Orbital AI sounds like science fiction until it is placed beside the current numbers. Apple has more than 2.5 billion active devices. SpaceX operates the world’s largest low-Earth-orbit broadband constellation. Starlink satellites use optical inter-satellite links, with SpaceX describing three lasers per satellite operating at up to 200 Gbps. T-Mobile is already selling Starlink-powered satellite messaging for compatible phones. SpaceX has filed plans for a far larger satellite architecture tied to solar-powered AI data centers in orbit.
This is Part II of the Three-Part Story
Part I examined SpaceX as the orbital gatekeeper and why Apple may need to rethink its relationship with Elon Musk’s space company during the Ternus era. Part II looks at the next layer: xAI, orbital servers, solar-powered compute, Starlink routing, and Apple’s massive installed base of personal devices. Part III will explore the strategic choices Apple may face if this new stack becomes real.
The most aggressive SpaceX plan is no longer only broadband. SpaceX sought FCC approval for a constellation of up to 1 million solar-powered satellites engineered as orbital data centers for AI.
A separate analysis described the proposed system as a solar-powered, optical-link-driven orbital data-center network. The filing did not specify how many Starship launches would be required to scale the network to an operational degree, and the technical hurdles remain enormous. But the direction is clear: Musk wants SpaceX and xAI to use space not only for connectivity, but for compute.
The reasoning is easy to understand, even if execution remains uncertain. AI data centers on Earth face power limits, water use, land constraints, permitting delays, grid bottlenecks, and cooling requirements. Space offers constant solar access in the right orbital design, but introduces its own problems: launch cost, radiation, thermal management, maintenance, latency, orbital debris, hardware failure, chip replacement, compute density, and data downlink capacity. Jeff Bezos has said orbital data centers may eventually make sense but called Musk’s two-to-three-year timeline too optimistic.
That skepticism is warranted. Space-based AI is not guaranteed. But Apple cannot dismiss the possibility because the companies pursuing it are not small. SpaceX has launch cadence, Starlink, Starship ambitions, laser-link experience, satellite manufacturing, and xAI’s compute appetite. Apple has devices, silicon, users, privacy architecture, and personal AI ambitions. The gap between the two is becoming strategically important.
Apple’s Device Base Is a Planet-Scale AI Endpoint
Apple’s active installed base gives it a position no AI model company can easily copy. More than 2.5 billion active devices across iPhone, iPad, Mac, Apple Watch, Apple TV, Vision Pro, AirPods, and related products create a distributed personal-computing network. Each device contributes something different.
The iPhone is the daily sensor and communications hub. The Apple Watch adds health, motion, identity, and safety context. The Mac provides local compute, productivity, files, and creative work. The iPad supports education, entertainment, design, and mobility. Apple TV sits in the living room. AirPods handle voice, audio, and future health-adjacent possibilities. Vision Pro adds spatial perception, eye tracking, hand input, and immersive computing. Future home devices or robotics could add physical-world presence.
Apple Intelligence tries to coordinate this ecosystem through on-device models, Private Cloud Compute, Siri, App Intents, and personal context. Apple’s strength is not only model size. It is proximity to the user. The device sees the screen, stores the messages, receives the notifications, handles payments, knows the calendar, tracks health signals, and connects to home devices.
That is why orbital AI matters. If future AI compute can be delivered through space-based infrastructure, Apple devices would be among the most valuable endpoints. They are premium, global, trusted, sensor-rich, and already tied to user identity and payment. A space-based AI network without high-value devices is infrastructure without an intimate interface. Apple devices without enough AI infrastructure risk becoming dependent on outside providers.
The strategic opportunity would be a three-layer model: Apple devices as the personal edge, Apple Silicon as local intelligence, and space-based connectivity or compute as a remote layer when terrestrial infrastructure is absent, constrained, or strategically secondary.
Starlink’s Laser Network Is the Bridge
Orbital AI would need a data-routing layer, and Starlink’s optical inter-satellite links are the obvious bridge. SpaceX says each Starlink satellite includes three space lasers operating at up to 200 Gbps. Those lasers allow satellites to route data between each other before reaching ground stations, which can reduce dependence on local ground gateways and improve coverage over oceans, remote regions, polar areas, aircraft, ships, and areas with limited infrastructure.
This is not the same as saying Starlink can already serve massive real-time AI workloads directly to every phone. Direct-to-cell capacity remains limited. Satellite connections can be delayed, obstructed, or bandwidth-constrained. AI inference, especially multimodal real-time inference, demands low latency and high reliability. Orbital AI data centers would require far more power, cooling, compute density, and routing capacity than current Starlink internet traffic.
But Starlink’s architecture points toward a future where space is no longer only a relay to the ground. It becomes a mesh. If satellites can route, compute, and communicate with each other, the network begins to resemble a distributed infrastructure layer above the planet.
The scale numbers matter. Starlink already has more than 10,000 satellites in orbit, according to public industry counts. The FCC has approved SpaceX to reach 15,000 satellites under current authorizations, while broader SpaceX filings and Musk’s public ambitions point toward much larger constellations. T-Mobile’s direct-to-cell service uses more than 650 Starlink satellites for satellite-to-mobile connectivity.
