICS.

The Direct-to-Cellular satellite architecture described in this article would reshape not only the technical landscape of mobile communications but also its economic foundation. The very premise of delivering mobile broadband directly from space, bypassing terrestrial towers, fiber backhaul, and urban permitting, undermines one of the most entrenched capital systems of the 20th and early 21st centuries: the mobile infrastructure economy. Once considered irreplaceable, the sprawling ecosystem of rooftop leases, steel towers, field operations, base stations, and fiber rings has been gradually rendered obsolete by a network that floats above geography.

The financial implications of such a shift are enormous. Before such an orbital transition described in this article, the global mobile industry invested well over 300 billion USD annually in network CapEx and Opex, with a large share dedicated to the site infrastructure layer, construction, leasing, energy, security, and upkeep of millions of base stations and their associated land or rooftop assets. Tower companies alone have become multi-billion-dollar REITs (i.e., Real Estate Investment Trusts), profiting from site tenancy and long-term operating contracts. As of the mid-2020s, the global value tied up in the telecom industry’s physical infrastructure is estimated to exceed 2.5 to 3 trillion USD, with tower companies like Cellnex and American Tower collectively managing hundreds of billions of dollars in infrastructure assets. An estimated $300–500 billion USD invested in mobile infrastructure represents approximately 0.75% to 1.5% of total global pension assets and accounts for 15% to 30% of pension fund infrastructure investments. This real estate-based infrastructure model defined mobile economics for decades and has generally been regarded as a reasonably safe haven for investors. In contrast, the 2045 D2C model front-loads its capital burden into satellite manufacturing, launch, and orbital operations. Rather than being geographically bound, capital is concentrated into a fleet of orbital base stations, each capable of dynamically serving users across vast and shifting geographies. This not only eliminates the need for millions of distributed cell sites, but it also breaks the historical tie between infrastructure deployment and national geography. Coverage no longer scales with trenching crews or urban permitting delays but with orbital plane density and beamforming algorithms.

Yet, such a shift does not necessarily mean lower cost, only different economics. Launching and operating tens of thousands of advanced satellites, each capable of supporting thousands of beams and running onboard compute environments, still requires massive capital outlay and ongoing expenditures in space traffic management, spectrum coordination, ground gateways, and constellation replenishment. The difference lies in utilization and marginal reach. Where terrestrial infrastructure often struggles to achieve ROI in rural or low-income markets, orbital systems serve these zones as part of the same beam budget, with no new towers or trenches required.

Importantly, the 2045 model would likely collapse the mobile value chain. Instead of a multi-layered system of operators, tower owners, fiber wholesalers, and regional contractors, a vertically integrated satellite operator can now deliver the full stack of mobile service from orbit, owning the user relationship end-to-end. This disintermediation has significant implications for revenue distribution and regulatory control, and challenges legacy operators to either adapt or exit.

The scale of economic disruption mirrors the scale of technical ambition. This transformation could rewrite the very economics of connectivity. While the promise of seamless global coverage, zero tower density, and instant-on mobility is compelling, it may also signal the end of mobile telecom as a land-based utility.

If this little science fiction story comes true, and there are many good and bad reasons to doubt it, Telcos may not Ascend to the Sky, but take the Stairway to Heaven.

FURTHER READING.

Kim K. Larsen, “Will LEO Satellite Direct-to-Cellular Networks Make Traditional Mobile Networks Obsolete?”, A John Strand Consult Report, (January 2025). This has also been published in full on my own Techneconomyblog.

Kim K. Larsen, “Can LEO Satellites close the Gigabit Gap of Europe’s Unconnectables?“ Techneconomyblog (April 2025).

Kim K. Larsen, “The Next Frontier: LEO Satellites for Internet Services.” Techneconomyblog (March 2024).

Kim K. Larsen, “Stratospheric Drones & Low Earth Satellites: Revolutionizing Terrestrial Rural Broadband from the Skies?” Techneconomyblog (January 2024).

Kim K. Larsen, “A Single Network Future“, Techneconomyblog (March 2024).

ACKNOWLEDGEMENT.

I would like to acknowledge my wife, Eva Varadi, for her unwavering support, patience, and understanding throughout the creative process of writing this article.