The first practical result is communication. The future possibility is compute.
xAI Turns SpaceX Into More Than Connectivity
xAI changes the SpaceX story because it adds a direct AI infrastructure motive. Reuters has reported that Musk’s SpaceX and xAI strategy includes proposed space-based AI data centers powered by solar energy. Wired has also reported that SpaceX’s AI infrastructure ambitions include major data-center capacity tied to xAI’s Colossus facilities on Earth, with huge computing costs already shaping the company’s finances.
The economics of AI are becoming brutal. Training and inference require chips, power, cooling, fiber, land, and capital. If SpaceX can lower launch costs through Starship and use solar-powered orbital hardware for some kinds of AI workloads, it could create a new infrastructure category. Not every AI task belongs in orbit. Latency-sensitive tasks, high-throughput model training, and hardware maintenance may still favor terrestrial data centers for a long time. But some inference, caching, global routing, remote-area service, disaster-resilient compute, or edge-adjacent tasks could eventually be tested in space.
For Apple, this is where the risk becomes strategic. Apple is building a privacy-led AI model through Apple Intelligence and Private Cloud Compute. But Apple does not currently own a satellite network, launch system, or orbital compute layer. If SpaceX/xAI becomes the first company to combine orbital connectivity with AI compute, it could gain leverage over future personal AI distribution.
That does not mean Apple should outsource Apple Intelligence to xAI. The cultural, privacy, governance, and brand risks would be substantial. It does mean Apple needs to understand whether SpaceX’s infrastructure could become a necessary part of global device connectivity and AI availability.
The Apple-to-SpaceX Correlation
The correlation between Apple and SpaceX is not obvious until the user is placed at the center. Apple’s 2.5 billion devices generate local demand for communication, search, personal assistance, maps, media, safety, translation, health, payments, identity, and increasingly AI. SpaceX’s satellites generate global coverage, backhaul, direct-to-cell messaging, mobility connectivity, maritime internet, aviation internet, and future orbital compute possibilities.
One side owns the endpoint. The other side owns the sky path.
A future iPhone could rely on local Apple Intelligence for private tasks, Private Cloud Compute for complex personal requests, carrier 5G for ordinary connectivity, Wi-Fi for home and work, and satellite connectivity for remote areas. If orbital AI matures, another layer could appear: compute or inference routed through a space-based network when it is cheaper, available, or strategically preferred.
The personal-device metric makes this important. If Apple’s 2.5 billion active devices each become more AI-dependent, even a small share of requests shifting to remote inference creates enormous infrastructure demand. A simple hypothetical shows the scale: if only 10% of Apple’s active devices made one cloud-assisted AI request per day, that would be 250 million requests daily. If the average were five requests, the number would become 1.25 billion daily. Apple will not route all of that through space, but the exercise shows why AI infrastructure and device scale are now inseparable.
Apple’s device base is also decentralized. A single user may have multiple Apple devices that could cooperate locally: iPhone for sensors, Watch for health, Mac for compute, AirPods for voice, Vision Pro for spatial input, Apple TV for display, and HomePod for room audio. Apple could eventually use this personal cluster to reduce some cloud dependence. But larger models, live multimodal agents, global knowledge, translation, video generation, and app-level automation will still need remote infrastructure.
That is where orbital connectivity and compute become a potential strategic partner or threat.
The Hurdles Are Massive
Orbital AI should not be presented as inevitable. The obstacles are real. Launching a million satellites would create enormous regulatory, environmental, orbital-debris, astronomical, spectrum, and safety questions. Space-based data centers would need radiation-hardened hardware or rapid replacement cycles. AI accelerators generate heat, and cooling in space is difficult because there is no air to carry heat away. Maintenance is far harder than swapping servers in a terrestrial data center. Latency may work for some tasks but not others. Building, launching, operating, and replacing orbital compute hardware would be financially extreme.
The network also has to return results to Earth. AI compute in orbit only helps if input and output move efficiently. Direct-to-cell is useful for messaging today, but rich AI workloads need far more bandwidth. Starlink terminals, ground stations, laser links, and future mobile satellite links would all need to scale.
This is why Bezos’s caution is credible. Space data centers may be economically attractive someday, but the timeline is uncertain. Terrestrial data centers will remain dominant for the near future because they are easier to build, maintain, cool, connect, and upgrade.
The point for Apple is not that orbital AI is guaranteed in 2028. The point is that SpaceX is one of the only companies with the launch capacity, satellite network, and AI motive to attempt it seriously. Apple builds strategy years before categories mature. If SpaceX is even partially successful, Apple needs to know where it stands.
The Stakes Move Beyond Satellite Messaging
This story leads to a larger conclusion. Satellite messaging is only the first visible feature. The deeper shift is the potential merger of connectivity and AI infrastructure above the device layer. SpaceX wants to own more of orbit. xAI wants compute. Starlink provides a live network. Starship aims to raise launch capacity. Apple owns the most valuable personal-device base.
Those pieces may define a new tech-titan moment: devices, personal ecosystem, orbital connectivity, and AI compute. Apple can remain independent in many layers, but it cannot pretend the sky layer is irrelevant. The next decade of iPhone, Apple Watch, Mac, Vision Pro, Home, and robotics may depend on coverage and compute that extend beyond terrestrial carriers and data centers.
Part III will examine Apple’s strategic choices. A partnership with SpaceX could create powerful iPhone and Apple Intelligence possibilities. A cautious multi-partner strategy could reduce dependence on Musk. Building Apple’s own satellite layer would be expensive and slow. Ignoring the orbital shift could leave Apple with the world’s best personal devices attached to someone else’s next-generation infrastructure.
